当前位置:首页 >> 能源/化工 >>

HYDROGEN INTEGRATION IN OIL REFINERIES


HYDROGEN INTEGRATION IN OIL REFINERIES

A thesis submitted to the

UNIVERSITY OF MANCHESTER INSTITUTE OF SCIENCE AND TECHNOLOGY

for the degree of

Doc

tor of Philosophy
by

Fang Liu

under the supervision of

Dr. Nan Zhang

Department of Process Integration October 2002

DECLARATION

I declare that no portion of the work referred to in this thesis has been submitted in support of an application for another degree or qualification of this or any other university or other institute of learning.

Fang Liu

i

ACKNOWLEDGEMENTS

Firstly of all, I would like to thank my supervisor, Dr. Nan Zhang for his instruction, advice and help during my time in UMIST. Without his support, this thesis would not be possible. I would like to express my gratitude to the Department of Process Integration, UMIST and the member of Process Integration Research Consortium for offering me the opportunity to carry out this research and providing the financial support. Many thanks to Dr. Nick Hallale for his assistance and valuable input to this research. Those contributions can be found in Chapter 1 and 2. I am grateful to my colleagues and the staff in the department for their companionship, inspiration and always lending me a hand. Thanks to all my friends in Manchester for their help and friendship which I have been fortunate to have. Special thanks to my parents and brother for their love and support throughout my life. Finally, my thanks should go to my beloved wife Ying. Thank you for your understanding and encouragement, for always being my side through all those difficult times and for the happiness we will share in the future.

ii

ABSTRACT

Several trends in the petroleum industry are leading to the increased demand for hydrogen in oil refineries caused by the tighter environmental regulations and heavy-end upgrading. Previous work developed a graphical method for the analysis of hydrogen distribution systems. Targets are set for hydrogen recovery and hydrogen plant production. However, the approach has neglected pressure constraints leading to an unrealistic solution. In this thesis, an automated approach has been developed to include pressure issues in the design. The method is based on the optimisation of a reducible superstructure. Multiple practical constraints can be implemented to achieve optimal, realistic designs. Retrofit options (for example, additional compression, purification and piping changes) are decided automatically through optimisation. A methodology is also proposed to select appropriate purifiers from pressure swing adsorption processes and membranes or hybrid systems for recovering hydrogen from refinery off-gases. Through the understanding of the trade-offs between hydrogen saving, compression costs and capital investment, a superstructure is built to include possible purification scenarios. The shortcut models for different purification units are developed via the insights of the processes. The recovery rate of purifiers is also modelled to optimise process parameters. This method achieves the optimal design for overall hydrogen networks at a conceptual level. To incorporate hydrogen generation into hydrogen networks, a method for integrating hydrogen generation and hydrogen recovery is developed. The hydrogen plant is modelled by means of correlating process data from compressive process simulation. The hydrogen plant model covers a wide feed range from natural gas, refinery off-gas hydrocarbon to light naphtha. A superstructure is then built to account for the integration of hydrogen plants and purifier operations. The refinery off-gases are evaluated as the possible feed to both hydrogen plants and purification units. The method is applied to find the optimum operational solution for the integration of hydrogen generation and purification. Extensive solution procedures are developed to overcome the computational difficulties and help to find optima. Case studies demonstrate the effectiveness of these methodologies.

iii

TABLE OF CONTENTS DECLARATION........................................................................................................I ACKNOWLEDGEMENTS .......................................................................................II ABSTRACT ............................................................................................................III
TABLE OF CONTENTS………………………………………………………………………….IV LIST OF FIGURES…….…………………………………………………………………………VII LIST OF TABLES………………………………………………………………………………….X

1

INTRODUCTION ..............................................................................................1

1.1 Background........................................................................................................................... 1 1.1.1 Briefing oil refinery ......................................................................................................... 1 1.1.2 Trends in modern refinery and impacts on hydrogen balance ...................................... 2 1.1.3 Hydrogen consumers and sources in refinery ............................................................... 5 1.2 Former research – hydrogen pinch analysis..................................................................... 9 1.2.1 Sink and source location................................................................................................ 9 1.2.2 Hydrogen composite curve and hydrogen surplus curve ............................................ 11 1.2.3 Purity trade-off and purification analysis...................................................................... 12 1.2.4 Linear network design.................................................................................................. 14 1.2.5 Limitation of Hydrogen Pinch approach....................................................................... 14 1.3 Approaches developed in this thesis for hydrogen integration.................................... 15

2 HYDROGEN NETWORK OPTIMISATION WITH PRESSURE CONSIDERATION.................................................................................................18
2.1 Match sources and sinks in hydrogen distribution networks ....................................... 18

2.2 Optimal hydrogen allocation with pressure constraints................................................ 20 2.2.1 Network re-routing without interception ....................................................................... 20 2.2.2 Source interception ...................................................................................................... 26 2.3 Automated design for hydrogen distribution system .................................................... 30 2.3.1 Compressor model.......................................................................................................31 2.3.2 Piping cost ................................................................................................................... 33 2.3.3 PSA capital cost........................................................................................................... 35 2.3.4 Optimisation strategy ................................................................................................... 36 2.4 Case studies ....................................................................................................................... 38 2.4.1 Case 1 – retrofit to save operating costs ..................................................................... 39 2.4.2 Case 2 – retrofit for debottlenecking............................................................................ 41 2.5 Conclusion .......................................................................................................................... 44

iv

3 STRATEGY OF PURIFIER SELECTION AND INTEGRATION IN HYDROGEN NETWORK.......................................................................................46
3.1 Introduction......................................................................................................................... 46

3.2 Purification process analysis............................................................................................ 47 3.2.1 Pressure swing adsorption (PSA)................................................................................ 47 3.2.2 Membrane.................................................................................................................... 51 3.2.3 Cryogenic separation ................................................................................................... 53 3.2.4 Hybrid system .............................................................................................................. 55 3.3 Review of existing methods for the selection and integration of purification processes ......................................................................................................................................... 56 3.3.1 Industrial approach ...................................................................................................... 56 3.3.2 Other research for integration of purification process.................................................. 58 3.3.3 New rules for the hydrogen pinch method ................................................................... 59 3.4 MINLP model for purifier selection and integration........................................................ 62 3.4.1 Superstructure ............................................................................................................. 62 3.4.2 Mass balance model for hydrogen consumers............................................................ 64 3.4.3 Compressor model ....................................................................................................... 65 3.4.4 PSA shortcut model .....................................................................................................67 3.4.5 Membrane model ......................................................................................................... 70 3.4.6 Piping model and pressure constraints........................................................................ 72 3.4.7 Objective function ........................................................................................................73 3.5 Solution procedure............................................................................................................. 74 3.5.1 Mixed integer and nonlinear features in the superstructure model ............................. 74 3.5.2 Linear relaxation of nonlinear terms ............................................................................ 75 3.5.3 Solution procedure....................................................................................................... 77 3.6 Case studies and analysis................................................................................................. 79 3.6.1 Purification process integration for retrofit ................................................................... 79 3.6.2 Selection and optimisation of purification process....................................................... 83 3.7 Conclusions ........................................................................................................................ 90

4 INTEGRATION OF HYDROGEN GENERATION AND HYDROGEN RECOVERY...........................................................................................................91
4.1 Introduction......................................................................................................................... 91

4.2 Hydrogen plant process introduction .............................................................................. 92 4.2.1 Available feedstock in refinery and pre-treatment ....................................................... 94 4.2.2 Steam reforming .......................................................................................................... 95 4.2.3 Reformed gas shifting .................................................................................................. 95 4.2.4 PSA and fuel system.................................................................................................... 96 4.2.5 Steam generation system ............................................................................................ 96 4.2.6 New trends in process design of steam reforming ...................................................... 96 4.3 Hydrogen plant process modelling .................................................................................. 97 4.3.1 Previous research ........................................................................................................ 97 4.3.2 Process simulation for hydrogen plant......................................................................... 98 4.3.3 Linear model for hydrogen plant ................................................................................103 4.3.4 Model verification .......................................................................................................109 4.4 Integration of hydrogen plant and PSA..........................................................................111 4.4.1 Building superstructure ..............................................................................................111

v

4.4.2 4.4.3 4.4.4 4.4.5 4.5 4.6

Nonlinear model.........................................................................................................115 Mixed integer linear model.........................................................................................118 Objective function ......................................................................................................122 Solution procedure.....................................................................................................123 Case study and analysis..................................................................................................126 Conclusion ........................................................................................................................132

5
5.1 5.2

CONCLUSION AND FUTURE WORK .........................................................133
Conclusion ........................................................................................................................133 Future work .......................................................................................................................135

REFERENCES ....................................................................................................137 NOMENCLATURE ..............................................................................................142 APPENDIX ..........................................................................................................150

vi

LIST OF FIGURES

Figure 1.1 Oil refinery processing sequence.............................................................................. 2 Figure 1.2 Mild hydrocracking unit flow diagram........................................................................ 6 Figure 1.3 World and US’s hydrogen capacity and planned capacity in the future.................... 8 Figure 1.4 Simplified diagram of a hydrogen consumer showing source and sink locations ................................................................................................................ 9 Figure 1.5 Hydrogen composite curve ..................................................................................... 10 Figure 1.6 Hydrogen surplus diagram...................................................................................... 11 Figure 1.7 Targeting the minimum utility by varying the hydrogen utility flowrate until a pinch is formed. ................................................................................................... 12 Figure 1.8 Effect of utility purity increment on hydrogen surplus curve.................................... 13 Figure 1.9 Evaluate purification scenario using hydrogen surplus curve ................................. 14 Figure 2.1 An example of hydrogen distribution networks ....................................................... 18 Figure 2.2 Decomposing hydrogen consumers into sinks and sources................................... 19 Figure 2.3 Final network configuration after stream match ...................................................... 20 Figure 2.4 An existing hydrogen network with pressure constraints ........................................ 21 Figure 2.5 Example of reduced superstructure ........................................................................ 22 Figure 2.6 Decomposing compressor into sink and source ..................................................... 24 Figure 2.7 The minimum hydrogen utility achievable with the existing compressor is 195.9 Mscfd......................................................................................................... 25 Figure 2.8 Sensitivity analysis showing that the bottleneck is the capacity of the make-up compressor to unit B ............................................................................ 25 Figure 2.9 If a new compressor is installed, the minimum utility target of 182.9 Mscfd can be met. ......................................................................................................... 27 Figure 2.10 A purifier in the superstructure of hydrogen networks .......................................... 28 Figure 2.11 Compressor capital cost vs. power ....................................................................... 32 Figure 2.12 Linear approximation of piping cost ...................................................................... 34 Figure 2.13 Estimating piping lengths in refinery plot plan....................................................... 35 Figure 2.14 Solution strategy for MINLP in automatic network design .................................... 37 Figure 2.15 Base case – hydrogen distribution network in existing refinery ............................ 37 Figure 2.16 Solution for Case 1: minimising operating cost..................................................... 40 Figure 2.17 Solution for Case 1: minimising operating cost with capital limit of 5 MUS$ .................................................................................................................. 41

vii

Figure 2.18 Conventional way of accommodating the increased hydrogen demand in Case 2................................................................................................................. 43 Figure 2.19 Debottlenecking design for minimum TAC in Case 2 ........................................... 43 Figure 3.1 PSA cycle sequence chart ...................................................................................... 48 Figure 3.2 Effect of tail gas pressure on PSA recovery ........................................................... 49 Figure 3.3 Effect of feed pressure levels on PSA system recovery ......................................... 50 Figure 3.4 Effect of product purity on PSA system hydrogen recovery.................................... 50 Figure 3.5 Single-stage membrane process ............................................................................ 51 Figure 3.6 Relative cost vs purity for membrane system ......................................................... 52 Figure 3.7 Hydrogen recovery vs purity for membrane system ............................................... 53 Figure 3.8 Partial condensation cryogenic process ................................................................. 54 Figure 3.9 Trade-offs in purification process selection............................................................. 62 Figure 3.10 Superstructure to select and integrate hydrogen purification processes .............. 63 Figure 3.11 The comparison of compressor power calculation using (2.14) and (3.20) .......... 65 Figure 3.12 PSA system - Recovery vs PH/PL (feed purity = 70%) .......................................... 68 Figure 3.13 PSA system - Recovery vs feed purity (PH/PL = 20) ............................................. 69 Figure 3.14 Solution procedure for superstructure MINLP model............................................ 78 Figure 3.15 Existing hydrogen network with PSA unit ............................................................. 78 Figure 3.16 New design with the integration of membrane and PSA....................................... 80 Figure 3.17 Sensitivity analysis for membrane purification capacity........................................ 83 Figure 3.18 Petro-Canada base case ..................................................................................... 83 Figure 3.19 Optimum network design with membrane unit...................................................... 86 Figure 3.20 Competitive design scenarios for purification process design .............................. 87 Figure 3.21 Improved design by merging compressors ........................................................... 89 Figure 4.1 Typical steam reforming hydrogen plant................................................................. 93 Figure 4.2 Simulation structure for hydrogen plant .................................................................. 99 Figure 4.3 Procedure for makeup fuel and flue gas calculation ............................................. 102 Figure 4.4 Hydrogen plant model interface in integration with hydrogen networks................ 104 Figure 4.5 Conventional position of hydrogen plant and purifier in hydrogen network of refinery .......................................................................................................... 112 Figure 4.6 Superstructure for integration of hydrogen plant and PSA ................................... 113 Figure 4.7 Mass balance model for PSA................................................................................ 116 Figure 4.8 Stream splitter....................................................................................................... 118

viii

Figure 4.9 Solution procedure for integration of hydrogen plants and purification units ........ 125 Figure 4.10 Base case – integration of hydrogen plant and purification processes (Flowrate: Nm3/h) .............................................................................................. 126 Figure 4.11 Superstructure – covering possible operating schemes ..................................... 128 Figure 4.12 Optimum scheme directly given by optimisation (Nm3)....................................... 129 Figure 4.13 Optimum scheme after evolution (Nm3) .............................................................. 130 Figure 4.14 Optimum solution for maximum hydrogen generation (Nm3) .............................. 131

ix

LIST OF TABLES

Table 1.1 Typical hydrogen consumption data .......................................................................... 4 Table 1.2 Typical hydrogen production data .............................................................................. 7 Table 2.1 Process data in base case ....................................................................................... 38 Table 2.2 Piping distance between units (m) ........................................................................... 39 Table 2.3 Cost breakdown – minimum operating cost ............................................................. 40 Table 2.4 Cost breakdown – minimum operating cost with capital limit of 5 MUS$................. 42 Table 2.5 Cost breakdown – conventional debottlenecking solution of Case 2 ....................... 42 Table 2.6 Cost breakdown – improved debottlenecking solution of Case 2 ............................ 44 Table 3.1 Typical content of some refinery off-gases .............................................................. 46 Table 3.2 Relative adsorptivity of typical components ............................................................. 49 Table 3.3 Relative permeability of typical components ............................................................ 52 Table 3.4 Selection guide for hydrogen purification process ................................................... 57 Table 3.5 Process data for existing hydrogen network with PSA unit...................................... 79 Table 3.6 TAC comparison for existing network and optimum design ..................................... 81 Table 3.7 Process data of sinks and sources .......................................................................... 84 Table 3.8 Distances between sources and sinks (m)............................................................... 85 Table 3.9 DICOPT executive results to converge .................................................................... 86 Table 3.10 TAC comparison for Petro-Canada case study...................................................... 88 Table 4.1 Simulation result of the steam reforming plus HT-shifting...................................... 106 Table 4.2 Heat of combustion for fuel components................................................................ 108 Table 4.3 Candidate feedstock composition for model verification ........................................ 110 Table 4.4 Comparison of linear model and comprehensive simulation.................................. 110 Table 4.5 Process data for hydrogen consumers .................................................................. 127 Table 4.6 Process data for hydrogen consumers .................................................................. 130

x

Chapter 1

Introduction

Chapter 1: Introduction
1.1. Background
1.1.1. Briefing oil refinery
The goal of typical oil refineries is to convert as much as possible of the barrel of crude oil into transportation fuel. Although there are thousands of products made from refineries and related petrochemical plants, the major profitable yields remain in transport fuel such as motor gasoline, diesel and jet fuel. Refineries can sometime produce specific oil with higher price margin such as lubricating oils, however the amount is generally less than five percent of the total crude oil. The operations in refineries are sophisticated. Figure 1.1 shows an example of processing sequence for a fuel type refinery. Crude oil is heated and charged to an atmospheric distillation unit (ADU), where it is separated into wet gas, light straight run naphtha, heavy naphtha, kerosene, atmospheric gas oil and atmospheric residue. Heavy naphtha from the ADU is hydrotreated and sent to a continuous catalytic reformer (CCR) to improve octane number. Kerosene is hydrotreated in a kerosene hydrotreator (KHT) to produce jet fuel. Atmospheric middle distillate is hydrodesulphurised and blended into a diesel pool. Atmospheric residue is further separated into vacuum gas oil (VGO) and vacuum residue (VR). VGO can be used as feedstock of a fluid catalytic cracker (FCC) or a hydrocracker (HCU), where the heavy petroleum molecules are cracked into lower molecular weight compounds within the boiling range of gasoline and distillate fuel ranges. VR is treated in a solvent deasphalting unit, where deasphalted oil (DAO) with lower sulphur and metal is lifted and used as FCC feedstock. The pitch and FCC slurry are sent to a delayed coker (DCU). The heavy coker gas oil can feed FCC. FCC gasoline is hydrodesulphurised and blended into gasoline pool. Middle distillate from FCC and DCU is hydrotreated and blended into the diesel pool. Coker naphtha is

1

Chapter 1

Introduction

Gas and LP G Refinery fuel gas LSR Napht ha LPG NHT Crude A D U CCR Reformate Alkylation CGHT KHT DHT FCC V D U VGO VR Deas phalted oil Slurry Solvent Extraction HCU Coker gas oil Coker diesel DCU LCOHT Alkylate Cracked Gasoline Jet fuel Diesel Diesel Petrochemical naphtha Gasoline Jet fuel Diesel Lube basis oil Sulphur

Gas,LPG, Naphtha and middle distillates CNHT Gas,LPG Coke

Coker naphtha

Figure 1.1 Oil refinery processing sequence hydrotreated and used as petrochemical naphtha or the CCR feedstock. The DCU yields petrochemical coke as one of its products. An alkylation unit produces high-octane alkylate that can be blended into gasoline range and improve the octane number of gasoline pool. Wet gas generated from FCC, DCU, HCU and other units are separated and light naphtha is stabilised to produce refinery fuel gas and liquid petroleum gas (LPG). After treating and blending, the refinery can produce gasoline, jet fuel, diesel and lubricating base oil etc. The crude processing structure can be different according to crude properties and the situation of the fuel market. Gary and Handwerk (1994) explained the general structure of modern refineries in the same order in which the crude flows through the refinery. Meyers (1997) collected the information of well-proven petroleum refining processes and provides more technology details.

1.1.2. Trends in modern refinery and impacts on hydrogen balance
The concern of hydrogen balance is increasing in the last fifteen years. Haun et al (1990) recollected the history of the hydrogen issue in refining. During the early age of oil refining, refiners paid little attention to hydrogen production, consumption or 2

Chapter 1

Introduction

distribution among the products. By the middle-50’s, after the prevalent employment of catalytic reforming, refiners had a cheap hydrogen resource which can be used as a reagent in their refining schemes. The hydrogen supply generally exceeded the demand. Although hydrogen became more and more important in lifting process performance, hydrogen management was of little importance in refineries. The hydrogen availability becomes a focal point because the refiners are facing challenges of stringent environmental regulations and increasing demand of transport fuel. Along with the legislation of environment protection, tougher gasoline and diesel quality specifications in the European Union and USA have been implemented to reduce smog-forming and other pollutants in the automotive exhausts. Such trend can be foreseen clearly in the nearly future. For example, the gasoline fuel specifications in the European Union will decrease the maximum sulphur from current 150ppm to 50ppm before 2005, and the maximum aromatics from 42%vol to 35%, while the maximum sulphur in diesel fuel specifications will decrease from current 350ppm to 50ppm. To stay in business, one of the options for refiners is to switch their feedstock to light sweet crude if they believe such supply is ample and under decent price in the future. At the same time, they have to look at significant investment in desulphurisation. Lower sulphur fuel means more hydrogen is necessary for deeper hydrodesulphurisation. In the mean time, lower aromatic gasoline specification will decrease the operation severity in catalytic reformers leading to the reduction of the by-product hydrogen. Knott (1998) reckoned that in Europe and US, the rate of hydrogen available from refinery by-products and recovery will decrease from 75% to 35% while the hydrogen demand will increase from 10 MNm3/h to 15 MNm3/h from 1992 to around 2000. Another future environmental impact is that the legislation of greenhouse gas abatement may urge refineries to reduce hydrogen production. Other big impacts on the refinery hydrogen balance are caused by bottom-of-barrel upgrade. According to more and more strict limitation of pollutant emissions, the fuel oil market has been in decline for a long period. On the other hand, the market trends indicate a very large increase in the share of middle distillates, reflecting spectacular growth in diesel oil and jet fuels production (Aitani and Ali, 1995). While crude is expected to become heavier and contain more sulphur, the refiners are pressured to add more conversion capacity to maintain competitive. There are two different kinds of 3

Chapter 1

Introduction

processes to increase the hydrogen-to-carbon ratio in the refinery product slate. One rejects carbon from petroleum streams such as fluid catalytic cracking and delayed coking. The other one adds hydrogen such as residue hydrotreating and hydrocracking. Although the latter is much more expensive in both capital and operating cost than the former, it is still predictable that hydrocracking processes will play the main role in the heavy-end conversion because of its considerable flexibility of the feedstock, processability of the yields and quality of the products. Especially when a refinery processes sour feedstock, the poor product quality after “carbon-reject” processes causes further headache on product blending and requires extra hydroprocessing. McGrath and Houde (1999) described the availabe processes for heavy crude upgrading. These processes are compatible and can be integrated to achieve the optimal economic results. Generally the additional “on-purpose” hydrogen capacity is required if the hydrocracking process is implemented. Many efforts have been contributed to analyse how these impacts affect hydrogen

Table 1.1 Typical hydrogen consumption data Chemical consumption only Process HT Str. Run Naptha HT FCC/TC Naphtha HT Kerosene HDS LS Gasoline to 0.2% S HDS HS Gasoline to 0.2% S HDS LS Gasoline to 0.05% S HDS HS Gasoline to 0.05% S HDS FCC/TC Gasoline Cycle oils hydrogenation Hydrocracking VGO Deep residue conversion %wt of feed 0.05 1 0.1 0.1 0.3 0.15 0.35 1 3 2-3 2-3.5 %wt of crude 0.01 0.05-1 0.01-0.02 0.03 0.04 0.04 0.05 0.1 0.3 0.5-0.8 1-2 Lamber et al (1994) 4

Chapter 1

Introduction

balance in refineries. Haun et al (1990) looked at the amount of hydrogen present in the feedstock versus the amount in the desired product slate as the ultimate determinant of hydrogen balance problem. The impact of the refinery evolution from hydroskimming to complex conversion is illustrated by comparing the hydrogen content in feeds and products. Two upgraded hypothetical refineries – gasoline refinery and diesel refinery are investigated. The case studies show that modern refineries can not only achieve anticipated product specifications but also dramatically improve their profitability. However, the hydrogen requirement is also largely increased. For the gasoline refinery, the hydrogen from catalytic reforming can not supply sufficient hydrogen and hydrogen has to be recovered from other refinery off-gases to settle the hydrogen balance. For the diesel refinery, hydrogen has to be produced through a hydrogen plant. The hydrogen management becomes essential for profitable operations in upgraded refineries. Lamber et al (1994) analysed how the environmental impacts on product qualities and changes in product slates shifted the hydrogen balance in refineries. It can be seen from Table 1.1 that vacuum distillate and residue hydroprocessing give main incentive to hydrogen demand. Heavy end gasification is suggested as an attractive option complying with the hydrogen demand, residue disposal and clean energy requirement of refineries. Philips (1999) discussed the approach to find the optimal hydrogen scenarios from the engineering point of view. An example project shows that through hydrogen management, hydrogen production capacity is decreased, resulting not only in a reduction in capital and operating expense but also in a significant decrease in CO2 emissions. However, it has been pointed out that any hydrogen solution must be tested not just for economic viability, the technical robustness, refinery integration and constructability should also be taken into account.

1.1.3. Hydrogen consumers and sources in refinery
1.1.3.1. Hydrogen consumers in refinery The main hydrogen consumption in refinery is from hydrocracking and hydrotreating units. There are other consumers such as lubricant units and paraffin isomerisation units, but they are not remarkable in fuel refineries. 5

Chapter 1

Introduction

The objective of hydrotreating is to remove undesired materials such as sulphur, nitrogen and metals, and saturate olefins and aromatics. Hydrotreating processes do not change the boiling range of their feedstock. Hydrocracking accomplishes the same objectives as hydrotreating, plus using hydrogen to break up the larger feedstock molecules and form lower-boiling products. The process therefore requires more severe conditions than hydrotreating, and consumes more hydrogen. Because the hydrogen partial pressure is one of the major design factors that affects capital cost and operating cost, a hydrocraker prefers high purity hydrogen. Typical hydrogen consumption data of hydroprocessing can be seen in Table 1.1. An example flow diagram of a mild hydrocracking unit is shown in Figure 1.2 The liquid feed is mixed with hydrogen-rich gas and heated. The reactions occur in a fixed bed reactor. The reactor effluent is cooled and sent to a gas-liquid separator. The gas from the separator is treated in an amine unit to remove hydrogen sulphide and a part of it is compressed and recycled. The rest of the hydrogen-rich gas is purged in order to prevent the build-up of contaminants in the recycle loop. The liquid product is further sent to a low-pressure separator where an off-gas stream is taken and typically sent to the fuel gas system.

1.1.3.2. Hydrogen sources in refinery There are several possible sources of hydrogen in refineries. Catalytic reforming units are typical sources of hydrogen. The aim of catalytic reforming is to produce aromatic
Make-up Hydrogen Recycle Hydrogen

Purge

Lean Amine H2S Removal Reactor Feed Water Rich Amine Gas

Sour Water Treated product

Figure 1.2 Mild hydrocracking unit flow diagram (Hiller et al, 1987) 6

Chapter 1

Introduction

compounds through the cyclisation and dehydrogenation of hydrocarbon molecules and is used to increase the octane number of naphtha. At the same time, large amounts of hydrogen are produced as a by-product. In modern refineries, catalytic reforming is operated as the largest source of hydrogen. The hydrogen-rich gas from a catalytic reformer can be used as make-up hydrogen to some hydrotreating units directly or sent to a hydrocracker after purification. However, with the implementation of restrictions on the aromatic contents in gasoline, the hydrogen available from catalytic reforming units is predicted to be decreased. Steam reforming and partial oxidation are on-purpose hydrogen generation processes in refineries. These hydrogen generating units use hydrocarbon as feedstock to produce high purity hydrogen. Lighter fractions, such as methane, refinery off-gas, hydrotreated LPG and light naphtha can be converted to hydrogen by steam reforming or partial oxidation. Heavier fractions, such as vacuum residue, refinery sludge and petroleum coke can only be processed into hydrogen by partial oxidation. Lambert et al (1994) summarised typical hydrogen production data as shown in Table 1.2. Many off-gas streams in refineries and petrochemical operations contain hydrogen, such as purge gas from hydrotreating and hydrocracking units, off-gas from delayed cokers, fluid catalytic crackers etc. The most efficient way to recover hydrogen from off-gases is the direct usage by hydrogen consumers. However, these streams

Table 1.2 Typical hydrogen production data Process Continuous Regeneration Reformer Semi-regeneration Reformer Residue Gasification Catalytic Cracking Thermal Cracking Ethylene Cracker Steam Reformer %wt of feed 0.05 1 0.1 0.1 0.3 0.15 0.35 %wt of crude 0.01 0.05-1 0.01-0.02 0.03 0.04 0.04 0.05 Lamber et al (1994) 7

Chapter 1

Introduction

sometimes have very low hydrogen purity or contain fatal impurities to hydroprocessing reactions so that they can not be utilised economically. Instead of releasing them to fuel gas head, purification units, such as pressure swing adsorption (PSA), membrane and cryogenic system can be employed to remove impurities and increase the hydrogen concentration. These recovery technologies are based on different separation theories. Hydrogen concentration, impurity characteristics and available pressure of off-gas candidates determine the selection criteria. More details will be discussed in Chapter 3. Hidalgo-Vivas and Towler (1997) pictured the spectrum of hydrogen facilities in refineries worldwide, and their possible incremental capacity in the future. Figure 1.3 shows that in the future, “on-purpose” hydrogen manufacturing processes including steam reforming and partial oxidation provide most of the hydrogen in refinery.

Me mb r a n e 1 5 % Cr y g e n i c 3 %

t Sp e c i f i e d Ot h e rNo 2 % 2 %

St e a m Me t h a n e Re f o r mi n g 3 3 %

Hy dr ogen Recov er y 39%

PSA 3 4 %

Pa r t i a l Ox i d a t i o n 2 % St e a m Na p t h a Re f o r mi n g 9 %

Hy dr ogen Pr oduct i on 61% W or l d Pl anned I ncr em ent alCapaci t y

W or l d Hydr ogen Capaci t y 1997

Me mbr an e 3 % Cr y ge n i c % P S A4 1 3 % S t e am Napt h a R e f o r mi n g 2 %

Ot h e r 2 %

Hy dr ogen Recov er y 22%

P ar t i al Ox i dat i o n 4 %

S t e am Me t h an e R e f o r mi n g 7 2 % US Hydr ogen Capaci t y 1997

Hy dr ogen Pr oduct i on 78%
US Planned Incremental Capacity

Figure 1.3 World and US’s hydrogen capacity and planned capacity in the future 8

Chapter 1

Introduction

1.2. Former research – hydrogen pinch analysis
Alves (1999) proposed a Pinch approach for targeting the minimum hydrogen utility. This work is based upon the pinch technology and exploits an analogy with heat exchanger network synthesis (Linnhoff, 1993). The method identifies sources and sinks of hydrogen, which are analogous to hot and cold streams in heat exchanger networks.

1.2.1. Sink and source location
The sinks of hydrogen correspond to the hydrogen demand of the various consumers. Each consumer needs a gas flowrate FSink and has a hydrogen purity requirement of ySink. The sources of hydrogen are the streams containing hydrogen, which can be used to feed the consumers. Hydrogen sources in a refinery include hydrogen streams from hydrogen plants, catalytic reformers or off-gases from the hydrogen consumers. The flowrate of a source is denoted as FSource and its purity is ySource. A typical hydrogen consumer including a hydrotreating reactor and a separator can be simplified as shown in Figure 1.4. Hydrogen is used to react with liquid hydrocarbon. The partial pressure in the reactor is a very important variable in the reaction. In this work, the partial pressure of hydrogen is assumed constant as well as other parameters such as operating temperature, reactor feedstock, products, etc. Under the above assumptions, a constant flowrate and hydrogen purity is imposed on the reactor gas inlet stream. Therefore, if the operating condition does not change, the inlet of the reactor and the outlet of the separator will be fixed. Here the mixture of the make-up

Make-up (FM,yM) Sink Liquid feed

Recycle (FR,yR)

Purge (FP,yP) Source Separator

Reactor

Liquid product

Figure 1.4 Simplified diagram of a hydrogen consumer showing source and sink locations 9

Chapter 1

Introduction

hydrogen and the recycle is defined as the sink and the mixture of the purge and the recycle is defined as the source. The sink and the source data can be determined from the make-up, the purge and the recycle data as follows: Fsink = FM + FR y sink = FM y M + FR y R FM + FR (1.1) (1.2) (1.3) (1.4)

Fsource = FP + FR
y source = y P = y R

where FM, FR and FP are the make-up, the recycle and the purge flowrate, and yM, yR and yP are the make-up, the recycle and the purge purity respectively.

1.2.2. Hydrogen composite curve and hydrogen surplus curve
The mass balance of each sink and source in a hydrogen distribution network can be conveniently represented in a two-dimensional plot with the flowrate of total gas on the

1 0.9 0.8 0.7

+

_

_ +

Purity (-)

0.6 0.5 0.4 0.3 0.2 0.1 0 0 50 100 150 200

250

300

Flowrate (MMscfd)

Figure 1.5 Hydrogen composite curve 10

Chapter 1

Introduction

horizontal axis and the purity on the vertical axis. Plotting the hydrogen demand profile and the hydrogen supply profile gives the hydrogen composite curves (Figure 1.5). This purity profile contains the hydrogen sinks and sources ordered by decreasing purity. Separately the sink and the source curves start at zero flowrate and continue until the lowest purity is represented. Where the hydrogen supply curve is above the hydrogen demand curve, the area between the two profiles is marked as surplus (+), which means the sources provide more hydrogen than required by the sinks; if the hydrogen supply is below the hydrogen demand curve, the area between the two profiles is marked as deficit (-), which means sources do not provide enough hydrogen to the sinks. The hydrogen composite curves can be divided into several regions with alternating surplus and deficit of hydrogen. Calculating these surpluses and deficits (area) of hydrogen, and plotting them against the purity level constructs the hydrogen surplus diagram (Figure 1.6). One of the necessary conditions for a feasible network is that there is no negative hydrogen surplus anywhere in the hydrogen surplus diagram, because if so the

1 0.9 0.8 0.7

Purity (-)

0.6 0.5 0.4 0.3 0.2 0.1 0 0 10 20 30 40 50

Hydrogen surplus (MMscfd)

Figure 1.6 Hydrogen surplus diagram 11

Chapter 1

Introduction

1 0.9 0.8 0.7

Purity (-)

0.6 0.5 0.4 0.3 0.2 0.1 0 0

Pinch

10

20

30

40

50

Hydrogen surplus (MMscfd)

Figure 1.7 Targeting the minimum utility by varying the hydrogen utility flowrate until a pinch is formed. sources can not provide enough hydrogen to the sinks. For an existing network, parts of the surplus curve are always positive. The hydrogen utility can be reduced through moving the curve towards the vertical axis until a vertical segment between the purity of the sink and the source overlaps with the zero axes (Figure 1.7). The purity at which this occurs is defined as the “hydrogen pinch” and is the theoretical bottleneck on how much hydrogen can be used from the sources to the sinks. The hydrogen utility flowrate that results in a pinch is the minimum target and is determined before any network design.

1.2.3. Purity trade-off and purification analysis
Alves (1999) then looked at hydrogen system analysis by using hydrogen pinch. One of the aspects is to reduce hydrogen utility flowrate by increasing the purity of one or more sources. It takes advantage of the fact that if two streams have the same flowrate of hydrogen, the one with higher purity will provide the hydrogen system with more hydrogen surplus. The resulting effect on the hydrogen surplus curve is shown in

12

Chapter 1

Introduction

Not pinched! Flowrate Hydrogen surplus

Figure 1.8 Effect of utility purity increment on hydrogen surplus curve Figure 1.8. The system initially pinched (dotted line) becomes unconstrained (solid line) with the increase in the utility purity. The additional hydrogen surplus thus created can be used to reduce the hydrogen utility, resulting in a lower target. This gives an option for debottlenecking the hydrogen distribution system. The purification of hydrogen sources is also analysed. The installation of a hydrogen purification unit adds one more sink and two sources to the hydrogen distribution system. The sink is the feedstock to purification. The sources are the purified product stream and the residue stream. The introduction of a new purification unit usually affects the entire hydrogen system, even if the unit is captive to an individual consumer process. The savings generated by the purification unit are assessed in the steps of placing the purification unit inside the network, applying the pinch method to find a new target. The multiple purification options can be evaluated one by one. Consequently, three possible placements for a purification unit in the hydrogen surplus curve are discussed: above the pinch, across the pinch or below the pinch. The general conclusions are then made to quantify different purification scenarios. It is found that purification across the pinch can reduce the requirement of the utility, and at the same time since the hydrogen loss happens below the pinch, the utility flowrate will not be affected. Their appearances are shown in Figure 1.9.

13

Chapter 1

Introduction

1

0.9

Possible reduction in utility Definite reduction in utility

0.8

0.7

0.6

0.5

No reduction in utility

0.4

0.3

0.2

0.1

0 0 2 4 6 8 10 12

Hydrogen surplus Figure 1.9 Evaluate purification scenario using hydrogen surplus curve

1.2.4. Linear network design
To achieve the target from the hydrogen pinch, a mathematical method using Linear Programming (LP) is employed to design hydrogen distribution networks. Mass balance for sinks and sources is imposed through equality constraints. The objective function is chosen to minimise the total cost in hydrogen networks, which includes the production cost of hydrogen from sources, the fuel credit to fuel systems and the hydrogen distribution cost between sources and sinks. To avoid the nonlinear formulation, the hydrogen distribution costs are simplified to be prorated to the flowrate with numerical parameters.

1.2.5. Limitation of Hydrogen Pinch approach
Hydrogen Pinch is a graphical approach to find the minimum hydrogen utility in distribution networks. It can provide insights to hydrogen distribution and is easy to access. However, it also has some drawbacks.

14

Chapter 1

Introduction

One of the major limitations with the method is that the targets are set based only on the flowrate and purity requirements and pressure is ignored. The targeting method assumes that any streams containing hydrogen can be sent to any consumers, regardless of the stream pressure. In reality, a source can only feed a sink if it is in a sufficient pressure level. Thus the targets generated may be too optimistic and unachievable in a real design. In addition, the targets from the hydrogen pinch can underestimate the maximum utility saving, because the potential installation of purification units is not considered during the targeting stage although it can further reduce the hydrogen utility usage. The analysis of the placement of purification units is processed on the basis of arbitrary selection. The Hydrogen Pinch can give an achievable target through the optimal network topology and essential compression before design. However, because the purification is also an important design option, this target is not sufficient to be the guide for the overall optimal design or debottlenecking. Some analysis carried out by the Hydrogen Pinch is based on the assumption of a static pinch point during system changes. For example, the conclusion that the purification below the pinch will not affect hydrogen utility is only true when the pinch point does not move after the installation of a purification unit. One must be very careful that if a new pinch point is formed below the former pinch point because of the purification, the utility usage can be increased. The mathematical approach is applied to the hydrogen distribution network design. The objective function is the minimum total cost of the network instead of the minimum hydrogen utility. However the linear programming approach simplifies the practical constraints and often gives unrealistic design solutions.

1.3. Approaches developed in this thesis for hydrogen integration
There are several options available to solve the hydrogen problems in refineries. They are: ? Reuse refinery off-gas by header modification ? Recoverd hydrogen from refinery off-gas by purification 15

Chapter 1

Introduction

? Increasing catalytic reformer capacity to produce more by-product hydrogen ? Increase hydrogen generation capacity by revamping existing hydrogen plants or building new capacity ? Buy hydrogen from third parties The hydrogen integration is aiming to fulfil the hydrogen requirement for the refinery operation by the optimal economic solution. The hydrogen balance is determined by the refinery crude properties and refining procedure; the latter is related to legalised transportation fuel specification and product slate for higher gross profit. To make these decisions, overall refinery planning and optimisation can be done under a sensible hydrogen price. Once the principle configure of a refinery is in place, the hydrogen solution for refinery is then searched by integrating hydrogen network in refinery. Philips (1999) suggested different hydrogen scenarios according to the tightened clean fuel specifications that will be applied in stages. Refiners may look at the economic performance as the criteria, e.g. little process change, lowest investment, short downtime etc. Direct reusing refinery off-gas by header modifications provides easy access to the answer, however extra compression may be necessary to overcome pressure problems. Both the piping changes and the installation of compressors lead to investment and the latter increases operating cost. More hydrogen can be recovered from refinery off-gas by purification. The design of new purification processes takes into account the trade-offs between recovery and investment, while supplementary compressors and piping changes may be required. Therefore, the optimal strategy for purification can be achieved by looking at purification, compression and piping changes simultaneously. An automated design approach is developed in Chapter 2 with more insights into hydrogen distribution networks. The objective function of the optimal design is to minimise total annual cost. Chapter 3 focuses on selecting purification strategy. When hydrogen recovered from refinery off-gas can not satisfy the hydrogen balance in a refinery, new hydrogen generation capacity should be considered. A hydrogen plant is a straightforward solution for more hydrogen requirement. Nevertheless, selecting feedstock for a hydrogen plant relates to further utilisation of refinery off-gas.

16

Chapter 1

Introduction

Chapter 4 addresses the integration of hydrogen plants and purification processes. Again, the aim is to find the optimum solution for an overall hydrogen network instead of individual parts. Chapter 5 concludes the thesis and discusses the further integration opportunities in hydrogen integration.

17

Chapter 2

Hydrogen network optimisation with pressure consideration

Chapter 2: Hydrogen network optimisation with pressure consideration
2.1. Match sources and sinks in hydrogen distribution networks
In refineries, hydrogen distribution networks are composed of sources including hydrogen utility, hydrogen-rich off-gas, and sinks for example hydrotreaters and hydrocrackers. Hydrogen is supplied by hydrogen utility, consumed by hydrogen sinks and disposed to a refinery fuel gas flare and system. To make the best use of hydrogen resources, sources and sinks shall be matched through an ingenious network design. Hydrogen Pinch can find the minimum hydrogen utility via a hydrogen surplus curve under the consideration of changing the network layout. However, the network design to achieve this target can not be derived through the surplus curve. Linear programming has been used to find the optimum design in previous research. However, the solution is often too trivial to apply. In this section, a systematic method
90.00 Mscfd 92.8% Unit A 310.00 Mscfd 40.00 Mscfd 91.0%

111.43 Mscfd Hydrogen Plant 278.13 Mscfd 99.0% 87.6% Unit B

488.57 Mscfd

11.43 Mscfd 85.0% 89.13 Mscfd 81.9% Offgas

26.70 Mscfd 77.7% Unit C

213.30 Mscfd

9.70 Mscfd 75.0%

50.00 Mscfd 75.4% Unit D

220.0 Mscfd

28.00 Mscfd 70.0%

Figure 2.1 An example of hydrogen distribution networks

18

Chapter 2

Hydrogen network optimisation with pressure consideration

is developed to match sources and sinks in order to have the minimum hydrogen utility usage. The method will be introduced through an example shown in Figure 2.1. In this hydrogen distribution network, hydrogen is supplied by a hydrogen plant. There are four hydrogen consumers - Unit A to Unit D. Part of purge streams from consumers are recycled and the rest of them are released to a fuel gas system. There is potential hydrogen saving because the purity of off-gas (81.9%) is higher than the demand of Units C (77.7%) and Unit D (75.4%). ? Step 1 – Decomposing consumer into sinks and sources ? The hydrogen consumers can be decomposed into sinks and sources. These sinks and sources have certain purity and flowrate. The hydrogen plant is considered as a source with constant purity and unknown flowrate. The flue gas system is considered as a sink with no limitation of purity and flowrate. ? Step 2 – Ranking ? Rank the sources and the sinks from high to low according to their purities as shown in Figure 2.2. ? Step 3 – Feed sources to sinks

Sinks
400 Mscfd 92.8% 600 Mscfd 87.6% 240 Mscfd 77.7% 270 Mscfd 75.4% Offgas (? Mscfd, ?%) Unit C Unit A

Sources
350 Mscfd 91.0% 500 Mscfd 85.0% 223 Mscfd 75.0% 248 Mscfd 70.0%

Unit B

Unit D

Hydrogen Plant (? Mscfd, 99.0%)

Figure 2.2 Decomposing hydrogen consumers into sinks and sources

19

Chapter 2

Hydrogen network optimisation with pressure consideration

? Apart from the hydrogen utility, feed the highest sink with the highest source. Following rules are applied: ? ? ? ? If the source purity is not high enough, introduce hydrogen utility. If the source quantity is not enough, introduce the next highest source. If the source purity is too high, introduce the next highest source. If the source quantity is more than enough, feed the next highest sink.

? Step 4 – Repeat step 3 to the remaining sources and sinks from high to low according to their ranks until the last sink is fed. The sources left are sent to the fuel system. ? Figure 2.3 presents the final network configuration. The hydrogen utility is saved by 13.2% and the off-gas purity is reduced from 81.9% to 70%.

2.2. Optimal hydrogen allocation with pressure constraints
2.2.1. Network re-routing without interception
? This section will describe a new method that can account for the pressure constraints and existing compressors while optimising the allocation of hydrogen.

90.00 Mscfd Unit A 92.8%

310.00 Mscfd 91.0%

40.00 Mscfd

94.29 Mscfd Unit B Hydrogen Plant 241.55 Mscfd 99.0% 12.38 scf Unit C 77.7% 87.6%

465.71 Mscfd 85.0%

34.29 Mscfd

Offgas 193.33 Mscfd 75.0% 52.55 Mscfd 29.67Mscfd 70%

44.88 scf Unit D 75.4%

195.45 Mscfd 70.0%

52.55 Mscfd

Figure 2.3 Final network configuration after stream match

20

Chapter 2

Hydrogen network optimisation with pressure consideration

310 Mscfd AR 90 Mscfd 200 Mscfd Hydrogen Plant 90% 360psi 110 Mscfd BM 2200 psi AM AR BM BR 40 Mscfd 91% 1500 psi Fuel 80 psi

AM

1600 psi

Unit A 490 Mscfd BR

Unit B 85% 10 Mscfd 1700 psi

make-up compressor for unit A recycle compressor for unit A make-up compressor for unit B recycle compressor for unit B

Figure 2.4 An existing hydrogen network with pressure constraints

No interception will be considered in this section. Based on industrial experience, this type of problems is common. It occurs when refineries are only interested in low-cost modifications with short-term downtime such as re-routing and allocation changes. Often, budgets are limited and so there will be no capital for interception equipment such as compressors or purifiers. ? The method is based on optimising a reducible superstructure and will be illustrated using a simple problem. This example consists of two hydrogen consumers (Unit A and B) and a hydrogen plant. The hydrogen system is shown in Figure 2.4. Each consumer has both a make-up and a recycle of hydrogen. In the base case, 200 Mscfd (2.23 x 105 Nm3/h) are supplied from the hydrogen plant. ? Applying the targeting method of Hydrogen Pinch (Alves, 1999) gives a minimum target from the hydrogen plant of 182.9 Mscfd (2.04 x 105 Nm3/h) – a potential hydrogen saving of 8.7%. This is a significant amount, but may not be achievable in practice. ? Because the purge from Unit A has a high purity, any design to meet the target will require this stream to be reused in Unit B. However, the pressure of the purge from Unit A (1500 psi or 10.34 MPa) is lower than that at the inlet of B (2200 psi or 15.17 MPa). Thus, direct reuse from Unit A to Unit B is not feasible and the target may not be achievable. Determining exactly what is feasible is a tedious task as the existing

21

Chapter 2

Hydrogen network optimisation with pressure consideration

compressors need to be included in the problem. This was not addressed in the targeting method of Alves (1999) and will be covered in the discussion that follows. ? Before moving on to the mathematical aspects of the method, the physical, engineering issues need to be discussed. Obviously, direct reuse of hydrogen between consumers is only possible if the pressure is sufficient. However, it is possible to reuse a hydrogen stream indirectly, i.e. by routing through an existing compressor, provided that certain conditions are met. Firstly, there has to be sufficient capacity in a compressor to accommodate the stream. Reusing hydrogen will change the make-up and recycle flowrate throughout the system and some spare capacity may be available in one or more compressors. Also, the initial pressure of the reused stream needs to be high enough to be fed to the compressor as compressors are designed for a specific inlet pressure. In addition, the compressor should be able to compress the stream to a high enough pressure that it can be used in the required consumer. ? The first step in the method is to set up a superstructure that embeds all possible connections while incorporating the physical insights discussed above. First connecting every sink with every source forms the superstructure; then the

360 psi

H2 Plant Source A Source B

Sink A Sink B Fuel

1600 psi 2200 psi 80 psi

1500 psi

1700 psi

1600 psi

AM

AM

360 psi

2200 psi

BM

BM

360 psi

1600 psi

AR

AR

1500 psi

2200 psi

BR

BR

1700 psi

Figure 2.5 Example of reduced superstructure

22

Chapter 2

Hydrogen network optimisation with pressure consideration

superstructure can be reduced by eliminating the connection where the source pressure is less than the sink pressure (Figure 2.5). The source pressure does not have to be equal to the sink pressure because the source can readily be dropped to a lower pressure, for example by passing through a valve. In this formulation the compressors are included as both sources and sinks; the inlet of a compressor is treated as a sink and the outlet of a compressor is treated as a source. ? Setting up the superstructure in this way assumes that the inlet and outlet pressures of the compressors are fixed at their design values. An alternative approach would be to allow the inlet and outlet pressures of the compressors to be variables which could be related by compressor model, which would make the optimisation problem more difficult to solve. It is important to note, however, that only the existing compressors are treated in this way. This makes sense because the machines are designed to operate under certain conditions. New compressors are different because their inlet and outlet pressure levels are undetermined and these pressures will be varied in the optimisation. More will be discussed in the following sections. The second step is to formulate a mathematical program for the design. The mathematical method is capable of addressing a much wider range of objectives, for example the minimum hydrogen utility or the minimum total cost, which is a significant improvement over the pinch method and will be illustrated in section 2.3. For now, the objective function will be the minimum hydrogen utility to investigate the impact caused by pressure constraints.

2.2.1.1. Sink requirements Regardless of the objective, there are several constraints that must be met. In order to maintain the operation of all the hydrogen consumers, the amount of gas fed as well as the hydrogen purity (partial pressure) at the reactor inlet must be constant. These constraints are expressed as:

∑F
i

i , Sk

* = FSk

(2.1) (2.2)

∑F
i

i ,Sk

* ? y i = FSk ? y Sk

23

Chapter 2

Hydrogen network optimisation with pressure consideration

2.2.1.2. Source availability Hydrogen in a source can be used in sinks or sent to the fuel gas system. The total amount of gas sent to the sinks and fuel system must be equal to the amount available from the source:



j

* FSr , j = FSr

(2.3)

2.2.1.3. Existing compressors Compressors behave as both sinks and sources, but these differ from those associated with the hydrogen consumers. Both the flowrate and the purity in the compressors are now variables, whereas in the hydrogen consumers these were constants. The constraints on the compressors are as follows. As shown in Figure 2.6, the flowrate of gas entering a compressor must be equal to the flowrate leaving the compressor:


leaving:

j

FCompi , j = ∑iFi ,Compj

(2.4)

The amount of pure hydrogen entering a compressor must be equal to the amount



j

FCompi , j ? y Compi = ∑i Fi ,Compj ? y i

(2.5)

Finally, an existing compressor will have been designed for a specific flowrate and so there will be a maximum flowrate constraint on each compressor.

∑F

i i ,Compj

Maximum ≤ FCompj

(2.6)

This constraint may be too simplistic. The maximum or ‘stonewall’ flowrate of a

Fi, Compj

FCompi, j

Figure 2.6 Decomposing compressor into sink and source

24

Chapter 2

Hydrogen network optimisation with pressure consideration

310 Mscfd AR 90 Mscfd 200 Mscfd Hydrogen Plant 30.4 Mscfd 91% 1500 psi Fuel BR 80 psi

AM

1600 psi

Unit A

9.6 Mscfd 484.5 Mscfd 90% 360 psi 105.9 Mscfd BM 2200 psi

Unit B 85% 15.5 Mscfd 1700 psi

Figure 2.7 The minimum hydrogen utility achievable with the existing compressor is 195.9 Mscfd

compressor is usually an actual volume flowrate. Therefore, the capacity limit would be a function of both molar or standard volume flowrate and the pressure. However, in this work the compressor inlet pressure is fixed and so it is sufficient to consider standard volume or molar flowrate. Sometimes, the limitation on a compressor can be in terms of the power required rather than the flowrate. The formulation developed so far is sufficient to solve the example shown in Figure 2.4, Figure 2.5 and Figure 2.7. The introduction of new equipment will be discussed in a design as well as cost optimisation. Because several sources may be mixed before entering a compressor, the hydrogen purity in the compressors is not known a priori and so the problem is non-linear. In this

Hydrogen Import Mscfd

202 200 198 196 194 192 190 188 186 184 182 0.00% 10.00% 20.00% 30.00% 40.00% 50.00%

Capacity increment %

Figure 2.8 Sensitivity analysis showing that the bottleneck is the capacity of the make-up compressor to unit B

25

Chapter 2

Hydrogen network optimisation with pressure consideration

example, it is assumed that all the compressors are currently operating at their design flowrates and have an additional 5% margin. Optimising with the objective function being the minimum hydrogen utility gives the design shown in Figure 2.7. Even though the pressure of the purge gas from Unit A is too low to be fed directly to Unit B, it can be fed to unit B’s make-up compressor as this is designed for an inlet pressure of 300 psi. The minimum utility is 195.9 Mscfd (2.19 x 105 Nm3/h)and so the initial target of 182.9 Mscfd (2.04 x 105 Nm3/h) is clearly too optimistic. It is interesting to note that the hydrogen recovery in this problem is bottlenecked by the capacity of the make-up compressor of Unit B. This can be seen by performing a sensitivity analysis in which the minimum utility is determined under a range of different compressor capacities as shown in Figure 2.8. This shows that the theoretical target of 182.9 Mscfd (2.04 x 105 Nm3/h) can only be reached if the bottlenecked compressor can have its flowrate increased by 21%. Of course, another way to achieve the target would be to install one or more new compressors.

2.2.2. Source interception
So far, the introduction of interception equipment has not been considered. However, if budgets allow for it, new equipment can be added to the network and this section will examine how this can be achieved in the best way. The major items of interception equipment that can be added to a hydrogen network are new compressors and purifiers.

2.2.2.1. Compressors New compressors can be added into the superstructure developed earlier, but there are some important points to note. As with an existing compressor, the total gas flowrate and the amount of pure hydrogen must be conserved across the unit and so equations (2.4) and (2.5) still apply. However, the machine has not yet been built and so there is no maximum flowrate limit (other than manufacturing limitations). Another major difference between a new compressor and an existing one is that the inlet and outlet pressures are not known before the network design. In other words,

26

Chapter 2

Hydrogen network optimisation with pressure consideration 310 Mscfd AR 90 Mscfd 182.9 Mscfd Hydrogen Plant 90% 360 psi BM 2200 psi 91% 1500 psi AM 1600 psi Unit A 17.4 Mscfd 1700 psi 467.1 Mscfd

22.6 Mscfd BR

92.9 Mscfd

Unit B

Fuel 80 psi 85% 1700 psi 32.9 Mscfd

Figure 2.9 If a new compressor is installed, the minimum utility target of 182.9 Mscfd can be met.

pressures are now also optimisation variables. The inlet pressure will be the lowest pressure over all the sources that feed the compressor. The compressor will be designed to be able to receive the lowest pressure source and all sources with higher pressures that also feed the machine will be passed through a valve to drop their pressure. Similarly, the design pressure for discharge will be the highest pressure over all the sinks that are fed by the new compressor. For an existing hydrogen distribution network, the pressure levels of sources and sinks are limited. One of the practical approaches to decide the inlet and outlet pressure of a new compressor is by enumeration. However, once the design pressure of new compressor is nominated, the problem is still non-linear. To demonstrate the use of a new compressor, consider the example shown in Figure 2.4 again. One possible solution using a new compressor is shown in Figure 2.9. The dotted lines indicate new equipment. The hydrogen utility is 182.9 Mscfd, which is exactly on target. Many different solutions exist that achieve the target. However, some of them would involve unnecessary compression. In Figure 2.9, the new compressor takes some of the purge from unit A and compresses it so that it can be fed to unit B’s recycle compressor. This is more sensible than compressing the gas to feed unit B directly. The pressure ratio of the new compressor is very small (1700/1500=1.13) whereas if it were going to feed unit B directly, the pressure ratio would have to be 2200/1500 =1.47. By keeping a low pressure, the compressor power requirement is

27

Chapter 2

Hydrogen network optimisation with pressure consideration

reduced and this gives lower operating cost as well as a smaller capital cost. Inclusion of equipment sizing and cost optimisation will be discussed in more details later. In some systems, more than one new compressor will have to be added in order to meet the minimum utility target. At first, it may seem that one compressor which delivers gas at the highest sink pressure will be sufficient. However, if there are too few compressors, gas streams with different hydrogen purity levels may have to be mixed together before being compressed. The mixing could degrade the purity of certain hydrogen sources, making them unsuitable for use even though they may be at very high pressures. Even if this is not the case, it may not be economical to have one large compressor designed to feed the highest pressure sink, as this could involve streams being compressed to pressures much higher than necessary, only to be throttled before use. Clearly this would be waste of power. On the other hand, space limitations on a site may restrict the number of new compressors that may be added.

2.2.2.2. Purifiers The pinch method of Alves (1999) is valuable in a sense that it predicts the minimum utility target that can be reached by reconnecting the existing sources and sinks regardless of pressure constraints. No matter how many new compressors are added, this target can not be improved upon without introducing purification. As discussed earlier, purifiers are interception units that upgrade the hydrogen purity of sources. The most commonly used purifiers are PSA and membrane units. For

Product

To sinks

From sources

Feed

Residue

To sinks

Figure 2.10 A purifier in the superstructure of hydrogen networks

28

Chapter 2

Hydrogen network optimisation with pressure consideration

hydrogen network design, there are three main issues that need to be addressed. These are product purity, product pressure and hydrogen recovery. PSA units are capable of achieving very high purity – often exceeding 99.9% – while membranes typically give lower purity – up to approximately 98%. Another advantage of PSAs over membrane separation is that the pressure drop over a PSA is very low (typically approximately 0.05 – 0.1 MPa), so that product pressure is virtually the same as the feed pressure. The residue from a PSA is at very low pressure and this fact, coupled with the low purity of the stream means that it is typically sent to the fuel system. In contrast, the pressure drop over a membrane unit is much higher. In fact, a large pressure drop is required to give satisfactory product purity. The residue or retentate stream from a membrane is close to the feed pressure, but has a lower purity. One area where membranes are superior is the hydrogen recovery, which is defined as the amount of pure hydrogen in the purifier feed that goes to the product stream. For PSA units, recoveries up to 90% are achievable whereas for membranes it can be as high as 95%. Both PSA and membrane units suffer from a reduction in recovery as the product purity increases. More details about hydrogen purification processes can be found in Chapter 3. Installing a purifier into a network means that one new sink (the feed stream) and two new sources (the product and residue streams) are introduced into the superstructure. Figure 2.10 shows a schematic diagram of a purifier. The feed stream flowrate is:
feed FPurifier = ∑i Fi , Purifier

(2.7)

and the feed purity is:
feed y Purifier =

∑F ∑F
i i

i , Purifier

? yi

(2.8)

i , Purifier

The inlet pressure is the lowest pressure over all the sources feeding the unit. If several streams are mixed before being purified, the ones with higher pressures need to have their pressures reduced, e.g. by passing through a valve. The product and residue pressures are related to the feed pressure. The product pressure is equal to the inlet pressure subtracting the pressure drop over purifier. If the pressure drop 29

Chapter 2

Hydrogen network optimisation with pressure consideration

across a purifier is large, a new or existing compressor may be required to compress either the feed or the product steams. The residue pressure depends on which purification process is selected. The hydrogen recovery RPurifier of a purifier is a function of feed purity and operating pressure. This chapter will only focus on the impact of pressure limitation. Therefore by specifying the product purity, yProduct, and hydrogen recovery, RPurifier, of the purifier, the product and residue streams can be calculated: Product flowrate:
product = FPurifier feed feed ? y Purifier R Purifier ? FPurifier product y Purifier

(2.9)

Residue flowrate:
redsidue feed product = FPurifier ? FPurifier FPurifier

(2.10)

Residue purity:
residue = y Purifier feed feed ? y Purifier (1 ? R Purifier ) ? FPurifier residue FPurifier

(2.11)

If an existing network already has a purifier, this methodology can be used for relocating it. It would usually be necessary to include some constraints about the existing purifier such as the maximum feed flowrate or the designed pressure.

2.3. Automated design for hydrogen distribution system
To find the minimum hydrogen utility under pressure constraints, the objective function to achieve the optimum design of a hydrogen distribution network is the minimum hydrogen utility flowrate. However, because the complexity of the network has not been taken into account, the solution of the network design may not be realistic. The number of solutions can be infinite and all of them give the same minimum hydrogen utility flowrate. Many matches with very small flowrate between sources and sinks will appear. To reduce these matches by banning streams manually can be very arbitrary.

30

Chapter 2

Hydrogen network optimisation with pressure consideration

The objective function of the minimum hydrogen utility can only find the target instead of realistic design to achieve the target. To find the optimum network design automatically, the objective function should be set as the minimum cost (operating and capital). For new designs, the problem becomes to optimise hydrogen distribution to satisfy the hydrogen requirement from sinks. For retrofit design of an existing hydrogen distribution network, one aspect of the problem is to fulfil the increased hydrogen demand, and the other aspect can be to reduce the operating cost of the network. Three options are available: piping changes, extra compression, and purification. Simultaneous application of these three options is essential. The selection criteria are based on their performance and costs. In order to achieve the optimum design with the consideration of hydrogen utility, operating cost and capital cost for new equipment, the objective function should be to minimise the total annual cost. TAC = Cost Operating + Af ? (Cost Equipment ) (2.12)

The total annual cost refers to the annual operating costs and annualised capital cost for new equipment. The operating costs include hydrogen cost, fuel cost (credit if exported) and power. The cost of hydrogen is a part of operating cost and can be calculated from hydrogen generation cost. If hydrogen is imported from a third party, it is equal to contract value for annual spending on hydrogen. The fuel cost refers to the annual value of hydrogen-rich off-gas eventually released to a fuel system, which is operating credit. The power cost is normally from compressors. Equipment costs include new compressors, new purifiers and necessary piping changes. Af means annualising factor. Smith (1995) addressed Af as: Af = fi ? (1 + fi ) (1 + fi )ny ? 1 (2.13)

where fi is the fractional interest rate and ny is the number of years.

2.3.1. Compressor model
Peters and Timmerhaus (1991) provided the equations to obtain the power and cost of the compressors. For an ideal gas undergoing an isentropic multistage compression, assuming equal division of work between cylinders and inter-cooling of gas to original intake temperature, the compressor power can be calculated by the following equation: 31

Chapter 2

Hydrogen network optimisation with pressure consideration
?5 γ ?1 γ ? ?N s

? ? P2 3.03 × 10 γ ? N s hp = P1 ? q fm1 ?? ?P ? η ? ( γ ? 1) ? 1 ? ?

? ? ?

? ? 1? ? ? ?

(2.14)

where hp = power requirement, horsepower γ = ratio of specific heat of gas at constant pressure to specific heat of gas at constant volume P1 = intake pressure, lbf/ft2 P2 = final delivery pressure, lbf/ft2 qfm1 = throughput, ft3/min Ns = number of stages of compression η = efficiency For ideal gas,
CP γ = γ ? 1 8.314 JK ?1 mol ?1

[

] ]

(2.15)

P1 ? q fm1 = 2116.224 lbf/ft 2 ? q sm1

[

(2.16)

where qsm1 is the volume in standard condition.

Centrifugal Reciprocating 1400

Cost fob 10 US$(1990)
3

1200 1000 800 600 400 200 0 0 500 1000 1500 2000 2500 3000 3500

Power / bhp

Figure 2.11 Compressor capital cost vs. power
Data from Peters, M.; Timmerhaus, K.; Plant Design and Economics for Chemical

Engineers, 4th edition, McGraw-Hill, 1991

32

Chapter 2

Hydrogen network optimisation with pressure consideration

The specific heat of the mixture of hydrocarbon and hydrogen is in certain range and is assumed to be a constant. With the specified efficiency, the power of compressor is calculated as:
? ? P2 Power[kW ] = 158 N s ? q sm1 [Mscfd ]?? ?? P ? 1 ? ? ? ? ? ?
0.286 Ns

? ? 1? ? ? ?

(2.17)

The number of stages can be decided by the ratio of outlet pressure and inlet pressure:
N s = int( P2 1 ) P1 ratio per stage

(2.18)

To avoid the situation that the outlet temperature of a compressor is higher than the flash point of lubricant oil, the maximum pressure ratio of hydrogen-rich gas compression for a single stage is about 3. When the number of stages is more than one, an internal cooler is necessary to guard the inlet temperature allowable to the next stage. The contribution of cooling water to the operating cost is much smaller than the power consumption of the compressor. Therefore it is omitted in operating cost calculation. The capital cost of a compressor is related to the compression power as shown in Figure 2.11. The following equation is for calculating the capital cost by linear correlation:
I COMP [k$US ] = a COMP ? If COMP + bCOMP ? PowerCOMP [kW ]

(2.19)

Updated by Marshall & Swift index (1999 2nd Quarter), the cost coefficients for centrifugal compressors are evaluated as aCOMP=764.86, bCOMP=1.7596. The capital cost of an inter-cooler is included. The usage of cooling water is ignored because of the small duty comparing to electricity cost.

2.3.2. Piping cost
Previous research can obtain the network design to achieve minimum hydrogen utility with piping changes. The assumption behind the solution is that piping is so cheap that it could be ignored during design. However, it is not true especially for high-pressure piping. 33

Chapter 2

Hydrogen network optimisation with pressure consideration

1000 900 800 700 600 CS SS Linear (S S ) Linear (CS )

?/m

500 400 300 200 100 0 0 0.02 0.04 0.06 0.08
2

0.1

0.12

Cross section area (m )

Figure 2.12 Linear approximation of piping cost

Data from IChemE; A Guide to Capital Cost Estimating, 3rd edition, IChemE and The Association of Cost Engineers, London, 1988; Linear approximation by Argaez (1999) The cost function of piping is linearly approximated using the data reported by IChemE as shown in Figure 2.12. The cost includes the supply and erection of pipe, flanges, fittings and welding requirements.
I PIPE [$US ] = (a PIPE ? If PIPE + bPIPE Acs [sf ]) ? Length PIPE [m ] Acs [sf ] = 0.02352q sm1 [Mscfd ] Pressure [MPa ]

(2.20) (2.21)

The cost coefficients aPIPE and bPIPE are constant but can be varied by different pipe materials and operation conditions, updated using Marshall & Swift index (1999 2nd Quarter). If the pipe is for the process operating in high pressure or temperature, the cost can be estimated by multiplying a factor. For example, if the material of pipe is stainless steel and under high operating pressure (factor =1.2), the corresponding cost coefficients are aPIPE =420.74, bPIPE=1484.76. The length of piping, LengthPIPE, required to link a source and a sink is not simply the straight-line distance between them. In real plants, there are often pipe bridges, which

34

Chapter 2

Hydrogen network optimisation with pressure consideration

SR

CCR

?y

HCU

DHT

KHT

Length = ?x + ?y

CNHT
?x

NHT

HDA
Figure 2.13 Estimating piping lengths in refinery plot plan

can take a convoluted route for reasons of layout or safety. Figure 2.13 shows one possible way of estimating piping lengths if no detailed information is available.

2.3.3. PSA capital cost
PSA is used as an example for purifiers. Ruthven et al. (1994) stated that the PSA capital cost is a linear function of throughput:

I PSA = a PSA + bPSA ? FPSA

(2.22)

Towler et al. (1996) fitted constants aPSA and bPSA by the cost data from Vervalin (1994) and rearranged the cost function to:
I PSA [US$1994 /kscf ] = 0.4330 0.2986 + q sm1 [Mscfd ] Y ? Z

(2.23)

where qsm1 is the flowrate of hydrogen, Y is the recovery yield of hydrogen and Z is the feed mole fraction of hydrogen. Using Marshall & Swift index (1999 2nd Quarter), the equation (2.22) can be updated to:
I PSA [kUS$ ] = 503.8 + 347.4 ? FPSA [Mscfd ]

(2.24)

35

Chapter 2

Hydrogen network optimisation with pressure consideration

2.3.4. Optimisation strategy
The automatic retrofit design requires Mixed-Integer Non-linear programming. Integer variables are used to identify the existence of new equipment and the availability of new piping. Mixing and splitting functions causes non-linearity. All of these nonlinear terms are bilinear. Quesada and Grossmann (1995) proposed an algorithm dealing with global optimisation of networks consisting of splitters, mixers and linear process units, which only contain nonlinear equations in bilinear forms. All the bilinear terms are relaxed to four linear inequalities. Thus a linear relaxation of the original nonconvex nonlinear problem can be generated. Solving this LP problem provides a valid lower bound to the global optimum for the original problem. A spatial branch and bound search is then employed until all the sub-regions have been evaluated and the global optimum is achieved. The bilinear term x1?x2 can be relaxed into the following linear constraints:
Low Low ? x1Low ? x 2 x 12 ≥ x1Low ? x 2 + x1 ? x 2

(2.25) (2.26) (2.27) (2.28)

Up Up Up ? x 2 + x1 ? x 2 ? x1 ? x Up x 12 ≥ x1 2

Up ? x1Low ? x Up x 12 ≤ x1Low ? x 2 + x1 ? x 2 2

Up Up Low Low ? x 2 + x1 ? x 2 ? x1 ? x2 x 12 ≤ x1

Up Low ≤ x1 ≤ x1Low , x Up where x1 2 ≤ x2 ≤ x2

x 12 = x1 ? x 2

Based on the concept of relaxing bilinear terms into linear inequalities, a strategy is developed to improve robustness and help to locate good optima of MINLP. Firstly, the non-linear items are relaxed into linear forms. For example, equation (2.5) represents the hydrogen balance for a compressor. If the sources feeding this compressor have constant purity yi, the equation is nonlinear with a bilinear term of multiplication of flowrate through the compressor and hydrogen purity, both of which are variables. Thus this equation can be relaxed into inequalities:

36

Chapter 2

Hydrogen network optimisation with pressure consideration

Relax network models to linear models and solve problem by Mixed-Integer Linear Programming (MILP)

Use MILP solution as initialisation

Solve problem by Mixed-Integer Non-Linear Programming (MINLP)

Figure 2.14 Solution strategy for MINLP in automatic network design

Sum , Low Sum Low Sum , Low Low Fy ≥ FCompi ? yCompi + FCompi ? y Compi ? FCompi ? yCompi

(2.29) (2.30) (2.31) (2.32)

Sum ,Up Sum Up Sum ,Up Up Fy ≥ FCompi ? yCompi + FCompi ? yCompi ? FCompi ? yCompi

Sum , Low Sum Up Sum , Low Up Fy ≤ FCompi ? yCompi + FCompi ? y Compi ? FCompi ? yCompi

Sum ,Up Sum Low Sum ,Up Low Fy ≤ FCompi ? yCompi + FCompi ? y Compi ? FCompi ? yCompi

H2 Plant
44.35 92.00%

45.00 92.00% 0.65 5.57

23.50 75.00%

CCR

10.66

38.78

HCU
11.29 75.00% 8.65 2.64 8.21 86.53%

11.31 75.97%

0.04

DHT
8.61 70.00%

IS4
Flows in Mscfd

JHT
4.32 65.00%

CNHT
3.47 75.00% 3.47

12.08 71.44%

NHT
6.55 60.00%

12.80

Fuel

Figure 2.15 Base case – hydrogen distribution network in existing refinery

37

Chapter 2

Hydrogen network optimisation with pressure consideration

Sum , Low Sum Sum ,Up Low Up where FCompi ≤ FCompi ≤ FCompi , yCompi ≤ yCompi ≤ yCompi Sum ? yCompi , Fy = FCompi Sum = ∑ j FCompi , j FCompi

By applying this procedure to each bilinear term, the problem is then turned into MixedInteger Linear Programming (MILP). The solution from the MILP is then used as the initialisation for the Mixed-Integer Non-linear Program. The solution procedure is shown in Figure 2.14.

2.4. Case studies
In the case studies, the hydrogen distribution network in an existing refinery is used as the base case. The current system is shown in Figure 2.15. There are six consumers, which are hydrocracker (HCU), jet fuel hydrotreater (JHT), cracked naphtha hydrotreater (CNHT), diesel hydrotreater (DHT), naphtha hydrotreater (NHT) and isomerisation unit (IS4). Hydrogen is supplied from a catalytic reformer as well as a hydrogen plant. There are two make-up compressors in the system and all the consumers except the isomerisation unit have internal recycle compressors. Currently, 45 Mscfd (5.03 x 104 Nm3/h) of hydrogen are produced in the hydrogen plant with the maximum capacity of 50 Mscfd (5.58 x 104 Nm3/h). Process flowrate and pressure data

Table 2.1 Process data in base case
Process M
Mscfd

Make-up yM Pressure
Vol%H 2 75.97 86.53 75.00 71.44 75.00 92.00 psi 600 500 500 300 300 2000

P
Mscfd 8.61 3.47 4.32 6.55 11.29

Purge yP Pressure
Vol%H 2 70.00 75.00 65.00 60.00 75.00 psi 400 350 350 200 1200

Recycle R
Mscfd 1.56 36.75 3.6 3.59 85.7

DHT CNHT JHT NHT IS4 HC

11.31 8.21 8.65 12.08 0.04 38.78

Flow
Mscfd

Hydrogen supply y Maximum flow Pressure
Mscfd 50.00 23.50 Vol%H 2 92.00 75.00 psi 300 300

H2 Plant CCR

45.00 23.50

38

Chapter 2

Hydrogen network optimisation with pressure consideration

Table 2.2 Piping distance between units (m)
Sinks Sources

HC 250 450 0 250 200 100 400 100 300 200 500

CNHT DHT 400 400 250 0 150 150 150 350 250 250 450 350 250 200 150 0 300 200 300 100 200 300

JHT 350 550 100 150 300 0 300 200 400 300 600

NHT 550 450 400 150 200 300 0 500 300 400 500

IS4 450 250 300 250 100 400 300 400 200 300 200

C1 150 350 100 350 300 200 500 0 200 100 400

C2 250 150 300 250 100 400 300 200 0 100 200

Fuel 450 450 250 150 250 150 50 350 350 350 450

NC 150 250 200 250 200 300 400 100 100 0 300

PSA 250 50 500 450 300 600 500 400 200 300 0

H2 Plant CCR HC CNHT DHT JHT NHT C1 C2 NC PSA

Note:

C1 - compressor No.1 (to HCU) C2 - compressor No.2 (to DHT) NC - new compressor

of all hydrogen consumers and hydrogen producers are given in Table 2.1.

2.4.1. Case 1 – retrofit to save operating costs
The first part of the case study is to find the maximum possible saving on the operating cost. The utility prices in this case are hydrogen 2000 US$/Mscf (0.075 US$/Nm3), power 0.03 US$/kWh, fuel gas 2.5 US$/MBTU (8.53 US$/MWh). The piping distances are given in Table 2.2. Several practical constraints have been imposed by the refinery: ? The existing compressors have 5% spare capacity. ? The internal recycle compressors may not be used by other units for operation reasons. ? There is space on the site only for one new compressor and one new purification unit. ? PSA units are preferred for purification. The product purity of PSA is 99% and the recovery is 90%. ? The PSA tail gas has to be compressed and sent to fuel system. 39

Chapter 2

Hydrogen network optimisation with pressure consideration

Table 2.3 Cost breakdown – minimum operating cost Existing network Operating cost Hydrogen Power Fuel Total MUS$ 32.9 1.77 -12.2 22.54 Total Capital cost Operating cost saving = 6 MUS$/year Payback period = 1.6 year Compressor PSA Piping Total Retrofit network Operating cost Hydrogen Power Fuel MUS$ 20.9 1.87 -6.24 16.53 MUS$ 1.00 7.02 1.78 9.8

The network is designed by setting the objective function as the minimum operating first. The problem is solved using MINLP solver DICOPT provided by GAMS. The resulting design is shown in Figure 2.16. Dotted lines indicate new equipment. To minimise the operating cost, both a new compressor and a PSA are used. The new

H2 Plant
31.25 37.15

28.61 92.00% 2.72

CCR
11.53 0.35 9.21 11.84

23.5 75.00%

5.33

0.04

0.04

HCU
75.00% 8.65

1.01

IS4

JHT
65.00% 4.32 11.14 99.00% 0.35

CNHT
75.00% 18.16 66.78% 3.68 2.69

DHT
9.14 70.00% 6.37 60.00% 1.03

PSA
7.28 17.33%

NHT Fuel

New Compressor

Figure 2.16 Solution for Case 1: minimising operating cost

40

Chapter 2

Hydrogen network optimisation with pressure consideration

H2 Plant
37.03

35.40 92.00% 1.82

0.19

CCR
11.69

23.5 75.00%

0.04

38.04

5.46 0.91

11.88

HCU
75.00% 8.65

0.99

IS4 DHT
70.00% 6.14

JHT
65.00% 4.32 6.47 99.00%

CNHT
75.00% 10.17 1.63

PSA
3.70 19.29%

NHT
6.56 60.00%

3.81

Fuel

Figure 2.17 Solution for Case 1: minimising operating cost with capital limit of 5 MUS$

matches between sources and sinks are introduced by some piping changes. Notice that the new compressor is used to accommodate the increased recycle requirement for the NHT, as well as to compress the feed to the PSA. The total capital investment of the retrofit is 9.8 MUS$ and the operating cost saving is 6 MUS$ per year. The simple payback period (capital cost divided by annual operating cost savings) is therefore 1.6 years. A complete breakdown of all the costs is shown in Table 2.3. Occasionally refineries have limited capital budgets so it would be very valuable to know what the maximum achievable savings are with a fixed capital budget e.g. 5 MUS$. Adding the maximum capital expenditure as an additional constraint, optimisation gives the solution shown in Figure 2.17. The best investment is in a PSA and no new compressor is used. Comparing the retrofit design without the capital limit, fewer new pipes have been installed. The operating cost savings are smaller (only 3.5 MUS$ per year) as to be expected.

2.4.2. Case 2 – retrofit for debottlenecking
In this case, a future hydrogen scenario in the refinery is considered where fuel specifications change as a result of new environmental regulations. As a result, the HCU and CNHT capacities need to be increased by 40%. This means a significant 41

Chapter 2

Hydrogen network optimisation with pressure consideration

Table 2.4 Cost breakdown – minimum operating cost with capital limit of 5 MUS$ Existing network Operating cost Hydrogen Power Fuel Total MUS$ 32.9 1.77 -12.2 22.54 Total Capital cost Operating cost saving = 3.5 MUS$/year Payback period = 1.4 year Compressor PSA Piping Total Retrofit network Operating cost Hydrogen Power Fuel MUS$ 25.84 1.86 -8.70 19 MUS$ n/a 4.04 0.96 5.0

increase in the hydrogen demand. However, the existing hydrogen plant has a maximum capacity of 50 Mscfd (5.58 x 104 Nm3/h) and will not be able to cope. Figure 2.18 shows the solution using the conventional approach, which is to expand the affected units, but not to change the structure of the network. It is necessary to revamp the hydrogen plant to produce an additional 12.7 Mscfd of hydrogen and to install new make-up and recycle compressors to deal with the larger flowrate. Table 2.5 Cost breakdown – conventional debottlenecking solution of Case 2 Existing network Operating cost Hydrogen Power Fuel Total MUS$ 45.8 2.77 -14.5 34 Retrofit network Capital cost Compressor Piping Hydrogen plant Total MUS$ 5.6 0.4 18 24

TAC = 47 MUS$/year 42

Chapter 2

Hydrogen network optimisation with pressure consideration

62.74 92.00% 46.57 92.00% 40.72 30

H2 Plant existing H2 Plant
5.85 13.57 1.95 3.70 8.65 0.65 15.52

Revamp

CCR
10.66

23.50 75.00%

HCU
15.81 75.00%

11.31 75.97%

0.04

DHT CNHT
8.61 70.00% 12.86 3.47

IS4

JHT
3.46 4.32 65.00%

4.86 75.00% 1.39

12.08 71.44%

NHT
6.55 60.00%

12.80

Fuel

Figure 2.18 Conventional way of accommodating the increased hydrogen demand in Case 2

Because the aim is not to save the operating cost, but rather to meet new requirements, it is more appropriate to evaluate designs based on the total annual cost rather than the payback. Capital costs are to be annualised over 2 years with a 5%

New Compressor 44.78

44.97 H2 Plant 92.00%

23.5 75.00% 23.46 0.19 0.19

CCR

7.80 40.72 41.57 7.23

HCU
8.65

20.47 12.68

0.04

75.00%

IS4 DHT
10.17 70.00%

JHT
65.00% 4.32 6.29 7.77

CNHT
75.00% 18.01 72.18% 11.82 99.00% 6.19 21.00% 7.84

PSA

NHT
6.56 60.00%

Fuel

Figure 2.19 Debottlenecking design for minimum TAC in Case 2

43

Chapter 2

Hydrogen network optimisation with pressure consideration

Table 2.6 Cost breakdown – improved debottlenecking solution of Case 2 Existing network Operating cost Hydrogen Power Fuel Total MUS$ 32.8 3.04 -8.93 26.9 Total Retrofit network Capital cost Compressor Piping PSA MUS$ 9.04 1.2 6.76 17

TAC = 36 MUS$/year interest rate. The design from the conventional approach has a total annual cost of 47 MUS$ per year and the breakdown is given in Table 2.5. The design is simpler as very few structural changes have been make, but it may not be the most economical. The capital cost of this design is dominated by the hydrogen plant revamp cost, which is estimated using data in Gary and Handwerk (1994) and updated to 1999 2nd quarter. Applying the new approach discussed above, the obtained design under the objective of the minimum total cost is shown in Figure 2.19. Both a new compressor and a PSA have been introduced, but the hydrogen plant is no longer bottlenecked and so the expensive revamp is no longer necessary. The total annual cost is now 36 MUS$ per year, which is a reduction of 11 MUS$ per year compared to the conventional design. Notice that the new design performs better in terms of both capital and operating costs (Table 2.6).

2.5. Conclusion
A new mathematical approach to the design of refinery hydrogen networks is presented. The method is based upon setting up a superstructure that includes all the possible connections and then subjects this to non-linear or mixed-integer non-linear programming. The primary strength of this method is that it can handle pressure constraints and account for existing equipment. It is therefore better suited for retrofitting existing processes than the previous pinch-based approach.

44

Chapter 2

Hydrogen network optimisation with pressure consideration

Another advantage of this method is that it is a systematic design approach. For example for retrofit design problem, retrofit options such as piping change, new compressors and new purifiers can be selected automatically. It can account for capital costs and therefore be used with a wider range of objective functions. The case studies demonstrate how the minimum operating cost, the maximum capital budgets and the total annual cost can all be considered. Better hydrogen management is not only limited to saving operating costs. The case studies also show how debottlecking objectives can be achieved. Besides pressure constraints, the mathematical nature of the new approach means that a designer is free to include any other practical constraints related to safety, operability, layout, contamination, etc. During the design, the hydrogen recovery of purifier (PSA) is specified as a constant. In reality, it is the function of several operating parameters such as feed and product purity, and operating pressure. These issues will be covered in Chapter 3.

45

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

Chapter 3: Strategy of Purifier Selection and Integration in Hydrogen Network
3.1. Introduction
Alternatives available to satisfy the hydrogen requirement in refineries are limited. Hydrogen can be generate by steam reforming or partial oxidation, or recovered from refinery off-gases. In some cases, the refiners can buy hydrogen from a third party. Among these options, recovering hydrogen from refinery off-gases can be considerably cheaper in both operating cost and capital investment. It is worth to prioritise recovering hydrogen from refinery off-gases with reasonable amount. The off-gases containing hydrogen are from catalytic reformers, hydroprocessors, fluid catalytic cracking (FCC) units and other refining or petrochemical units. The typical content of some off-gases is listed in Table 3.1. Table 3.1 Typical content of some refinery off-gases
Off-gas sources Hydrogen concentration (v%) Pressure (psig)

Catalytic reformer Hydroprocessor High pressure Low pressure Fluid catalytic cracking unit (FCC)
Delayed cocker (DCU)

70-90+

250-400

75-90 50-75 15-20

800-2500 100-250 100-250

Toluene hydrodealkylation (TDA)
Miller and Stoecker(1989)

55

400-500

The purification processes include pressure swing adsorption (PSA), membrane, cryogenic process, and gas-liquid absorption. Each of these processes is based on different separation principles, and therefore they have specific process characteristics. The selection of these purification processes depends on the economic aspects as well 46

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

as process flexibility, reliability and ease of future expansion. Tremendous effort has been made to find the guidelines for the proper selection. Although most methods give the physical insights, they are only instructive to the purification process design. In this chapter, a systematic approach is developed to integrate the purification with hydrogen networks in refineries. The method generates a superstructure through screening the available options of purifiers. The suitable design is then decided by an optimisation procedure. The physical insights are valued to decrease the size of superstructure and the range for solution searching. The appropriate purification system can decrease the hydrogen plant capacity in a new design or provide cheaper hydrogen in a retrofit project.

3.2. Purification process analysis
3.2.1. Pressure swing adsorption (PSA)
A PSA process is based on the principle that the specific adsorbents are capable of adsorbing different gas molecules with different affinity based on partial pressure, size and polarity. Two basic stages are involved: adsorption and regeneration or desorption. The operating pressure in the adsorption stage is higher than in the desorption stage. Because the adsorbent capability for impurities is much higher than for hydrogen in certain partial pressure, most of the impurities are adsorbed together with only a small amount of hydrogen. The impurities can be then removed from the adsorbent by reducing the pressure. The process operates on a cyclic basis. Multiple adsorbers are used in order to purify continuously a feedstock and provide a constant product and a tail gas. A typical sequence chart is shown in Figure 3.1 for a system with four adsorbers. A hydrogen stream is separated from the feedstock in the adsorption phase. The adsorber then goes through co-current depressurisation to repressure other adsorbers, remove impurities from the adsorbent while producing a tail gas. The purge from other adsorbers and finally the product hydrogen are used to repressure the adsorber until it is ready for next adsorption cycle.

47

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

FC

Product

Feed Off-gas
ADSORBER 1 2 3 4 Adsorption E2 BD E1 R P PP E1 PP BD E1 P PP 411 CYCLE E2 EB E1 R P PP

Adsorption E2 BD R P

Adsorption E2 R

Adsorption

E1= Equalisation (Co-Current Depressurisation) E2 = Equalisation (Counter-Current Repressurisation) PP = Provide purge (Co-Current Depressurisation) P = Purge BD = Blowdown (Counter-Current Depressurisation) R = Final Repressurisation Figure 3.1 PSA cycle sequence chart

The product hydrogen is available at roughly the same pressure as the feed. The pressure drop between feed and product is nominal 10 psi (0.07MPa). The product hydrogen is always in very high purity (up to 99.999%) and the impurities will appear in product in the sequence of adsorption strength to adsorbent. Relative adsorptivity of typical feed impurities is given in Table 3.2.

48

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

Table 3.2 Relative adsorptivity of typical components Non-Adsorbed H2 He Light O2 N2 Ar Intermediate CO CH4 C2H6 CO2 C3H8 C2H4 Heavy C3H6 C4H10 C5+ H2S NH3 BTX H2O
Miller and Stoecker(1989)

The performance of a PSA unit can be evaluated by the hydrogen recovery that is defined as the ratio of the amount of hydrogen contained in the product by the amount of hydrogen contained in the feedstock. The hydrogen recovery is influenced by tail gas pressure, feed pressure, feed purity and product purity, and unit configuration, numbers of equalisation phases etc. Low tail gas pressure can significantly improve the hydrogen recovery (Figure 3.2). However, compressing tail gas may be necessary in order to match the fuel system pressure in a refinery or for other usage, which perils the economics of PSA units. Therefore, the selection of the appropriate tail gas pressure is extremely important. The impact of feed gas pressure to hydrogen recovery

BASE -5 -10 -15 -20 5 60

Feed Pressure = 300 psig

Tail Gas Pressure, psig

Figure 3.2 Effect of tail gas pressure on PSA recovery 49

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

BASE -2 -4

Tail Gas Pressure = 5 psig
-6

200

350

Feed Pressure, psig.

Figure 3.3 Effect of feed pressure levels on PSA system recovery is less than that of the tail gas pressure. Figure 3.3 shows there is an optimal feed pressure. Miller and Stoecker (1989) suggested that the minimum pressure ratio between the feed and the tail gas is approximate 4:1, and the optimal range of feed pressure is 200 - 400 psig (1.38 - 2.76 MPa(g)). A low feed purity is not recommended because of poor hydrogen recovery. The low product hydrogen purity can increase the hydrogen recovery, but the effect is relatively small. Most of PSA units are designed to achieve high hydrogen purities.

BASE -1 -2 -3

Impurity = CH4

97

99.9

Product Hydrogen Purity, vol%

Figure 3.4 Effect of product purity on PSA system hydrogen recovery
Data of Figure 3.2, Figure 3.3, Figure 3.4 are from Miller and Stoecker(1989).

50

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

Feed
(High pressure)

Membrane

Residue
(High pressure)

Permeate
(Low pressure)

Figure 3.5 Single-stage membrane process The advantage of using PSA processes to separate hydrogen from refinery off-gas is that the product purity can be very high and the impurities can be controlled in ppm levels. However, the tail gas is difficult to be reused because of its low pressure. A computer aid control gives PSA a wide operation range, and its reliability has been proven by long term operations.

3.2.2. Membrane
A membrane separation is achieved by different permeations between hydrogen and impurities. Table 3.3 gives relative permeabilities of some typical components. The most popular membrane used in hydrogen recovery is composite hollow-fibre membrane composed by an active layer and a support layer. Gases pass through a membrane in two sequential steps: solution and diffusion. A simplified process is shown in Figure 3.5. The permeate is the fast gas which has higher permeability, and enriches in the low-pressure side of the membrane. The pressure difference between the permeate and the residue provides the driving force for the diffusion of gas across the membrane. The membrane performance is much more dependent upon the feed to permeate pressure ratio rather than the operating pressure (Spillman, 1989; Miller and Stoecker 1989). The strength of a membrane limits the design of pressure difference. Because hydrogen is always recovered as a permeate, there is a trade-off between the hydrogen recovery and the product pressure drop. Figure 3.6 illustrates that when the same recovery is maintained, increasing the permeate pressure decreases hydrogen purity and increases the membrane area — the cost of compression is not taken into account.

51

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

Table 3.3 Relative permeability of typical components High H2 H2O H2S CO2
Miller and Stoecker(1989)

Medium C1 O2

Low C2+ N2

Unlike PSA, a membrane process can not remove impurities to a very low level thus it is not suitable when a process requires fine impurity removal. The hydrogen purity in the permeate can be low. The hydrogen recovery increases while the product purity drops as shown in Figure 3.7. The performance of a membrane system can be improved by a multistage design. Spillman (1989) reviewed the performance principles of gas membrane separation and demonstrated some commercial applications of membrane separation processes including the hydrogen recovery. Through the introduction of designs for single-stage

1.25
(HDS off-gas, 800 psig, 75% H2)

1.20 1.15 1.10 1.05 1.00

450 psia permeate 315 psia permeate 140 psia permeate 95
90

% H2 recovery
80 60

40

90 91

92

93

94 95

96

97

98 99 100

Hydrogen purity (v%)
Figure 3.6 Relative cost vs purity for membrane system
Spillman (1989)

52

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

100 90 80 2.5 70 60 50 92 94 96 98 100 Relative permeability : 30 5.0 Pressure ratio

Hydrogen purity (%)

Figure 3.7 Hydrogen recovery vs purity for membrane system
Miller and Stoecker(1989)

and multistage membranes, it is indicated that the membrane designs are very much case dependent. The advantages of membrane gas separation are low capital cost even at low gas volumes, ease of operation, low energy consumption, good space efficiency. Comparing to other gas separation processes, many efforts made for the research of new membrane technologies have brought membrane gas separation into an increasingly important process for gas separation and production.

3.2.3. Cryogenic separation
Cryogenic separation is a low temperature process, which exploits high relative volatility of hydrogen compared to other gas components to separate hydrogen. Figure 3.8 shows the flow diagram for a typical partial condensation process. The feed is cooled in exchanger X-1 to a temperature in which the majority of the C2+ hydrocarbons condenses. The two-phase stream is then separated in separator S-1. The hydrogen-methane vapour from S-1 is sent to exchanger X-2 where it is cooled to a temperature low enough to provide the required hydrogen product purity. The cooled stream enters separator S-2 and the vapour from S-2 becomes the hydrogen product after it is warmed in X-1 and X-2. The hydrocarbon liquids from S-1 are throttled to a vaporisation pressure when exchanged against the incoming feed stream in exchanger X-1. This stream can be withdrawn separately at its highest pressure as a by-product, or mixed with the methane reject stream at a lower pressure. The methane-rich liquid 53

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

from S-2 is throttled to a pressure in which it will boil and provide the necessary temperature difference to the feed to S-2. The S-2 temperature sets the hydrogen product purity by controlling the amount of methane remaining in the vapour phase. The separators S-3 and S-4 are used to provide the proper distribution of liquid and vapour into the multiple passes of the heat exchangers. As shown in the diagram, the refrigeration required by the process is obtained by Joule-Thomson expansion of hydrocarbon. If the process itself can not provide sufficient coolant, external refrigeration is required. Therefore, high hydrogen purity in feed can increase operation cost dramatically. Thermodynamically cryogenic process has higher hydrogen recovery than other purification processes (92-97%). The hydrogen purity in the product is controlled by equilibrium and has less impact on recovery than that in membrane. High product purity leads to large investment.

L.P. Fuel Feed

Hydrogen product M.P.Fuel

X-1 S-1 S-4

X-1 S-2 S-3

“cold box”

Figure 3.8 Partial condensation cryogenic process
Miller and Stoecker(1989)

54

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

The advantage of using cryogenic separation is that the process can deal with low feed purity and give high hydrogen recovery. However, pre-treatment is necessary sometimes to remove low boiling impurities such as N2 and CO before cryogenic separation, and the components such as CO2, H2O, H2S, C5+ to appropriate level in order to avoid freezing. The application is only economically attractive in large-scale units because of high capital cost. The hydrogen recovery cost can be reduced largely if the value of hydrocarbon by-products is considered.

3.2.4. Hybrid system
Because different hydrogen purification processes employ different separation principles, the characteristics of one process are distinctive from others. Efficient Integration of those processes can combine the merits and achieve competitive purification results. The process characteristics that can be taken into account in the hybrid system design are: ? PSA: produce high purity product and remove low boiling point impurity completely. ? Membrane: high hydrogen recovery with high residue pressure ? Cryogenic process: high hydrogen recovery with easy recovery of hydrocarbon byproduct Ratan (1994) proposed hybrid system designs by the integration of membrane – PSA, cryogenic-membrane, PSA-cryogenic and the possible applications. Pacalowska et al. (1996) analysed the economics and flexibility of combination of PSA-cryogenic by case studies and concluded that this combination has a lower hydrogen production cost after accounting for the by-product value comparing to a PSA process alone and a hydrogen plant.

55

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

3.3. Review of existing methods for the selection and integration of purification processes
3.3.1. Industrial approach
How to design and manage hydrogen purification from refinery off-gas is a hot topic in both industry and academic research. The reason is that purification processes rely on different separation theories. They have different features in both processes and economics. Therefore, none of them has dominated the hydrogen purification of refinery off-gases. Many publications addressed the problem of finding general criteria for purification process selection. (Miller and Stoecker, 1989; Ratan, 1994; Whysall and Picioccio, 1998; Allen 1999; Sabram et al. 2001). The advantages and disadvantages of the purification processes have been classified and represented by operating conditions in certain ranges. On the other hand, the refinery off-gases are grouped through the level of hydrogen purity, pressure and volume. The procedure of selecting suitable purification processes for a specific refinery off-gas is to match the range of process requirement and the stream data. This work can be done systematically by building the binary matrix Sq,p, where subscript q is the range of process requirement such as feed purity, product purity, feed pressure etc., and p represents the purification process. If the operating conditions of a gas stream meet the range of process requirement, the cell of the matrix is equal to 1, otherwise 0. The decision of whether the purification process is suitable for the specific gas stream can be made by:
Dp =

∏S
q

q, p

(3.1)

If Dp is equal to 1, purification process p can be selected. Otherwise, process p is not suitable for the purification of this stream. Beside process requirement, operational performance is also important to the appropriate choice of purification processes. Miller and Stoecker (1989), Whysall and Picioccio (1998) compared the operational factors in purification processes, such as operating flexibility, turndown ratio and reliability. The possibility of future expansion should be considered from the beginning of a project. These factors can not be quantified. Both these process and operational requirements are summarised in Table 3.4. 56

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

The general selection guidelines are experience-oriented and extensively accepted in industry. However, the answer can still be ambiguous when applying these guidelines because of the following reasons: ? The guidelines are specified by the extreme operating conditions of purification processes. There are overlaps for applying different purification processes against the same range of operating parameters. ? The match between a hydrogen-rich stream and a purification process can only deal with one stream or a group of streams in very similar process conditions. No guideline is given for the selection of purification processes in the hydrogen networks that include different streams in different process conditions

Table 3.4 Selection guide for hydrogen purification process Factors Process consideration Feed purity (v%) Maximum product purity (v%) Maximum hydrogen recovery (%) Unit hydrogen capacity Mscfd Feed pressure (psig) Product pressure (psig) Operational consideration Flexibility Turn-down (%) Reliability Other consideration By-product recovery Ease of expansion No Average Possible High Yes Low Very high 10 – 65 High High 15 – 120 High Average 10 - 80 Average >40 99.9+ Up to 90 1 – 200 150 -1,000 Approximate feed >25 98+ Up to 95 1 – 50 200 - 2,000 Much less than feed 15 - 80 97 Up to 98 10 - 75 200 - 1,200 Approximate feed PSA Membrane Cryogenic

57

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

simultaneously. ? The guidelines are only indicative. Although some of them are concluded from economic considerations, the cost comparison is excluded in the guideline. To achieve the optimal design, the evaluation of the purification processes referring to specified cases is still necessary. The accessory costs such as compressor cost, piping cost, and additional operation cost are also indispensable to consider, which are ignored by the guidelines somehow. ? To utilise the advantages of different purification processes, the possible hybrid designs are suggested. However the guidelines can not cover this issue. ? The guidelines can not help decide the process parameters of purification processes referring to specified cases. Peramanu et al. (1999) examined the PSA, membrane and gas-liquid contacting in industrial cases and determined the most suitable design by economic comparison and analysis. However, the resulting two networks and the process parameters are designed manually.

3.3.2. Other research for integration of purification process
Some research dedicates to integration of purification processes by using new technologies. Feng et al. (1998) proposed a novel process that incorporates membrane permeation into the cyclic process of PSA to improve gas separation performance. The membrane processes were analysed in varied pressures to study the feasibility of integrating permeation with cyclic PSA processes. Two configurations of process integration were discussed: membrane-assisted feed gas pressurisation and membrane-assisted co-current depressurisation. It was shown that as compared to the simple adsorption process, both the product purity and the recovery could be improved by using the integrated process. Sircar et al. (1999) investigated the integration of PSA and nanoporous selective surface flow (SSF) carbon membrane, which can selectively adsorb CO2, CH4 and CO and produce hydrogen at high pressure side of the membrane. The SSF membrane is used to recover hydrogen from PSA tail gas. The study shows that the integrated process can increase the net

58

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

hydrogen recovery to 84–85% from a hydrogen recovery value of 77-78% by the conventional PSA process. Although results demonstrate remarkable improvement by integration of purification processes, these new technologies have not been proven by industrial applications. Therefore this chapter will focus on the conventional purification processes.

3.3.3. New rules for the hydrogen pinch method
Alves (1999) analysed the installation of hydrogen purification units in hydrogen distribution networks by observing a hydrogen pinch graph with the assumption that the residue is in a very low purity level and goes to a fuel system, and proposed three possible placements for a purification unit: ? Above the pinch – both feed and product purities are higher than the pinch purity, which possibly saves hydrogen utility. ? Below the pinch – both feed and product purities are lower than the pinch purity, which gives no saving in hydrogen utility at all. ? Across the pinch – the feed purity is lower than the pinch purity and the product purity is higher than the pinch purity, which always saves hydrogen utility. More details are discussed in Chapter 1. However, the above rules are only true when the pinch purity does not change, which may actually happen after adding a purification unit. Therefore, these rules are modified as follows: ? Above the pinch - possible reduction in utility if no change on pinch purity whereas possible increase in utility if a new pinch appears due to low recovery ? Below the pinch - no reduction in utility and possible increase of utility if a new pinch appears due to low recovery ? Across the pinch - definite reduction in utility with limitation. These rules show that installing a purifier whose product purity is higher than the pinch purity can reduce the hydrogen utility. With poor recovery, the residue purity may be high enough to break the assumption that the residue goes to fuel system. To reduce

59

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

the hydrogen utility by purification, the residue purity of a purifier must be lower than the highest purity among the sources that are currently sent to a fuel system. This chapter provides a method to evaluate minimum recovery of a purifier in order to save hydrogen utility. Equations (3.2), (3.3) describe the mass balance of process sources and sinks, equation (3.4) describes the hydrogen balance of process sinks.



j

Fi , j + Fi ,Fuel = Fi*
+ FH 2U , j = F j*

?i
?j ?j

(3.2) (3.3) (3.4)

∑F ∑F

i i, j

i i, j

* * ? y* i + FH 2U , j ? y H 2U = F j ? y j

They can be further derived to equations (3.5) to (3.7).

∑∑ F
i j

i, j

+ +

∑F
i

i , fuel

=

∑F
i

i

*

(3.5) (3.6)

∑ ∑F
j

i i, j



j

FH 2U , j =



j

F j*

∑ ∑F
j

i i, j

? y i* +



j

FH 2U , j ? y H 2U =



j

F j* ? y * j

(3.7)

Adding a purification unit introduces one sink (feed) and two sources (product and residue) into a network. The mass balance of these sinks and sources can be described by equations (3.8) and (3.9):

∑ ∑

j

F product , j + ∑ j Fresidue, j + F product , fuel + Fresidue, fuel = ∑iFi , feed
F product , j ? y product +

(3.8)

j

+ F residue , fuel ? y residue

∑ F =∑ F
j i

residue , j

? y residue + F product , fuel ? y product

i , feed

? y feed

(3.9)

Considering the product from a purifier will not go to a fuel system and the residue will not feed other process sinks, Fproduct,fuel and Fresidue,j are equal to 0. By adding (3.8) and (3.9) to (3.5), (3.6) and (3.7), the mass balance and hydrogen balance of the hydrogen network with a purifier can be represented by (3.10) and (3.11):

60

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network
* i

∑ F ?∑ F
i i

i , fuel

+

∑F
j

H 2U , j

+

∑F
j

product , j

+ Fresidue , fuel =

∑ F +∑F
j * j i

i , feed

(3.10)

∑F

i i

*

? y i* ?

∑F

i i , fuel

? y i* +



j

FH 2U , j ? y H 2U +



j

F product, j ? y product

+ Fresidue, fuel ? y residue =



F * ? y* j + j j

∑F

i i , feed

? y feed

(3.11)

Fresidue,fuel can be eliminated from (3.10) and (3.11) because the right side of the equation excludes the fuel as a sink. To maintain the equality, ΣiFi,fuel must decrease by the same amount, which means other sources send less to the fuel system. If yresidue is greater than the purity level of sources formally sent to fuel, the hydrogen sent to fuel from other sources will be less than necessary in (3.11) in order to keep the balance in (3.10). This will increase the hydrogen utility because the other terms at the left side of (3.10) are constant. Vice versa, the hydrogen utility decreases. Therefore, to decrease the hydrogen utility by adding a purification unit, the product purity must be higher than the existing pinch purity and the residue purity must be lower than the highest purity of existing sources to the fuel system. Once the product purity is selected, the residue purity of a purification unit is decided by the recovery rate.
F feed = Fproduct + Fresidue F feed ? y feed ?Recovery y product 1 ? Recovery ) y residue

(3.12) (3.13)

F product =

Fresidue = F feed ? y feed (

(3.14)

By combining the above equations and eliminating the flowrate terms, equations (3.15) and (3.16) are derived.
y product = y feed ? y feed ? Re cov ery y residue + y feed (Re cov ery ? 1 ) y feed ? y residue y feed (1 ? y residue ) ≤1

(3.15)

Recovery ≥

(3.16)

61

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

The minimum recovery of a purification unit to reduce the hydrogen utility can be calculated by (3.16).

3.4. MINLP model for purifier selection and integration
3.4.1. Superstructure
To select suitable purification processes for hydrogen recovery from refinery off-gases, one should make decisions of selecting separation technology, the capacity of the unit, and process parameters etc. The possible integration of different purification processes also needs to be considered. The off-gas streams to be purified need to be decided as well as their flowrate. The aim is to find the most cost efficient purification process under certain process requirements. However, the optimum solution can only be obtained by simultaneously considering the existing trade-offs shown in Figure 3.9. The hydrogen recovery from refinery off-gases reduces the requirement of hydrogen production whereas decreases the export to the fuel system. The larger the capacity of hydrogen recovery, the more the investment and operation costs. The capacity of a purification process depends on the hydrogen recovery affected by a few process parameters. For example, high hydrogen recovery could be achieved by operating a PSA in a high pressure ratio of the feed over the residue or a membrane in high pressure ratio of the feed over the product, which may result in extra compression cost

Fuel export

Hydrogen recovery Investment cost
Figure 3.9 Trade-offs in purification process selection

Operating cost

62

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

Sources . . . . . Membrane

Sinks . . . . .

Sources . . . . . PSA

Sinks . . . . .

Figure 3.10 Superstructure to select and integrate hydrogen purification processes in both investment and operation. Trade-offs still exist even to keep the same recovery performance. For example, either the larger membrane area or the higher pressure ratio of the feed over the product benefits the hydrogen recovery, but the former results in higher membrane capital and the latter requires extra compression. Finally, the appropriate selection of purification processes depends not only on the performance of itself but also on the accessories including compressors and piping, etc. The optimum solution can only be obtained by the optimal design of overall hydrogen networks with hydrogen purification processes. The superstructure shown in Figure 3.10 has been built to consider these trade-offs simultaneously. Hydrogen producers and hydrogen consumers are decomposed as the sources and the sinks in the strategy discussed in Chapter 2. Hydrogen purification processes are installed between these sources and sinks. In order to integrate different purification processes, they are assumed to be able to feed each other. The possible physical insights discussed in 3.2.4 can be used to remove unrealistic links between purification processes. Currently, only the integration of PSA and membrane is considered because they are the most popular purification processes in refineries. 63

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

Normally cryogenic process is only cost efficient when the by-product hydrocarbon is utilised. The superstructure considers the possible network design with purification processes. The optimal design may need one purification unit – PSA or membrane, integrated system or no purification process at all. The possible matches between streams and purification units are fully covered by the superstructure. The operating parameters of purification units are treated as variables such as operating pressure, and hydrogen recovery, etc. The relationship among these parameters is considered through shortcut models of purification processes. The installation of new compressors and piping is also included in this superstructure. Some physical insights can be used to reduce the size of the superstructure, which will be discussed later. A Mixed Integer Nonlinear Programming (MINLP) model has been built to describe the superstructure mathematically.

3.4.2. Mass balance model for hydrogen consumers
In the superstructure, the hydrogen producers such as hydrogen plants and catalytic reformers can be specified as sources. The hydrogen consumers such as hydrotreaters and hydrocrackers can be treated as sources and sinks. Since a hydrogen purification unit has one inlet and two outlets, it can be decomposed as one sink and two sources. The following equations show the mass balance of these sinks and sources:

∑F
j

Sr , j

* = FSr

?Sr

(3.17) (3.18)

∑F ∑
Sk

i i ,Sk

* = FSk

?Sk

FSr ,Sk ? y* Sr + FPSAp ,Sk ? y PSAp + FPSAr ,Sk ? y PSAr ?Sk

* + FMemP ,Sk ? y MEMp + FMEMr ,Sk ? y MEMr = FSk ? y* Sk

(3.19)

Sk ∈ j;

Sr ∈ i

where Sk and Sr represent the sinks and sources decomposed from hydrogen consumers. The above equations do not apply to the hydrogen producer and the sink of fuel system whose capacities need to be optimised. Hydrogen producers and fuel 64

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

gas systems have constant operating pressure. Equation (3.17) indicates each source can only provide hydrogen with certain amount. Equations (3.18) and (3.19) indicate the requirement of both total gas and hydrogen must be satisfied. The flowrate and purity of purification units are variables involved in nonlinear terms.

3.4.3. Compressor model
Sources can only feed sinks with higher pressures through compressors. The model of compressors has been discussed in Chapter 2. If the ratio of the outlet pressure over the inlet pressure of a compressor is variable, the number of stages is an unknown integer, which gives a discrete nonlinear expression in the compressor model. This can be avoided by a new compressor model as follows:
PowerCOMP = ( cCOMP + d COMP ? P2 )q sm1 P1

(3.20)

The coefficients in (3.20) are regressed from (2.14). Figure 3.11 shows that the calculation results using (3.20) are very much agreed by the results by (2.14) when the pressure ratios are 3, 6, 9, and the number of stages are 1,2,3. Further comparison has found that if the maximum pressure ratio per stage is set as 3, the relative difference is less than 7% when pressure ratios are from 3 to 9. Because of unknown operating pressure of some sinks and sources, the MINLP model
14000 12000

Power(KW )

10000 8000 6000 4000 2000 0 0 20 40 60 80 100 120 140

Ns=1 Pout/Pin=3 Ns=2 Pout/Pin=6 Ns=3 Pout/Pin=9

Flowrate(MMscfd)

Figure 3.11 The comparison of compressor power calculation using (2.14) and (3.20) 65

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

assumes initially that there is always a compressor between each sink and source in the superstructure. Binary variables are used to indicate the existence of compressors. Therefore, the shaftpower functions of a compressor are:

PowerCOMP =

{

0 ? P2 ? ? cCOMP + d COMP ? P 1 ? ? ? ? q sm1 ?

=0 when If i,comp j =1 when If i,comp j

(3.21)

The indicative variables If depend on the pressure of sinks and sources as well as the flowrate between them. The logic relationship will be discussed in 3.4.6. Because the indicative variables are variables during optimisation, the above expressions give discrete equations to the MINLP model, which can not be programmed properly by current optimisation package. This situation can be avoided by formulating equation (3.21) into three different inequalities: From the source with known pressure to the sink with known pressure
? PSkFixed COMP ? PowerCOMP ? U COMP If SrFixed,Sk Fixed ? 1 ≥ ? cCOMP + d COMP ? PSrFixed ?

(

)

? ? FSrFixed ,SkFixed ? ? ? ? FSrFixed ,SkFixed ? ?

(3.22)

? PSkFixed COMP ? PowerCOMP ? U COMP 1 - If SrFixed, SkFixed ≤ ? cCOMP + d COMP ? PSrFixed ?

(

)

(3.23)

COMP PowerCOMP ? U COMP ? If SrFixed, SkFixed ≤ 0

(3.24)

From the source with known pressure to the sink with unknown pressure
? PSkVaried COMP ? PowerCOMP ? U COMP If SrFixed, SkVaried ? 1 ≥ ? cCOMP + d COMP ? PSrFixed ?

( (

)

? ? FSrFixed ,SkVaried ? ? ? ? FSrFixed ,SkVaried ? ?

(3.25)

? PSkVaried COMP ? PowerCOMP ? U COMP 1 - If SrFixed, SkVaried ≤ ? cCOMP + d COMP ? i PSrFixed ?

)

(3.26)

COMP PowerCOMP ? U COMP ? If SrFixed, SkVaried ≤ 0

(3.27)

66

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

From the source with unknown pressure to the sink with known pressure
? PSkFixed COMP ? PowerCOMP ? U COMP If SrVaried, SkFixed ? 1 ≥ ? cCOMP + d COMP ? PSrVaried ?

(

)

? ? FSrVaried ,SkFixed ? ? ? ? FSrVaried ,SkFixed ? ?

(3.28)

? PSkFixed COMP ? PowerCOMP ? U COMP 1 - If SrVaried, SkFixed ≤ ? cCOMP + d COMP ? PSrVaried ?

(

)

(3.29)

COMP PowerCOMP ? U COMP ? If SrVaried, SkFixed ≤ 0

(3.30)

From the source with unknown pressure to the sink with unknown pressure
? PSkVaried COMP ? PowerCOMP ? U COMP If SrVaried, SkVaried ? 1 ≥ ? cCOMP + d COMP ? PSrVaried ?

( (

)

? ? FSrVaried ,SkVaried ? ? ? ? FSrVaried ,SkVaried ? ?

(3.31) (3.32) (3.33)

? PSkVaried COMP ? PowerCOMP ? U COMP 1 - If SrVaried, SkVaried ≤ ? cCOMP + d COMP ? PSrVaried ?

)

COMP PowerCOMP ? U COMP ? If SrVaried, SkVaried ≤ 0

Only the pressure levels of sinks and sources to be optimised are set as variables in order to cut the number of variables and reduce the number of nonlinear equations. Equations (3.22) to (3.27), (3.30) and (3.33) are linear equations, others are nonlinear equations but have different types of nonlinear terms. The cost function for compressors is the same as in 2.3.1.

3.4.4. PSA shortcut model
The fundamental of PSA has been described thoroughly by Ruthven et al. (1994). The hydrogen recovery is a function of process parameters such as feed purity, product purity, feed pressure, and residue pressure, and unit features such as the phase number of equalisation, adsorbent, the number of adsorbers and so on. Modelling PSA units has been the topic of many researches (Kvamsdal and Hertzberg, 1996; Warmuzinski and Tanczyk, 1997, 1998; Malek and Farooq, 1997, 1998). These works developed models for the design of PSA units under the consideration of mass transfer, adsorption isotherms, physical property of adsorbent, adsorption cascades, number of adsorbers etc. The model parameters are verified by experiment data. These models are applicable in the design and simulation of PSA units. The 67

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

performance of bulk separation can be analysed by simulation. However, some of the data required by these models are unavailable during the conceptual design stage. The models employ partial differential equations to account for dynamics, which will cumber the optimisation by unnecessary complexity. Chung (1998) proposed a short-cut method to evaluate the performance of a PSA system by category the basic cascades into equilibrium adsorption and desorption steps. However, the vessel volume and the density need to be specified. Lacava et al. (1998) introduced the principles in PSA system design. Knaebel (1999) concluded that the product recovery of hydrogen for bulk separation during adsorption could be expressed as the following for the case of linear isotherms:
R PSA = (1 ? θ )(1 ? 1 PH y PSAf PL )

(3.34)

where PH and PL are the high and low absolute pressures of the pressure swing cycle and θ is the adsorbent selectivity which value is between 0 and unity. The expression concentrates on the contribution of adsorbent affection to a PSA system by adsorbent selectivity θ, so that other effects can be analysed easily. θ relates to flow velocity and

1 0.9 0.8 0.7

θ
0.02 0.05 0.1 0.2 0.5 0.8

Recovery

0.6 0.5 0.4 0.3 0.2 0.1 0 0 5 10 15 20 25 30 35

PH/PL Figure 3.12 PSA system - Recovery vs PH/PL (feed purity = 70%) 68

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

can be determined in a laboratory or a pilot unit. Another way to decide θ is to correlate from industrial cases. The impacts of θ on hydrogen recovery are shown in Figure 3.12 and Figure 3.13. Figure 3.12 shows that the hydrogen recovery increases when the differences between adsorbing pressure and purging pressure enlarge. The hydrogen recovery also improves while the hydrogen purity of feed to PSA increases as shown in Figure 3.13. Both figures show that the lower the value of θ the higher the hydrogen recovery. The value of θ depends on PSA design and operating. Because the maximum recovery for practical PSA units is less than 0.92, the value of θ should be greater than 0.02. Therefore the mass balance of a PSA process can be modelled as the following equations while the hydrogen recovery varies according to the feed purity, adsorption pressure and tail gas pressure:

∑F
i

i , PSAf

=

∑F
j

PSAp , j

+

∑F
j

PSAr , j

(3.35) (3.36)

∑F

i i , PSAf

? y PSAf ? R PSA =



j

FPSAp , j ? y PSAp

1 0.9 0.8 0.7

θ
0.02 0.05 0.1 0.2 0.5 0.8

Recovery

0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.2 0.4 0.6 0.8 1

Feed purity
Figure 3.13 PSA system - Recovery vs feed purity (PH/PL = 20) 69

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

R ? ? PPSAr = PPSAp ? y PSAf ? ?1 ? PSA ? ? 1?θ ?

(3.37)

∑F
i i

i , PSAf

? y PSAf =
? y PSAf =

∑F
j

PSAp , j

? y PSAp +

∑F
j

PSAr , j

? y PSAr
FSrVaried ,Memf ? y SrVaried

(3.38) (3.39)

∑F

i , PSAf



SrFixed

* + FSrFixed , PSAf ? y SrFixed



SrVaried

As discussed in 3.2.1, the product purity has little impact on recovery. The product purity is set as a constant. Because the pressure drop in a PSA is very low, the product pressure is set as equal to the feed pressure.
PPSAp = PPSAf
up PPSAp ≤ PPSAp

(3.40) (3.41)

In Figure 3.12 and Figure 3.13, hydrogen recovery will approach the maximum along horizontal axis. However, industrial experience tells us that hydrogen recovery has optimum operating pressure (Figure 3.3). Equation (3.41) applies upper bound to PSA product pressure in order to achieve realistic optimisation results. Ruthven et al. (1994) stated that the investment of a PSA system is the function of throughput. The cost equation can be found in 2.3.3.

3.4.5. Membrane model
Winston and Sirkar (1992) addressed the fundamental research in membrane separation and categorised the commercial applications of membranes. For a gas permeation membrane, gases pass through a membrane in two sequential steps: solution and diffusion, which can be described by Henry’s law and Fick’s law respectively:
Henry' s Law : C n = S n ? Pn
Fick' s Law : J n = -D n dC n dz

(3.42) (3.43)

where Jn represents the flux of component n, Cn represents the concentration, Pn represents the partial pressure, Sn represents the solubility, Dn represents the 70

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

diffusivity, z represents the membrane thickness. n is a set of the hydrocarbon components through the membrane in which only hydrogen and methane are covered in this chapter. Therefore the gas flowrate passes through a membrane can be obtained by:
J n = Lg n ? AMEM Pnfeed ? Pnproduct Lg n = S n ? Dn z

(

)

(3.44) (3.45)

Lgn represents the permeability of component n, AMEM represents the membrane area. The membrane permeability is independent from pressure and varies with specific membranes. It has been reported that Ube has developed a polymer membrane for hydrogen/methane separation with the methane permeability of 4 GPU and hydrogen permeability of 500 GPU (Winston and Sirkar, 1992). Equations (3.46) to (3.50) describe the mass balance of a membrane process considering the permeation of two components – hydrogen and methane:

∑F
i i

i , MEMf

=

∑F
j

MEMp , j

+

∑F
j

MEMr , j

(3.46)

∑F ∑F
i j

i , MEMf

? y MEMf =
? y MEMf =

∑F
j

MEMp , j

? y MEMp +

∑F
j

MEMr , j

? y MEMr
FSrVaried , MEMf ? y SrVaried

(3.47) (3.48) (3.49)

i , MEMf



SrFixed

* + FSrFixed , MEMf ? y SrFixed



SrVaried

∑F ∑F
j

MEMp , j

? y MEMp =Lg H 2 ? AMEM PMEMf ? y MEMf ? PMEMp ? y MEMp

(

) )

MEMp , j

? (1 ? y MEMp ) =Lg CH 4 ? AMEM PMEMf ? (1 ? y MEMf ) ? PMEMp ? (1 ? y MEMp )

(

(3.50)

Some pressure features are considered in a membrane system. The pressure difference between the feed and the residue pressure is ignored because of little pressure drop along a membrane. The pressure ratio of the residue over the product is limited to the tolerance of the membrane strength. The hydrogen partial pressure in the

Note :

GPU =

?6 cm ( STD ) 2 × 10 H

3

71

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

residue is greater than in the product to maintain positive permeation.
PMEMr = PMEMf PMEMr ≤ Lm MEM ? PMEMp PMEMr ? y MEMr ≥ PMEMp ? y MEMp

(3.51) (3.52) (3.53)

Uppaluri (2000) employed the cost of membranes as 400US$/m2/year. To indicate the installation cost of a membrane, the capital cost of membranes can be calculated as:
I MEM = a MEM ? If MEM + bMEM ? AMEM

(3.54)

The membrane model discussed above only covers single-stage membranes. Nevertheless, the network superstructure allows multistage membranes if a proper model is available.

3.4.6. Piping model and pressure constraints
The pipe link between each source and sink is decided by the flowrate between them. The sufficient and necessary condition of pipe link existing is that the flowrate is not zero, and vice versa. If a binary variable IfPIPE is used to indicate the existence, the relation can be expressed as:
?U i , j ≤ 0 Fi , j ? If i PIPE ,j ? ui , j ≥ 0 Fi , j ? If i PIPE ,j

(3.55) (3.56)

where Ui,j is the upper bound of Fi,j, and ui,j is the lower bound when pipe existence is considered necessary. The cost of piping installation can be found in 2.3.2, and rearranged to fit the superstructure of purification process integration as following:
0.02352 FSrFixed , j [MMscfd ] ? ? PIPE PIPE PIPE PIPE ? ? I SrFixed , j [$US ] = ? a SrFixed , j ? If SrFixed , j + bSrFixed , j ? ? PSrFixed [MPa ] ? ?
PIPE ? LengthSrFixed ,j

[m]

(3.57)

72

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

0.02352 FSrVaried , j [Mscfd ] ? ? PIPE PIPE PIPE PIPE ? ? I SrVaried , j [$US ] = ? a SrVaried , j ? If SrVaried , j + bSrVaried , j ? ? [ ] P MPa SrVaried ? ?
PIPE ? LengthSrVaried , j [m ]

(3.58)

where equation (3.57) is linear and (3.58) is nonlinear. Sometimes only a pipe link is not enough for source to feed sink. When the pressure level of a sink is higher than the pressure level of a source, the installation of a compressor between them is required. However, the existence of a compressor is more difficult to judge. Williams (1999) described the principles of how integer programming describes the logic relationships. More efforts are spent to develop such a program in this chapter. The premises of compressor installation are both the flowrate between a source and a sink is greater than zero and the pressure difference is less than zero:
= 1 and If i PIPE ,j
COMP =1 If i dp , j = 1 → If i , j

(3.59)

To define when the source pressure is lower than the sink pressure, binary variable Ifdp is equal to 1, the following formulation is developed:
P j ? Pi ? If i ,dp j ? U dp ≤ 0 P j ? Pi + (1 ? If i ,dp j ) ? U dp ≥ u dp

(3.60) (3.61)

Thus equation (3.59) can then be reformulated by integer programming as follows:
? If i dp If i COMP ,j ,j ≤ 0 ? If i PIPE ≤0 If i COMP ,j ,j
PIPE ? If i COMP ≤1 If i dp , j + If i , j ,j

(3.62) (3.63) (3.64)

3.4.7. Objective function
To consider the trade-offs existing in purification process selection and integration simultaneously, the optimum design solution will provide the minimum Total Annual Cost (TAC) that includes operation costs and annualised capital costs: 73

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

TAC = Cost H 2 ? Cost Fuel + Cost Power + Af ?

(∑

I COMP COMP

+ I PSA + I MEM +



PIPE

I PIPE

)

(3.65)

where Af is the annualising factor. Smith (1995) addressed Af as: Af = fi ? (1 + fi ) (1 + fi )ny ? 1 (3.66)

where fi is the fractional interest rate and ny is the number of years. The cost of hydrogen imported in a network is calculated by summarising the expense of hydrogen generation and buying hydrogen from a thirty party.
Cost H 2 = oy ?



j

FH 2U , j ? PI H 2U

(3.67)

where oy is the annual operating hours. The credit of fuel gas is obtained by heat value calculation:
CostFuel = oy ? FH2U ,Fuel ? HVH2 + FCH4 ,Fuel ? HVCH4 ? PI HV

(

)

(3.68)

The shaftwork costs of compressors are calculated as :
Cost Power = oy ?



COMP

PowerCOMP ? PI Power

(3.69)

The heating steam for membrane pretreatment is not taken account due to the lack of information. Other utility costs such as instrument air etc. are negligible in the calculation.

3.5. Solution procedure
3.5.1. Mixed integer and nonlinear features in the superstructure model
Hydrogen network design with appropriate purification processes involves the following optimisation procedure:
Objective function : minimise TAC

under certain constraints........

74

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

The constraints are all the equations discussed from 3.4.2 to 3.4.6. There are nonlinear terms in (3.19)) of the mass balance model for hydrogen consumers, (3.25), (3.26), (3.28), (3.29), (3.31), (3.32) of the compressor model, (3.36), (3.37), (3.38), (3.39) of the PSA model, (3.47), (3.48), (3.49), (3.50),(3.53) of the membrane model, and (3.58) of the piping model. It should be noticed that all the nonlinear terms are in bilinear expression or can be converted to a bilinear form. Integer variables are used to indicate the existence of piping, compressors or purification units, which can be found from (3.22) to (3.33), and cost equations, from (3.54) to (3.58), and from (3.60) to (3.64). Therefore the overall superstructure model leads to a mixed integer nonlinear (bilinear) programming (MINLP) optimisation problem.

3.5.2. Linear relaxation of nonlinear terms
To solve an MINLP problem, it requires a large amount of computation efforts. The features of MINLP problems and the principles of algorithms to solve MINLP problems are introduced by published literature (Biegler et al 1997, Adjiman et al, 1998). One of the critical methods to improve an MINLP optimisation procedure is appropriate initialisation of the problem. For a nonconvex MINLP problem, the global optimum is not guaranteed. A high quality starting point can help find a good local optimum. In many of the cases, a converged solution can only be obtained through good problem initialisation. Quesada and Grossmamm (1995) developed a method to deal with the global optimisation of systems consisting of bilinear terms. In their approach, a non-convex bilinear term can be relaxed as four linear inequalities. Thus, the problem can be turned to a relaxed linear problem. The more detail of this algorithm is discussed in section 2.3.4. The bilinear terms formed by three variables are relaxed in two steps: firstly, they can be relaxed as bilinear terms with two variables by treating the product by multiplication of two variables as a new variable; the new variable then can be relaxed in the above procedure afterwards. For example, equations (3.36), (3.37) in the PSA shortcut model contain nonlinear terms that can be relaxed to generate linear equations: 75

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

Equation (3.36) becomes:
FyR PSAf =



j

FPSAp , j ? y PSAp

(3.70) (3.71) (3.72) (3.73) (3.74) (3.75) (3.76) (3.77) (3.78) (3.79)

Low Low Low Low ? RPSA + Fy PSAf ? RPSA ? Fy PSAf ? RPSA FyR PSAf ≥ Fy PSAf

Up Up Up FyR PSAf ≥ Fy Up PSAf ? R PSA + Fy PSAf ? R PSA ? Fy PSAf ? R PSA

Low Up Low Up ? RPSA + Fy PSAf ? RPSA ? Fy PSAf ? RPSA FyR PSAf ≤ Fy PSAf

Low Up Low FyR PSAf ≤ Fy Up PSAf ? R PSA + Fy PSAf ? R PSA ? Fy PSAf ? R PSA

Low Low Low Low ? y PSAf + FPSAf ? y PSAf ? FPSAf ? y PSAf Fy PSAf ≥ FPSAf

Low Low Low Low ? y PSAf + FPSAf ? y PSAf ? FPSAf ? y PSAf Fy PSAf ≥ FPSAf

Low Low Up ? y PSAf + FPSAf ? y Up Fy PSAf ≤ FPSAf PSAf ? FPSAf ? y PSAf

Up Low Up Low ? y PSAf + FPSAf ? y PSAf ? FPSAf ? y PSAf Fy PSAf ≤ FPSAf

FPSAf =

∑F
i

i , PSAf

Equation (3.37) becomes:
RPry PSA = (1 ? θ )(Pry PSA ? 1)
Low Low Low Low ? Pry PSA + RPSA ? Pry PSA ? RPSA ? Pry PSA RPry PSA ≥ RPSA

(3.80) (3.81) (3.82) (3.83) (3.84) (3.85) (3.86) (3.87)

Up Up Up RPry PSA ≥ R Up PSA ? Pry PSA + R PSA ? Pry PSA ? R PSA ? Pry PSA

RPry

PSA

Low Low Up ≤ R PSA ? Pry PSA + R PSA ? Pry Up PSA ? R PSA ? Pry PSA

Up Low Up Low ? Pry PSA + RPSA ? Pry PSA ? RPSA ? Pry PSA RPry PSA ≤ RPSA

Low Low Low Low ? y PSAf + PrPSA ? y PSAf ? PrPSA ? y PSAf Pry PSA ≥ PrPSA

Up Up Up ? y PSAf + PrPSA ? y Up Pry PSA ≥ PrPSA PSAf ? PrPSA ? y PSAf

Low Low Up ? y PSAf + PrPSA ? y Up Pry PSA ≤ PrPSA PSAf ? PrPSA ? y PSAf

76

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

Up Low Up Low ? y PSAf + PrPSA ? y PSAf ? PrPSA ? y PSAf Pry PSA ≤ PrPSA

(3.88) (3.89) (3.90) (3.91) (3.92)

Low Low Low Low ? PrPSA + PPSAr ? PrPSA ? PPSAr ? PrPSA PPSAp ≥ PPSAr

Up Up Up Up ? PrPSA + PPSAr ? PrPSA ? PPSAr ? PrPSA PPSAp ≥ PPSAr

Low Up Low Up ? PrPSA + PPSAr ? PrPSA ? PPSAr ? PrPSA PPSAp ≤ PPSAr

Low Up Low Up ? PrPSA + PPSAr ? PrPSA ? PPSAr ? PrPSA PPSAp ≥ PPSAr

Through the same procedure, the original MINLP model becomes an MILP model. The upper bounds and lower bounds of variables can be set as the physical limitation of networks. The process features of purification processes are analysed in 3.2. Table 3.4 lists the selection guide from industrial experience for purification processes. Applying these criteria to define the variable bounds has the following advantages: ? Improve the solvability of MINLP problems. Generally the bound definition by physical insights gives a narrower solution space than by network limitations, thus speeds up the optimisation. ? Overcome model shortcomings. Purification process models can be constricted in the viable regions. ? Achieve realistic solutions.

3.5.3. Solution procedure
Bilinear terms cause non-convexity in optimisation. Attempts to solve the MINLP problem explained in 3.4 have shown peculiar difficulty to converge. A novel initialisation procedure is applied to the MINLP model of purification process selection and integration. The idea is to convert original MINLP models into MILP models by exploiting linear relaxation of bilinear forms. The solution of MILP provides initial values of the original MINLP problems. The method extends the linear relaxation of mixer or splitter to the models of compressors and purification processes. Based on the procedure of linear relaxation, the original MINLP problem can be converted to an MILP problem with the same linear objective function. This problem is optimised and 77

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

Relax network models to linear models and solve problem by Mixed-Integer Linear Programming (MILP)

Use MILP solution as initialisation

Solve problem by Mixed-Integer Non-Linear Programming (MINLP)

Evolve design by merging compressors

Figure 3.14 Solution procedure for superstructure MINLP model

the solution for all the variables is then used as the initial value for the original MINLP problem. The MINLP problem is solved afterwards. The solution procedure is shown in Figure 3.14 . Merging some compressors can evolve the optimal design. The motivation is to reduce the investment of compressors and piping as well as simplify the network. If a retrofit problem is presented, existing compressors can be exploited to eliminate the need of

37.32 Mscfd 99.9%

H2 import
26.35 Mscfd 99.9% 8.09 Mscfd 99.8% 55.58 Mscfd 99.8%

Unit A

Unit B

3.09 Mscfd 91%

38.67 Mscfd 80%

12.32 Mscfd 37.42%

PSA

Fuel

Figure 3.15 Existing hydrogen network with PSA unit 78

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

installation of some new compressors. The operational flexibility of existing compressors and the scheme of unit start-ups and shutdowns should be considered to reuse these compressors.

3.6. Case studies and analysis
3.6.1. Purification process integration for retrofit
This example demonstrates how the integration of membranes and PSAs provides promising retrofit options for hydrogen networks. An existing hydrogen network with two hydrogen consumers and one PSA unit is illustrated in Figure 3.15. Currently 38.32 Mscfd (4.28 x 104 Nm3/h) hydrogen in 99.9% is imported. The purge stream from unit A

Table 3.5 Process data for existing hydrogen network with PSA unit Make-up
Mscfd

Recycle
Mscfd

Purge
Mscfd

%

%

%

Unit A Unit B

8.09 55.58

99.9 99.9

31.91 61.33

91.0 80.0

3.09 38.67

91.0 80.0

Sink
Mscfd

Source
MPa
Mscfd

%

%

MPa

Unit A Unit B H2 import PSA feed PSA product PSA residue

40 120

92.8 89.5

15.2 11.03

35 100 37.32

91.0 80.0 99.9

11.7 9.0 2.5

38.67

80.0

2.6 26.35 12.32 99.9 37.42 2.5 0.15

79

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

is reused as one of the make-up streams to unit B. The purge stream from unit B is purified by the PSA unit that provides 26.35 Mscfd (2.94 x 104 Nm3/h) hydrogen in 99.9%. The PSA unit has 85% hydrogen recovery. The PSA product is mixed with hydrogen import as the make-up to the hydrogen consumers, while the residue is compressed to a fuel system. The pressure level of the fuel system is 0.6 MPa. Table 3.5 lists process data for the hydrogen network. The target of retrofit is to reduce the system operating cost including hydrogen import, and compression, etc. Through observation and the pinch analysis, the existing distribution network is found in proper design giving the minimum hydrogen utility so far. The possible retrofit options are focusing on improving the hydrogen recovery. To replace the existing PSA with a new PSA in higher recovery or to revamp the existing PSA requires large investment and prolonged shutdown period. The integration of a membrane with the PSA to improve recovery has been discussed in 3.2.4. In this case study, a new membrane is designed to increase the total recovery of purification. The new design is shown in Figure 3.16 . The purge stream from unit B is pre-purified by the membrane. The membrane permeate is used as the make-up of unit B, and the membrane residue is taken as the feed to the PSA unit. The pressure level of the membrane permeate is designed as 2.5 MPa to reuse the

34.17 Mscfd 99.9%

H2 import
6.25 Mscfd 99.9% 8.09 Mscfd 99.9%

Unit A

Unit B
23.49 Mscfd 99.7% 3.09 Mscfd 91% 38.91 Mscfd 80% 15.42 Mscfd 50% 9.17 Mscfd 16.01%

MEM

PSA
Figure 3.16 New design with the integration of membrane and PSA

Fuel

80

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

existing compressor. The pressure level of the membrane inlet is the same as the pressure of the purge stream from unit B to take advantage of its high operating pressure. This design is then optimised by using the models discussed in 3.4. The objective function is to minimise total annual cost. The cost of hydrogen import is 2000 US$/Mscf. Electricity costs 0.03 US$/kWh, and fuel costs 2.5 US$/MBTU (8.53 US$/MWh). The hydrogen permeability in the membrane is set as 500 GPU that is equal to 125 times of the impurity permeability. The annual capital cost of the membrane is calculated as 400 US$/m2 with 50000 US$ installation charge. The capital cost is annualised in two years, with 5% interest rate. The piping distances between unit B and the membrane, the membrane and the PSA are 200 m and 50 m respectively. Once the topology of network has been determined, the problem can then be simplified to an NLP problem. The lower bound of the PSA inlet purity is defined as 50%. The optimal solution shows that all the purge streams from unit B is pre-purified by the membrane.

Table 3.6 TAC comparison for existing network and optimum design Existing network (MUS$) Operation cost Hydrogen import Electricity Fuel export Capital cost Membrane Piping TAC 20.889 20.889 27.241 2.561 -8.913 Optimum design (MUS$) 19.516 24.946 2.505 -7.935 0.568 0.369 0.199 19.821

81

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

The total annual cost is compared between the existing network and the optimal design in Table 3.6 The total annual cost of optimal design is 1.068 MUS$ less than the existing network. The optimal design costs 2.295 MUS$ per year less on hydrogen import than the existing network and exports 0.978 MUS$ per year less fuel, saves totally 1.373 MUS$ per year in operation costs. The saving in electricity is achieved from less fuel export when the capacity of recycle compressor is increased slightly by 0.4%. Although the installation of the membrane and piping costs 0.568 MUS$, the simple payback period is only 5 months. The total investment is only 4.07% of installing a new PSA (13.94 MUS$ for PSA only) to replace the existing one. Apparently, this option is also competitive to revamp the existing PSA. Because the membrane residue feeds the PSA, the purity decrease of the PSA inlet worsens the hydrogen recovery in the PSA. However, taking the membrane permeate as the make-up hydrogen improves the overall hydrogen recovery of the whole system from 85% to 95.8%. The sensitivity analysis is carried out for the purification capacity of the membrane. Figure 3.17 shows the results. While the purification capacity of the membrane increases, the operating costs decrease and the investment rises. The total annual cost of the design with a membrane is lower only when the membrane capacity reaches a certain level. In this case study, integration of the membrane and the PSA improves the system performance in the following aspects: ? Reduce hydrogen import ? Increase total hydrogen recovery by purification ? Less fuel export leads to less compression cost. If fuel gas in a refinery is redundant, such integration gives more benefits. ? Short shutdown period for revamp ? Small investment and short payback time.

82

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

30 28 26

Cost (MMUS$)

24 22 20 18 16 14 12 0 10 20 30 40

Operation costs MMUS$ (Capital costs+15) MMUS$ TAC MMUS$

Flowrate through membrane (MMscfd)

Figure 3.17 Sensitivity analysis for membrane purification capacity

3.6.2. Selection and optimisation of purification process
A Petro-Canada industrial example presented by Peramanu et al. (1999) is used as the base case in the second case study. The simplified process diagram in Figure 3.18
33.71 Mscfd 99.9% 2.5 MPa 192.77 Mscfd 94.4% 17.58 Pa

Gas Make

Hydrocracker

HP Cold Separator

Fuel 4.29 Mscfd 94.4% 17.58 MPa

Consumption

Water LPHot Separator

Gas Make

Naphtha Reformer LP Cold Separator

3.25 Mscfd 87.4% 2.21 MPa Fuel 9.99 Mscfd 83% 1.38 MPa Fuel

Water

Figure 3.18 Petro-Canada base case 83

Peramanu et al. (1999)

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

shows a hydrocracker. The make-up hydrogen comes from import. The high-pressure purge from the hydrocracker is partly recycled, and the rest is released as fuel gas. Beside the high-pressure purge, other available sources are the low-pressure purge from the hydrocracker, and the hydrogen-rich purge from a catalytic reformer. After decomposition of the network, the process data of sinks and sources are listed in Table 3.7. A hydrogen purification process is utilised to reduce hydrogen import. Because the pressure range of the refinery off-gas is varied from 17.58 MPa to 1.38 MPa, both a membrane and a PSA are considered as candidates. The purity of the refinery off-gas also is high enough to suit the requirement of the membrane and the PSA. Regarding the trade-offs between the selection of purification processes and network design discussed in 3.4.1, the superstructure is built to include the possible links between offgas streams and both purification applications. The selection of purification processes is the procedure to achieve the optimal network design. The objective function is represented by the total annual cost in order to consider the investment penalty during saving hydrogen import. The installation of piping, compressors, and purifiers is identified by integer variables. The flowrate of each refinery off-gas to a purification process is to be determined by optimisation. The operation parameters of each purifier are variables too. The varied operating pressure of purifiers determines the existence of new compressors. All the variables will be

Table 3.7 Process data of sinks and sources

Source Hydrogen Import HP Purge LP Purge Reformer Purge Sink Hydrocracker

Flowrate MMscfd 33.71 197.06 3.25 9.99

Purity 99.9% 94.4% 87.4% 83% 95.22%

Pressure MPa 17.58 2.21 2.21 1.38 19.00

226.48

84

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

decided simultaneously to achieve the optimal network design with appropriate purification processes. The cost of utilities and the investment coefficients are the same as the first case study. The investment coefficients of existing compressors and piping are fixed as zero. The capital cost is annualised in five years, with 5% interest rate per year. The highest operating pressure for PSA is assumed at 4.0 MPa. The piping distances between each unit are presumed in Table 3.8. The solution procedure of optimisation in Figure 3.14 is followed. The GAMS system developed by GAMS Development Corporation provides the access to both MILP and MINLP solvers. Brooke et al (1996) discussed GAMS in details. The algorithm and features of solvers can be found in WWW.GAMS.COM. OSL is used as the MILP solver and DICOPT is used as the MINLP solver. The statistics of the MINLP model shows 415 equations, 247 continuous variables and 96 binary variables. DICOPT takes 33.8 CPU seconds (Pentium II Celeron 300A) to converge to the solution in Figure 3.19. Table 3.9 lists the executive results.

Table 3.8 Distances between sources and sinks (m) HC H2 import HP purge LP purge Ref purge PSA product PSA residue MEM permeate MEM residue 500 50 50 500 500 500 500 500 PSA inlet 500 500 500 500 50 50 500 500 85 MEM inlet 500 500 500 500 200 200 50 50 Fuel 500 500 500 500 500 500 500 500

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

Table 3.9 DICOPT executive results to converge DICOPT Log File Major Major Step NLP MIP NLP MIP NLP MIP NLP MIP NLP Iter 1 1 2 2 3 3 4 4 5 Objective Function 11628.88548 14272.09094 17190.51024< 17004.47589 16958.23529< 17014.76760 16958.01104< 17014.76760 17030.94397 NLP = NLP = CPU time (Sec) 1.20 2.47 0.83 2.64 0.61 15.99 0.60 9.07 0.39 3.63 10.75 Itera- Evaluation Solver tions 67 431 84 712 63 1605 68 1098 54 MIP = MIP = Errors 0 0 0 0 0 0 0 0 0 30.17 89.25 conopt osl conopt osl conopt osl conopt osl conopt

Total solver times : Perc. of total :

19.41 Mscfd 99.9% 2.5 MPa 192.50 Mscfd 94.4% 17.58 MPa Gas Make Hydrocracker HP Cold Separator

14.56 Mscfd 99.8% 3.24 MPa

11.61 MPa

Consumption

Water LPHot Separator

Gas Make

Naphtha Reformer LP Cold Separator 3.25 Mscfd 87.4% 2.21 MPa

Fuel 3.24 Mscfd 27.85% 11.61 MPa 9.99 Mscfd 83% 1.38 MPa

Water

Figure 3.19 Optimum network design with membrane unit 86

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

A new membrane with relevant compressors and piping is selected in the optimal solution. The high-pressure and low-pressure purge streams from the hydrocracker, and the purge stream from the catalytic reformer are purified by a membrane. The inlet purity of the membrane is 87%. The 11.61 MPa inlet pressure of the membrane, while the pressure of the permeate is 3.24 MPa, gives pressure ratio of the feed over the permeate in 3.58. The area of membrane is calculated as 183.42 m2. Upon the above design, the membrane system can achieve high hydrogen recovery as 0.94. The design saves 14.3 Mscfd (1.60 x 104 Nm3/h) in the hydrogen import (42.4%), 5.694 MUS$ in operation cost (27.5%), and 3.731 MUS$ on TAC even considering new capital cost (18%). Sometimes the MINLP optimisation converges to a local optimum. To check the optimality of this solution, several scenarios are taken into account as competitive designs (Figure 3.20). Scenario A has only a PSA. In scenario B, all purge streams are

HC HPP LPP RefP

HC HPP LPP HC

PSA
Fuel

RefP

PSA Scenario B
Fuel

Scenario A

HC

HC

HPP LPP

HC

HPP

Fuel HC

LPP

PSA
RefP Fuel RefP

PSA
Fuel

Scenario C

Scenario D

Figure 3.20 Competitive design scenarios for purification process design 87

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

purified by a membrane first, and the residue of the membrane is purified by a PSA afterwards. Scenario C shows again the hybrid system of a membrane and a PSA, while the low-pressure streams are mixed with the membrane residue and sent to the PSA. Scenario D has a membrane and a PSA that deal with the high- pressure purge and low-pressure purge separately. For each scenario, only the topology of purification process design is fixed by presuming some binary variables. The match between streams and units, unit capacity and necessity of new compressors and piping are still determined by optimisation, which still leads to MINLP optimisation. The comparison among the base case, the optimal solution and the competitive designs is shown in Table 3.10. Although all the competitive designs have less TAC than the base case, the optimal design obtained from the direct solution enjoys the most benefits. Data indicate that in this case study the trade-off between hydrogen

Table 3.10 TAC comparison for Petro-Canada case study MUS$ Base Case TAC Operating cost H2 import Electricity Fuel export Capital cost Membrane PSA Compressor Piping 20.688 20.689 24.603 3.075 -6.990 Optimum Design 16.958 14.995 14.172 3.372 -2.549 8.497 0.368 6.837 1.292 Scenario A 18.745 15.271 14.847 3.261 -2.837 15.039 6.615 6.672 1.752 Scenario B 17.563 14.993 14.186 3.362 -2.555 11.127 0.387 0.622 8.294 1.824 Scenario C 18.516 15.447 15.159 3.258 -2.970 13.288 0.099 5.215 5.807 2.167 Scenario D 18.471 15.083 14.556 3.240 -2.713 14.666 0.099 5.104 7.373 2.089

88

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

import and fuel export dominates the economic performance of each scenario. Therefore the PSA residue is compressed to the fuel system in all the designs. The operating cost saving is in the same level. The selection of purification scenario is affected by the investment. In the circumstance when different designs give similar savings, the winning scenario will be the one with top process reliability and flexibility. The optimal design assumes that each new link requires compressor installation if pressure match requires so. Once the optimal solution is obtained, some of the new compressors can be combined and existing compressors can be utilised to reduce capital costs and simplify the design. However, there are still trade-offs between operation cost and capital cost on compressors. The improved design for this case study can be found in Figure 3.21. The design uses existing compressor to boost the membrane permeate under the assumption that the compressor for hydrogen import is able to be reused and has at least 0.8% capacity margin. Two new compressors for the low-pressure purge from hydrocracker and purge from catalytic reformer in optimal design are substituted by one compressor with larger capacity. The relaxed optimal design reduces the capital cost by 3.717 MUS$ and rises the operating cost by 0.08 MUS$. TAC is improved in 16.179 MUS$ after
19.41 Mscfd 99.9% 2.5 MPa 192.50 Mscfd 94.4% 17.58 MPa Gas Make Hydrocracker HP Cold Separator 11.61 MPa

14.56 Mscfd 99.8% 3.24 MPa

Consumption

Water LPHot Separator

Gas Make

Naphtha Reformer LP Cold Separator 3.25 Mscfd 87.4% 2.21 MPa

Fuel 3.24 Mscfd 27.85% 11.61 MPa 9.99 Mscfd 83% 1.38 MPa

Water

Figure 3.21 Improved design by merging compressors 89

Chapter 3

Strategy of Purifier Selection and Integration in Hydrogen Network

design evolution. The difference on piping installation is negligible.

3.7. Conclusions
A systematic methodology is developed in this chapter for the purpose of automated network design associated with the optimal selection of purification processes. The superstructure-based model is built upon understanding the trade-offs in selection of purification processes. The models of purification processes are developed. The recovery and capital cost of purification processes are related to their operating conditions. The pressure difference and the match between a source and a sink justify the necessary installation of new compressors. The superstructure covers the scenarios of a single purification process and a hybrid system. The final design is obtained automatically through optimisation. An MINLP algorithm is applied to solve the problem. A solution procedure is proposed to improve the solvability. The MINLP problem is relaxed to an MILP problem whose solution initialises the original problem. The optimisation experience notices that this MINLP problem with initialisation is easier to converge than the problem without such initialisation. The physical insights of the network and purification processes can help tighten solution space and achieve realistic design by confining variable bounds. The optimisation solution provides the strategies for the refinery off-gas purification as well as appropriate network designs. To consider the benefits and penalty in network design, the objective function is to minimise the operating cost and the annualised capital cost simultaneously. The method evaluates the purification scenarios quantitatively rather than qualitatively in traditional approaches. The case study shows impressive cost savings in retrofit projects. However, the method is also suitable for new designs of distribution network of refinery off-gases.

90

Chapter 4

Integration of Hydrogen Generation and Hydrogen Recovery

Chapter 4: Integration of Hydrogen Generation and Hydrogen Recovery
4.1. Introduction
When hydrogen consumers with large hydrogen requirement e.g. hydrocrackers are operating in a refinery, the supplemental hydrogen often needs to be provided by hydrogen generation processes besides catalytic reforming units. To produce hydrogen by a steam reforming process, the available feedstock in a refinery can be methane, saturated LPG, straight-run naphtha and some refinery off-gas. In Chapter 3, the hydrogen recovery from refinery off-gas is discussed. The management of hydrogen generation and hydrogen recovery depends on the refinery hydrogen balance, raw material prices and fuel prices of hydrogen generation, total operating cost and investment incurred. The methodology for integration of hydrogen generation and hydrogen recovery is proposed in this chapter. The hydrogen plant is modelled by means of process data from comprehensive simulation. A superstructure is then built to include the operation situation of a hydrogen plant with PSA purification and a separated PSA. The refinery off-gas is evaluated as the feed of both the hydrogen plant and the PSA unit. The optimisation results provide improved operations. An algorithm is also developed to find the global optimum solution. There are two major parts in this chapter. The first part is covered by sections 4.2 and 4.3. The process details of hydrogen plant are introduced in section 4.2. Section 4.3 builds up a comprehensive process simulation for hydrogen plant. The linear model of hydrogen plant is then developed and tuned based on the simulation results. The second part discusses the integration of hydrogen plant and PSA. In section 4.4, the linear model of hydrogen plant as well as the models for PSA and compressor are embedded into the superstructure of hydrogen networks. A solution procedure is developed in order to find global optimum solution for operational problems. Those methods are then applied in case studies in section 4.5.

91

Chapter 4

Integration of Hydrogen Generation and Hydrogen Recovery

4.2. Hydrogen plant process introduction
Hydrogen generation processes in a refinery are steam reforming and partial oxidation. Gardner (1998) addressed the distinguishable factors of hydrogen generation processes. For gaseous feedstock, the capital costs inside battery limits for steam reforming and partial oxidation are similar. However, the high-pressure oxygen requirement adds extra expense to partial oxidation. For heavy feedstock, the capital cost for partial oxidation is significantly higher than for gas-based steam reforming due to the requirement for sulphur removal and feedstock handling facilities. Except the cases that gaseous feedstock is extremely expensive and heavy feedstock is very cheap, high capital cost for partial oxidation processes plus oxygen supply requirements make the hydrogen cost substantially higher than that from steam reforming processes. Because of the capability to handle heavy feedstock, partial oxidation process is an opportunity to overcome bottom-of-barrel problems in refineries, whereas the amount of available heavy feedstock in refinery typically far exceeds the required amount to satisfy the hydrogen balance. Therefore, alternative consumers of hydrogen or synthesis gas need to be found to make this option viable. Currently, steam reforming dominates the hydrogen generation processes in refineries. There are two typical processes in steam reforming with different purification processes. In conventional processes, the low-temperature shift reactor is necessary to reduce residual carbon monoxide in synthesis gas further after the high-temperature shift conversion. The product is then cooled and scrubbed with circulating amine or hot potassium carbonate solution to remove carbon dioxide by absorption. The remaining small quantities of carbon monoxide and carbon dioxide are converted to methane by methanation. The hydrogen purity in the conventional processes is typically 95% to 97%. The impurities are mainly methane and carbon oxidant in ppm level. The pressure swing adsorption (PSA) was introduced as replacement of the conventional purification process in the 1970s. The high-temperature shift gas is cooled and purified by PSA. The hydrogen purity by PSA purification can reach 99.999%. The residue of PSA is recycled to the reformer as fuel. Because a part of generated hydrogen is discharged as residue in PSA, the hydrogen yield of unit feedstock with PSA is lower than with conventional processes, but needs less external fuel. The investment of hydrogen plants with PSA is higher than the conventional plants regarding 92

Chapter 4

Integration of Hydrogen Generation and Hydrogen Recovery

to the same hydrogen production. The economics of hydrogen generation depends on the price of feedstock and fuel. While feedstock is expensive and fuel is relatively cheap, the conventional process shows lower hydrogen cost, or vice versa. However, the hydrogen plants with PSA become a preferred choice of hydrogen generation in refineries nowadays due to the following reasons: ? Hydrogen purity has significant impacts on the design and operation of hydroprocessors. High purity is favourable to hydroprocessor design (Hiller et al, 1983). ? The state-of-art PSA technology provides high hydrogen recovery up to 90% (Vervalin, 1998), which enormously reduces hydrogen cost. ? A PSA process is easier to design and operate for a large-scale hydrogen plant than an adsorption process which normally uses packing columns. ? A hydrogen plant with PSA has low turndown ratio, high operation flexibility, modest control, short start-up and turn down period, long-term operation and relative ease of

Figure 4.1 Typical steam reforming hydrogen plant
? Copyright 1998, Haldor Tops?e A/S, http://www.haldortopsoe.dk/Technologies/Hydrogen/

93

Chapter 4

Integration of Hydrogen Generation and Hydrogen Recovery

expansion. The process characters and variables for different hydrogen generation can be found in many publications (Comstock, 1980; Gary and Handwerk, 1994; Armor, 1999). In this chapter, the steam reforming with PSA is taken as a default process of hydrogen plant. Such a hydrogen plant consists of feed preparation, steam reforming, high-temperature shifting, PSA, heat recovery and steam generation. Figure 4.1 shows a typical process with a pre-reformer.

4.2.1. Available feedstock in refinery and pre-treatment
Steam reforming refers to the reaction between hydrocarbon and water to generate hydrogen. However, not all the hydrocarbons in refineries are suitable to be used as the feedstock of steam reformer. While reforming reaction happens, hydrocarbons crack to liberate hydrogen and lay down carbon on catalyst. In gaseous hydrocarbon, olefins are much easier to crack and lay down carbon which causes catalyst coking. Therefore, streams containing olefin in high level, such as off-gases from FCC and DCU are excluded from the feedstock of hydrogen plants. To exploit these gas streams, olefins can be saturated in a pre-hydroprocessing reactor with special design to take away heat generated by olefin saturation. The qualified gaseous streams to feed a hydrogen plant are refinery off-gases from hydroprocessing units and CCR. Natural gas is also suitable if sufficient supply is available. The same thing happens to heavier hydrocarbon. The available feedstock to a hydrogen plant is saturated LPG and light straight naphtha. The hydrocarbons with low molecule weight are favourable to steam reforming processes not only because of high hydrogen yield but also because they have lower rate of carbon formation. The feedstock needs pre-treatment to remove sulphur, chlorine and other impurities that will poison steam reforming catalyst. The feedstock hydrotreating converts them into hydrides by the aid of cobalt/molybdenum or nickel/molybdenum catalyst and resists them by adsorption. A small amount of product hydrogen is recycled to a hydesulphurisation reactor if there is not adequate hydrogen in the feedstock.

94

Chapter 4

Integration of Hydrogen Generation and Hydrogen Recovery

4.2.2. Steam reforming
The pre-treated feedstock is mixed with super-heated steam and preheated to 500°C (932°F) before passing to the radiant section of the reformer. The steam reforming reaction takes place in the furnace tube packed with nickel-based catalyst. The steam reforming reaction is not just one reaction but can be described briefly by the following equations:
CH 4 + H 2 O ? CO + 3H 2 ? ?H m )H 2 2 ? ?H

(4.1) (4.2) (4.3)

C n H m + nH 2 O ? nCO + ( n + CO + H 2 O ? CO 2 + H 2

+ ?H

Although the reactions in equations (4.1) and (4.2) are endothermic and the reaction in equation (4.3) is exothermic, the overall reaction is highly endothermic. The burners in the furnace radiant section provide heat to maintain operating temperature of steam reforming to more than 800°C (1472°F). Addition of excess steam not only shifts the reaction equilibrium to produce more hydrogen but also avoids carbon formation in catalyst. However, high steam import increases the duty of a steam reformer. The steam to carbon ratio in industrial operations is typically between 2.5 and 3.5 depending on the feedstock and the process. Low operating pressure promotes hydrogen generation, but most of hydrogen plants with PSA operates above 20 atm (2.03 MPa), based on the consideration of the optimum PSA operating pressure, compression cost and equipment size.

4.2.3. Reformed gas shifting
The temperature of the gases exiting from a reformer is reduced to 340-360°C (644680°F) by generating steam. Carbon monoxide is converted further to hydrogen in a high-temperature shifting reactor on magnetite iron oxide based catalyst. Equation (4.3) shows the reaction procedure. Normally a low-temperature shifting reactor is only employed in the conventional processes.

95

Chapter 4

Integration of Hydrogen Generation and Hydrogen Recovery

The gases exiting from the shifting reactor are cooled in stages to generate steam, preheat boiler feed water (BFW), and finally by air cooling and water cooling to normal temperature 40°C (104°F). The gases are then purified by PSA. The condensed sour water is recycled to the boiler after stripping.

4.2.4. PSA and fuel system
The process characteristics of PSA are discussed in Chapter 3. The adsorption forces for the components in the shift gas are water >> carbon dioxide > carbon monoxide > methane > hydrogen on the molecule sieve adsorbent. The product of PSA is hydrogen in 99.9+%. The impurities in the product are mainly methane with carbon oxides in ppm level. The rest components appear in the residue. Because the discharge pressure of PSA is very low (about 1.5 atm or 0.15 MPa) in order to achieve high hydrogen recovery, the residue from PSA is sent back directly to the steam reformer with specific low operating pressure burners as fuel instead of compressing to a fuel system. The additional fuel is supplied by a refinery fuel system. The mixture of fuel gas is mixed with air preheated in the convective section of the reformer furnace and sent to burners in the radiant section. The high-temperature flue gas enters the convention section, heats the reformer feed, superheats exporting steam, generates steam and heats combustion air separately.

4.2.5. Steam generation system
Hydrogen plants with steam reforming generate steam to supply the reaction and recover energy. Boiler feed water is pumped and preheated by the shift gas before entering the boiler. The steam generation is completed by exchanging heat with the reformed gas, the shift gas and the flue gas. The steam generated in a hydrogen plant is always more than the reaction requirement. The extra is exported to a refinery steam system after super heating.

4.2.6. New trends in process design of steam reforming
Johansen et al (1992) briefly reviewed the state of the art of hydrogen production technology and outlined certain design concepts that are expected to set the trend for 96

Chapter 4

Integration of Hydrogen Generation and Hydrogen Recovery

future hydrogen plant design. The integrated designs of steam reformers with Potformers and Gas Heated Reformer (GHR) are addressed to utilise the energy from reformed gas, reduce steam export and lessen the reformer size. Madesen and Gol (1998) introduced advanced technologies including pre-reforming, auto-thermal reforming, oxygen-fired secondary reforming etc. Liu and Li (1997) analysed the industrial applications of integration of pre-reformers and traditional steam reformers.

4.3. Hydrogen plant process modelling
4.3.1. Previous research
In order to integrate hydrogen plants into hydrogen networks, an adaptive model of the hydrogen plant is essential. Although some research has been done for overall hydrogen plant modelling and optimisation, it pales in comparison to the tremendous efforts made on reforming and shift conversion kinetics, and modelling the corresponding reactors. The details were reviewed notably by Rajesh et al (2000, 2001). Rajesh et al (2000) simulated a steam reformer using a rigorous model with reaction kinetics, heat transfer and diffusion in the catalyst pellet. The non-dominated sorting genetic algorithm (NSGA) is applied to have the simultaneous minimisation of methane feed rate and the maximisation of the flowrate of carbon monoxide in the syngas. The research was then extended to the optimisation of industrial hydrogen plants (Rajesh et al, 2001). The steam reformer and shifter were modelled rigorously while steam was generated by heat exchange with syngas and shift gas, PSA was considered as a splitter with specific 90% of hydrogen recovery and 99.5% product purity. A multi-objective optimisation of simultaneous maximisation of product hydrogen and export steam flowrate was performed through NSGA. Bussani et al (1995) presented an application of the reconciliation and optimisation package ORO (On-line Reconciliation and Optimisation) to an existing hydrogen plant. The hydrogen plant simulation model was built by a software package and tuned by reconciling operation data and simulation data. The optimisation was carried out with the ORO package through a sequential modular approach to achieve new operating conditions for different production targets. 97

Chapter 4

Integration of Hydrogen Generation and Hydrogen Recovery

The discussion of advantages and shortcomings of NSGA is not referred to avoid distraction. The rigorous mathematics models used to describe a hydrogen plant take advantage of the research on steam reforming and shifting reactions, are suitable for the unit-level analysis of hydrogen plants. However, not only do the differential equations representing reaction kinetics require significant computational cost and encumber the employment of equation-oriented optimisation, but also many operation and design data are involved in simulation and optimisation. These data e.g. catalyst volume and length of reactor, are generally not available during conceptual design stage. When the integration of hydrogen plants and hydrogen networks is defined in site level, the optimisation of operating conditions inside a hydrogen plant is not the issue being focused on. The “black-box” model can be obtained by the support of commercial software packages. Another motivation to build a new hydrogen plant model is to remove unreality in the existing models. For example, the steam generation in a hydrogen plant uses heat not only from syngas and shift gas but also from flue gas in the reformer convective section, which was ignored by Rajesh‘s model. Also when PSA is considered to be the purification facility for shift gas and off-gases in a hydrogen plant, PSA recovery will vary due to the various purities of off-gases in refineries.

4.3.2. Process simulation for hydrogen plant
A simplified model for hydrogen plants is developed with the assistance of comprehensive process simulation. The model is built in three steps: firstly, process simulation of a hydrogen plant is accomplished by commercial simulation software. The simulation provides process data in specific operating conditions. The hydrogen production capacity and utility consumption are then calculated referring to unit feedstock. In the second step, a linear model of hydrogen plants is built and tailored by correlating coefficients through process data. The model is then verified by process simulation results in the third step. In the first step, the commercial simulation software PRO/II is used as the simulation tool. PRO/II is a comprehensive computer simulation system developed by Simulation Sciences Inc., for process engineers in the chemical, petroleum, natural gas, and synthetic fuels industries. It combines the data resources of a large chemical 98

Chapter 4

Integration of Hydrogen Generation and Hydrogen Recovery

component library and extensive thermodynamic property prediction methods, with the most advanced and flexible unit operation techniques. PRO/II has been accepted as one of the most popular computational facilities for process engineers to perform mass and energy balance calculations for many steady-state processes. The aim of process simulation is to obtain the hydrogen yields upon certain feedstock and corresponding utility consumption and/or generation such as fuel gas, steam, deionised water (DSW), electricity, cooling water (CW) and so on. Figure 4.2 shows the simulation structure and the mass flow of feedstock, yield and utility of a hydrogen plant except PSA purification. The keyword file of PRO/II input is listed in Appendix. The unit operations included in the process simulation are reactors, heat exchangers, compressors, pumps, flashers, stream mixers and splitters, stream calculators. A programming module is used to define streams by process requirement instead of specific process parameters, and compute utility consumption automatically by the access to a process data library. Some of important aspects in process simulation for hydrogen plants are explained as follows:
Fuel gas PSA tail gas Air Steam drum

Steam export hydrodesulphurised feedstock

HT shifter

Deionised water Shifted gas to PSA

Steam reformer

Cooling water

Electricity

Other utilities

Figure 4.2 Simulation structure for hydrogen plant 99

Chapter 4

Integration of Hydrogen Generation and Hydrogen Recovery

? Thermodynamic method SRK is selected as the thermodynamic property methods because it gives good results for the equilibrium of most hydrocarbon and hydrocarbon-water systems. ? Reactors General unit operation modules provided by PRO/II can cover all kinds of reactors in hydrogen plants. The steam reforming reaction is simulated by a Gibbs reactor module. Using the minimisation of Gibbs free energy, it calculates product rates, compositions and thermal conditions subject to an overall material balance in a single-phase reactor. The steam reformer operates isothermally. High-temperature shifting reaction is simulated adiabatically. Reaction completions in both reactions are controlled by the definition of a temperature approach to the equilibrium temperature. A negative temperature approach is specified for the steam reforming because the overall reaction is end

相关文章:
UNIT 1 INTRODUTION OF PLASTICS
(such as carbon, hydrogen, oxygen, or nitrogen)...The integration of several steps is quite common ...resemble oil refineries in their size and scope....
脱硫与脱硝论文
Integration technology for flue gas dry technique ...The sulfur content of crude oils ranges from 0 ...In petroleum refineries, the hydrogen sulfide gas ...
Forecast of development trends in Taiwan's machinery industry
A transition to integrated biorefineries as analogues of oil refineries has ...integration that challenges the competitive development of RRM, for which the...
The Gilded Age
Rockefeller used horizontal integration to monopolize the oil industry. Rockefeller controlled 40 oil refineries, accounting for about 90 percent of the nation...
专业英语 第八单元 翻译
In oil refineries, propylene mainly is used for alkylation reaction. 芳烃是煤的一种成分,在1980年,他在煤种含量非常少,含有4%的笨,0.9%的甲苯,0.1% 的...
voa
Nowhere is that determination more evident than outside the oil refineries ...President Wulff stressed that integration did not mean giving up Turkish ...
英汉翻译方法1——被动语态译法
译成“把”、“使”和“由”字句 ?When the crude oil is obtained from the field, it is taken to the refineries to be treated. 1The commonest form...
延迟石油焦介绍
Delayed coking, is one of the unit processes, used in many oil refineries...99.5 0.1 Component Fixed carbon, wt % Hydrogen, wt % Nitrogen, wt %...
重油催化裂化工艺的新进展
oil demand to rise very quickly, which requires oil refineries in China ... makes catalytic gasoline in the of olefin main for hydrogen transfer, and...
汽油资源的整合与优化利用(含中英文word,可自行改为pdf)
12,No. 2,pp 51-56 " 丨丨丨 Integration and...Therefore, China ’ s oil refineries have to ... hydrogen purity and light aromatic content had ...
更多相关标签:
refineries | teapot refineries | ysl tint in oil | ysl oil in stick试色 | ysl tint in oil 试色 | ysl oil in stick | roilling in the deep | oil in gel |