Optimal Design and Operation of Energy Polygeneration Systems by Yang Cen Submitted to te Department of Cemical Engineering in partial fulfillment of te requirements for te degree of Doctor of Pilosopy in Cemical Engineering at te MASSACHUSETTS INSTITUTE OF TECHNOLOGY February 2013 c Massacusetts Institute of Tecnology 2013. All rigts reserved. Autor.............................................................. Department of Cemical Engineering October 25, 2012 Certified by.......................................................... Paul I. Barton Lammot du Pont Professor of Cemical Engineering Tesis Supervisor Certified by.......................................................... Tomas A. Adams II Assistant Professor of Cemical Engineering Tesis Supervisor Accepted by......................................................... Patrick S. Doyle Professor of Cemical Engineering Cairman, Committee for Graduate Students
2
Optimal Design and Operation of Energy Polygeneration Systems by Yang Cen Submitted to te Department of Cemical Engineering on October 25, 2012, in partial fulfillment of te requirements for te degree of Doctor of Pilosopy in Cemical Engineering Abstract Polygeneration is a concept were multiple energy products are generated in a single plant by tigtly integrating multiple processes into one system. Compared to conventional single-product systems, polygeneration systems ave many economic advantages, suc as potentially ig profitability and ig viability wen exposed to market fluctuations. Te optimal design of an energy polygeneration system tat converts coal and biomass to electricity, liquid fuels (napta and diesel) and cemical products (metanol) wit carbon dioxide (CO 2 ) capture under different economic scenarios is investigated. In tis system, syngas is produced by gasification of coal and/or biomass; purified by a cleaning process to remove particles, mercury, sulfur and CO 2 ; and ten split to different downstream sections suc as te gas turbine, FT process and te metanol process. In tis tesis, te optimal design wit te igest net present value (NPV) is determined by optimizing equipment capacities, stream flow rates and stream split fractions. Te case study results for static polygeneration systems reveal tat te optimal design of polygeneration systems is strongly influenced by economic conditions suc as feedstock prices, product prices, and potential emissions penalties for CO 2. Over te range of economic scenarios considered, it can be optimal to produce a mixture of electricity, liquid fuels, and metanol; only one eac; or mixtures in-between. Te optimal biomass/coal feed ratio significantly increases wen te carbon tax increases or te biomass price decreases. An economic analysis of te optimal static polygeneration designs yielded a sligtly iger NPV tan comparable single-product plants. Te flexible operation is ten considered for te energy polygeneration system. In real applications, product prices can fluctuate significantly seasonally or even daily. Te profitability of te polygeneration system can potentially be increased if some operational flexibility is introduced, suc as adjusting te product mix in response to canging market prices. Te major callenge of tis flexible design is te determination of te optimal trade-off between flexibility and capital cost because iger flexibility typically implies bot iger product revenues and larger equipment sizes. 3
A two-stage optimization formulation for is used for te optimal design and operation of flexible energy polygeneration systems, wic simultaneously optimizes design decision variables (e.g., equipment sizes) and operational decision variables (e.g., production rate scedules) in several different market scenarios to acieve te best expected economic performance. Case study results for flexible polygeneration systems sow tat for most of market scenarios, flexible polygeneration systems acieved iger expected NPVs tan static polygeneration systems. Furtermore, even iger expected NPVs could be obtained wit increases in flexibility. Te flexible polygeneration optimization problem is a potentially large-scale nonconvex mixed-integer nonlinear program (MINLP) and cannot be solved to global optimality by state-of-te-art global optimization solvers, suc as BARON, witin a reasonable time. Te nonconvex generalized Benders decomposition (NGBD) metod can exploit te special structure of tis matematical programming problem and enable faster solution. In tis metod, te nonconvex MINLP is relaxed into a convex lower bounding problem wic can be furter reformulated into a relaxed master problem according to te principles of projection, dualization and relaxation. Te relaxed master problem yields an nondecreasing sequence of lower bounds for te original problem. And an nonincreasing sequence of upper bounds is obtained by solving primal problems, wic are generated by fixing te integer variables in te original problem. A global optimal objective is obtained wen te lower and upper bounds coincide. Te decomposition algoritm guarantees to find an ɛ-optimal solution in a finite number of iterations. In tis tesis, several enanced decomposition metods wit improved relaxed master problems are developed, including enanced NGBD wit primal dual information (NGBD-D), piecewise convex relaxation (NGBD-PCR) and lift-and-project cuts (NGBD-LAP). In NGBD-D, additional dual information is introduced into te relaxed master problem by solving te relaxed dual of primal problem. Te soobtained primal dual cuts can significantly improve te convergence rate of te algoritm. In NGBD-PCR, te piecewise McCormick relaxation tecnique is integrated into te NGBD algoritm to reduce te gap between te original problem and its convex relaxation. Te domains of variables in bilinear functions can be uniformly partitioned before solution or dynamically partitioned in te algoritm by using te intermediate solution information. In NGBD-LAP, lift-and-project cuts are employed for solving te piecewise lower bounding problem. In all tree enanced decomposition algoritms, tere is a trade-off between tigter relaxations and more solution times for subproblems. Te computational advantages of te enanced decomposition metods are demonstrated via case studies on te flexible polygeneration problems. Te computational results sow tat, wile NGBD can solve problems tat are intractable for a state-ofte-art global optimization solver (BARON), te enanced NGBD algoritms elp to reduce te solution time by up to an order of magnitude compared to NGBD. And enanced NGBD algoritms solved te large-scale nonconvex MINLPs to ɛ-optimality in practical times (e.g., a problem wit 70 binary variables and 44136 continuous variables was solved witin 19 ours). 4
Tesis Supervisor: Paul I. Barton Title: Lammot du Pont Professor of Cemical Engineering Tesis Supervisor: Tomas A. Adams II Title: Assistant Professor of Cemical Engineering 5
6
Acknowledgments I would like to begin my acknowledgments by tanking my tesis advisors Prof. Paul I. Barton and Prof. Tomas A. Adams II. Tey provided me excellent suggestions and detailed advisories for my researc, bot in matematical and engineering aspects. I appreciate tat tey elped me a lot overcome te obstacles, wile gave me abundant freedom to develop my own ideas. I would like to tank my tesis committee members: Prof. Jon G. Brisson, Prof. Amed F. Goniem and Prof. William H. Green. Tey are very entusiastic to my researc project and provided many useful suggestions to improve my modeling metodologies. My sincerely tanks to tem for providing me broader insigt to energy systems design from areas oter tan process systems engineering. I would express my gratitude to BP-MIT Advanced Conversion Project and Martin Fellowsip for Sustainability for funding my PD researc. I would appreciate te BP-MIT Conversion Project team for teir essential collaboration work. Randall Field, Huan Hsu and Robert Brasington, wo are current or former researc staffs at MIT Energy Initiative, constructed an Aspen Plus simulation model for te polygeneration process, wic provided tecnical parameters in my optimization model. Some former student working for different tasks of te conversion project gave me significant elp for te model construction. Dr. Rory Monagan (mecanical engineering) provided detailed parameters for biomass gasification. Dr. Barbara Botros (mecanical engineering) estimated te efficiency of low-temperature steam turbines. Sara Basadi (tecnology and policy) evaluated te economic performance of Selexol process. My gratitude will also expressed to engineers at BP, especially George Huff, Martin Sellers and Bruce Briggs, for teir suggestions on te process flowseet, modeling metod and tecnical and economic parameters. My researc was greatly benefited from te elp of members in Process Systems Engineering Laboratory (PSEL). I closely collaborated wit Prof. Xiang Li for te NGBD algoritm development. Part of tis tesis, e.g., Capter 5 and 6, sows 7
results of te collaboration work. Xiang is also a very good personal friend of mine, and we discussed a lot for many areas outside of current researc. Dr. Kai Hoeffner coordinated te Task 6A team in BP-MIT Conversion Project, and compiled all reports and presentations to BP for Kamil and me. Ajay Selot elped me a lot for algoritm development and C++ programming. Josep Scott provided me great elp for preparation of tesis proposal and committee meetings. Mattew Stuber and Acim Wecsung elped to solve many problems encountered wit computer and program issues. I sincerely tank Viet Pan and Adam Newby at Aspen Systems. Tey elped me solve multiple computing and operating problems of te cluster Banquo, not only benefiting my researc but also facilitating researc of oter PSEL members. I would like to greatly appreciate my family and my friends. My parents gave me a lot of insigt and elp for my life. Tey also took care of me well wen I wrote my tesis. My friends at Boston are a essential part of my life, especially for some of very good friends at MIT wo provided me countless elp. It is impossible for me to express te gratitude to tem in a few sentences, terefore I tank all friends wo elped me in my life. 8
Contents 1 Introduction 23 1.1 Energy Polygeneration Processes.................... 23 1.1.1 Clean Coal Conversion Processes................ 23 1.1.2 Biomass Conversion Processes.................. 28 1.1.3 Energy Polygeneration Processes................ 29 1.1.4 Flexible Energy Polygeneration Processes............ 32 1.1.5 Literature Review......................... 34 1.2 Stocastic/Multiperiod Optimization Problems............ 38 1.2.1 Problem Formulation & Applications.............. 38 1.2.2 Global Optimization Algoritms & Literature Review..... 40 2 Process Description of Energy Polygeneration Systems 45 2.1 Overview.................................. 45 2.2 ASU and Gasifier............................. 46 2.3 Syngas Cleaning and Upgrading Process................ 47 2.4 Fiscer-Tropsc Syntesis Process.................... 48 2.5 Metanol Syntesis Process....................... 49 2.6 Gas Turbine................................ 50 2.7 HRSG and Steam Turbine........................ 50 3 Optimal Design and Operation of Static Energy Polygeneration Systems 53 3.1 Matematical Model........................... 53 9
3.1.1 Overview............................. 53 3.1.2 Mass Balance........................... 55 3.1.3 Energy Balance.......................... 61 3.1.4 Entalpy Calculation....................... 62 3.1.5 Production Rates and Feedstock Consumption Rates..... 63 3.1.6 Capital Costs........................... 63 3.1.7 Economic Analysis........................ 63 3.1.8 Model Summary......................... 65 3.2 Case Study Results............................ 65 3.2.1 Detailed Results of Two Sample Case Studies......... 66 3.2.2 Power Price vs. Napta Price................. 68 3.2.3 Napta Price vs. Metanol Price................ 70 3.2.4 Biomass Price vs. Carbon Tax.................. 72 3.2.5 Carbon Tax witout Fuel vs. Carbon Tax wit Fuel...... 75 3.2.6 Polygeneration System vs. Single-product System....... 76 4 Optimal Design and Operation of Flexible Energy Polygeneration Systems 89 4.1 Matematical Model........................... 89 4.1.1 Overview............................. 89 4.1.2 Capital Costs........................... 91 4.1.3 Economic Analysis........................ 92 4.1.4 Model Summary......................... 94 4.2 Case Study Results............................ 95 4.2.1 Case Study Problems....................... 95 4.2.2 Optimization Results of a Sample Case Study......... 97 4.2.3 Operations in Flexible Polygeneration Systems......... 99 4.2.4 Comparison of Static Designs and Flexible Designs...... 102 5 Nonconvex Generalized Benders Decomposition Algoritm 115 5.1 Motivation................................. 115 10
5.2 Overview.................................. 116 5.3 Subproblems in te Decomposition Metod............... 118 5.3.1 Primal Bounding Problem.................... 118 5.3.2 Feasibility Problem........................ 119 5.3.3 Relaxed Master Problem..................... 119 5.3.4 Primal Problem.......................... 121 5.4 Decomposition Algoritm........................ 122 5.5 Conclusions................................ 123 6 Enanced Nonconvex Generalized Benders Decomposition Algoritms125 6.1 Overview of Enancement Tecnologies................. 125 6.2 Enanced Decomposition Algoritm wit Primal Dual Cuts..... 127 6.2.1 New Subproblems......................... 127 6.2.2 Teoretical Properties...................... 131 6.2.3 Enanced Decomposition Algoritm wit Primal Dual Cuts. 140 6.3 Enanced Decomposition Algoritm wit Piecewise Convex Relaxation 143 6.3.1 Piecewise Relaxation for Bilinear Functions.......... 143 6.3.2 New Subproblems......................... 145 6.3.3 Teoretical Properties...................... 149 6.3.4 Enanced Decomposition Algoritm wit Piecewise Relaxation 151 6.3.5 Adaptive Piecewise Convex Relaxation & New Subproblems. 154 6.3.6 Enanced Decomposition Algoritm wit Adaptive Piecewise Relaxation............................. 158 6.4 Enanced Decomposition Algoritm wit Primal Dual Cuts and Piecewise Convex Relaxation......................... 161 6.4.1 New Subproblems......................... 161 6.4.2 Enanced Decomposition Algoritm wit Primal Dual Cuts and Piecewise Relaxation..................... 162 6.5 Enanced Decomposition Algoritm wit Lift-and-Project Cuts... 166 6.5.1 Lift-and-Project Cuts for MILPs................ 166 11
6.5.2 New Subproblems......................... 170 6.5.3 Teoretical Properties...................... 178 6.5.4 Enanced Decomposition Algoritm wit Lift-and-Project Cuts 179 6.6 Conclusions................................ 183 7 Case Studies of Polygeneration Problems wit Decomposition Algoritms 185 7.1 Model Reformulations.......................... 185 7.1.1 Aggregate Equipment....................... 186 7.1.2 Discrete Capital Costs...................... 187 7.1.3 Oter Reformulations....................... 191 7.1.4 Model Summary......................... 192 7.2 Case Study Problems and Implementation............... 193 7.2.1 Description of Case 1 and 2................... 193 7.2.2 Description of Case 3....................... 195 7.2.3 Implementation.......................... 199 7.3 Optimization Results........................... 200 7.3.1 Optimization for Different Time Periods............ 200 7.3.2 Optimization under Market and Policy Uncertainties..... 202 7.4 Computational Performance....................... 203 7.4.1 NGBD and Enanced NGBD wit Primal Dual Cuts (NGBD-D and NGBD-MD)......................... 203 7.4.2 Enanced NGBD wit Piecewise Convex Relaxation (NGBD- PCR)............................... 204 7.4.3 Enanced NGBD wit Primal Dual Cuts and Piecewise Convex Relaxation (NGBD-D-PCR)................... 205 7.4.4 Enanced NGBD wit Lift-and-Project Cuts (NGBD-LAP). 206 8 Conclusions and Future Work 221 8.1 Conclusions................................ 221 8.2 Future Work................................ 225 12
8.2.1 Polygeneration Model...................... 225 8.2.2 Decomposition Algoritm.................... 228 A Detailed Matematical Model for Static Polygeneration Systems 231 A.1 Matematical Model........................... 231 A.1.1 Mass Balance........................... 231 A.1.2 Energy Balance.......................... 245 A.1.3 Entalpy Calculation....................... 250 A.1.4 Production Rates and Feedstock Consumption Rates..... 250 A.1.5 Capital Costs........................... 252 A.1.6 Economic Analysis........................ 255 A.2 Parameter Tables............................. 257 B Detailed Matematical Model for Flexible Polygeneration Systems275 B.1 Matematical Model........................... 275 B.1.1 Mass Balance........................... 275 B.1.2 Energy Balance.......................... 276 B.1.3 Entalpy Calculation....................... 276 B.1.4 Production Rates and Feedstock Consumption Rates..... 277 B.1.5 Capital Costs........................... 277 B.1.6 Economic Analysis........................ 281 B.2 Parameter Tables............................. 283 Nomenclature 289 Bibliograpy 301 13
14
List of Figures 1-1 Te flowseet of an example IGCC process wit CCS. [182]..... 25 1-2 Te flowseet of an example CTL process. [174]............ 27 1-3 Te flowseet of an example BTL process wit tree tar removal alternatives. [82].............................. 29 2-1 Simplified process flowseet of te polygeneration system........ 46 2-2 Detailed process flowseet of te polygeneration system........ 51 3-1 Product distributions in case studies under different power prices and napta prices. (Te axes are rotated to provide a favorable view.) [Grey circle : Case 1, Wite circle : Case 2.].............. 78 3-2 Net present values in case studies under different power prices and napta prices............................... 79 3-3 Annual CO 2 emission in case studies under different power prices and napta prices............................... 79 3-4 Product distributions in case studies under different napta prices and metanol prices. (Te axes are rotated to provide a favorable view.) 80 3-5 Net present values in case studies under different napta prices and metanol prices............................... 81 3-6 Annual CO 2 emission in case studies under different napta prices and metanol prices............................ 81 3-7 Annual gross CO 2 emission in case studies under different biomass prices and carbon taxes.......................... 82 15
3-8 Annual net CO 2 emission in case studies under different biomass prices and carbon taxes.............................. 82 3-9 Biomass usage in case studies under different biomass prices and carbon taxes.................................... 82 3-10 Net present values in case studies under different biomass prices and carbon taxes................................ 83 3-11 Product distributions in case studies under carbon taxes for process CO 2 emissions. [ : electricity, : liquid fuels, : metanol ]................................ 83 3-12 Product distributions in case studies under carbon taxes for total CO 2 emissions. [ : electricity, : liquid fuels, : metanol ]................................. 84 3-13 Net present values in case studies under two carbon tax cases. [ : carbon tax w/o fuel, : carbon tax w/ fuel ]........ 84 3-14 Annual process CO 2 emissions in case studies under two carbon tax cases. [ : carbon tax w/o fuel, : carbon tax w/ fuel ] 85 3-15 Annual total CO 2 emissions in case studies under two carbon tax cases. [ : carbon tax w/o fuel, : carbon tax w/ fuel ]... 85 3-16 Product distributions in te polygeneration systems wit te optimal designs. [ : electricity, : liquid fuels, : metanol ]................................. 86 3-17 Net present values of te polygeneration systems and different singleproduct systems. [ : polygeneration plant, : power plant w/ CCS, : power plant w/o CCS, : liquid fuels plant, : metanol plant ]............................ 86 3-18 Net present values of te polygeneration systems and several singleproduct systems (enlarged view). [ : polygeneration plant, : liquid fuels plant, : metanol plant ].............. 87 4-1 Scale factors for product prices in different scenarios.......... 96 16
4-2 Product distributions for te 50% operational flexibility case (%). [P = peak, OP = off-peak.]......................... 107 4-3 Product distributions for te 100% operational flexibility case (%). [P = peak, OP = off-peak.]......................... 108 4-4 Equipment capacity usages for te middle carbon tax and 50% operational flexibility case (%). [P = peak, OP = off-peak.]......... 109 4-5 Equipment capacity usages for te middle carbon tax and 100% operational flexibility case (%). [P = peak, OP = off-peak.]........ 109 4-6 CO 2 emission rates for te middle oil price and 50% operational flexibility case (tonne/r). [P = peak, OP = off-peak; Process Only = carbon taxes only apply to CO 2 emissions in te process, Plus Liquid Fuels = carbon taxes apply to bot te CO 2 emissions from te process, and to te carbon in te fuels wic will eventually be combusted.].... 110 4-7 CO 2 emission rates for te middle oil price and 100% operational flexibility case (tonne/r). [P = peak, OP = off-peak; Process Only = carbon taxes only apply to CO 2 emissions in te process, Plus Liquid Fuels = carbon taxes apply to bot te CO 2 emissions from te process, and to te carbon in te fuels wic will eventually be combusted.]110 4-8 Annual product distributions for tree different operational flexibilities (%)..................................... 111 4-9 Annual CO 2 emissions for tree different operational flexibilities (Mt/yr). [Process Only = carbon taxes only apply to CO 2 emissions in te process, Plus Liquid Fuels = carbon taxes apply to bot te CO 2 emissions from te process, and to te carbon in te fuels wic will eventually be combusted.].............................. 111 4-10 Capital investments in all cases ($billion)................ 112 4-11 Annual net profits in all cases ($billion/yr)............... 112 4-12 Net present values in all cases ($billion)................. 113 4-13 Increase of NPV in flexible polygeneration systems compared to te corresponding static polygeneration systems (%)............ 113 17
5-1 Flowcart for te decomposition algoritm................ 124 6-1 Flowcart for te enanced decomposition algoritm wit primal dual cuts..................................... 142 6-2 Flowcart for te enanced decomposition algoritm wit piecewise convex relaxation.............................. 153 6-3 Flowcart for te enanced decomposition algoritm wit adaptive piecewise convex relaxation........................ 161 6-4 Flowcart for te enanced decomposition algoritm wit primal dual cuts and piecewise convex relaxation................... 165 6-5 Flowcart for te enanced decomposition algoritm wit primal dual cuts and adaptive piecewise convex relaxation.............. 166 6-6 Flowcart for te enanced decomposition algoritm wit lift-andproject cuts................................. 182 7-1 Illustration of aggregate equipment.................... 187 7-2 Scale factors of product prices in all scenarios for Case 1........ 195 7-3 Scale factors of product prices in all scenarios for Case 2........ 197 18
List of Tables 3.1 Key decision variables in te model................... 54 3.2 Dry mass compositions of feedstocks.................. 56 3.3 Economic parameters in Case 1 and Case 2............... 66 3.4 Feedstock consumption rates and production rates in Case 1 and Case 2 67 3.5 Optimal results of key decision variables in Case 1 and Case 2... 68 3.6 Optimal product distributions in Case 1 and Case 2.......... 68 3.7 Economic parameters in case studies under different power prices and napta prices............................... 69 3.8 Economic parameters in case studies under different napta prices and metanol prices........................... 71 3.9 Economic parameters in case studies under different biomass prices and carbon taxes................................ 73 3.10 Economic parameters in case studies under different carbon tax policies 75 3.11 Economic parameters in case studies comparing te polygeneration and single-product systems.......................... 77 4.1 Key operational decision variables in te model............ 90 4.2 Fractions of occurrence of all scenarios................. 95 4.3 Te average prices for different oil prices................ 97 4.4 Te values of different carbon taxes ($/tonne of CO 2 )......... 97 4.5 Optimal values of key decision variables in te sample case study.. 104 4.6 Feedstock consumption rates and production rates for te sample case study in all scenarios.......................... 105 19
4.7 Annual feedstock consumption rates and production rates for te sample case study.............................. 106 7.1 Parameters for equipment capacities.................. 188 7.2 Parameters for equipment capital costs................. 190 7.3 Case study problems (Case 1 and 2)................... 193 7.4 Average market prices and carbon tax in Case 1 and 2........ 194 7.5 Fractions of occurrence of all scenarios for Case 1........... 195 7.6 Fractions of occurrence of all scenarios for Case 2........... 196 7.7 Case study problem (Case 3)....................... 197 7.8 Average market prices and carbon tax in Case 3............ 198 7.9 Scale factors of market prices under different oil price scenarios.... 198 7.10 Scale factors of te carbon tax under different carbon tax scenarios. 199 7.11 Optimal equipment designs for Cases 1 and 2............. 201 7.12 Optimal operations in Case 1...................... 207 7.13 Optimal Operations in Case 2...................... 208 7.14 Economics of Cases 1 and 2....................... 209 7.15 Optimal equipment designs for Cases 3................. 209 7.16 Optimal feedstock consumption rates in Case 3............ 210 7.17 Optimal production rates in Case 3 (electricity, napta and diesel). 211 7.18 Optimal production rates in Case 3 (metanol and sulfur)...... 212 7.19 Optimal CO 2 sequestration rates and emission rates in Case 3.... 213 7.20 Economics of Cases 3........................... 214 7.21 Computational performance of BARON, NGBD, NGBD-D and NGBD- MD for Case 1 (70 binary variables and 4904 continuous variables).. 214 7.22 Computational Performance of BARON, NGBD, NGBD-D and NGBD- MD for Case 2 (70 binary variables and 14712 continuous variables). 215 7.23 Computational Performance of BARON, NGBD and NGBD-D for Case 3 (70 binary variables and 44136 continuous variables)........ 215 20
7.24 Computational performance of NGBD and NGBD-PCR for Case 1 (70 binary variables and 4904 continuous variables)............ 216 7.25 Computational performance of NGBD and NGBD-PCR for Case 2 (70 binary variables and 14712 continuous variables)............ 216 7.26 Computational performance of NGBD and NGBD-PCR for Case 3 (70 binary variables and 44136 continuous variables)............ 217 7.27 Computational performance of NGBD and NGBD-D-PCR for Case 1 (70 binary variables and 4904 continuous variables).......... 217 7.28 Computational performance of NGBD and NGBD-D-PCR for Case 2 (70 binary variables and 14712 continuous variables).......... 218 7.29 Computational performance of NGBD and NGBD-D-PCR for Case 3 (70 binary variables and 44136 continuous variables).......... 218 7.30 Computational performance of NGBD and NGBD-LAP for Case 1 (70 binary variables and 4904 continuous variables)............ 219 7.31 Computational performance of NGBD and NGBD-LAP for Case 2 (70 binary variables and 14712 continuous variables)............ 219 A.1 Mole/mass compositions......................... 258 A.2 Mass/molar ratios............................. 259 A.3 Conversions................................ 259 A.4 Efficiency................................. 260 A.5 Selectivity................................. 260 A.6 Split fractions............................... 261 A.7 Temperatures ( C)............................ 262 A.8 Base case flow rates for power consumption/generation (Mmol/r).. 263 A.9 Base case power consumption/generation rates (MW)......... 264 A.10 Heat/power consumption coefficients.................. 264 A.11 Molar weigt (kg/kmol)......................... 265 A.12 Coefficients for entalpy calculations under 5.5 MPa.......... 266 A.13 Coefficients for entalpy calculations under 3.2 MPa.......... 267 21
A.14 Coefficients for entalpy calculations under 2 MPa........... 268 A.15 Coefficients for entalpy calculations under 1.6 MPa.......... 269 A.16 Coefficients for entalpy calculations under 1 MPa........... 269 A.17 Coefficients for entalpy calculations under 0.1 MPa.......... 269 A.18 Base case flow rates for capital costs.................. 270 A.19 Base case capital costs ($MM)...................... 271 A.20 Sizing factors for capital costs...................... 272 A.21 Maximum capacity (tonne/r)...................... 273 A.22 Economic parameters........................... 273 B.1 Base case flow rates for capital costs.................. 284 B.2 Base case capital costs ($MM)...................... 285 B.3 Sizing factors for capital costs...................... 286 B.4 Maximum capacity (tonne/r)...................... 287 B.5 Economic parameters........................... 287 22
Capter 1 Introduction 1.1 Energy Polygeneration Processes 1.1.1 Clean Coal Conversion Processes Energy and te environment are two crucial issues for te world s sustainable development. Te global energy demand is expected to grow by one-tird from 2010 to 2035 due to te increase of population and te economic growt [37]. Fossil fuels, wit advantages of low cost, large scale and ig stability, will still contribute over 80% of total energy supply in te next several decades [108]. At present te global economy is eavily dependent on te supply of crude oil, wic is limited and potentially unstable. Proven oil reserves are projected to be depleted in about 46 years globally and in fewer tan 20 years in most countries [2], based on te current production rates wit no additional oil discoveries. In addition, te geograpical concentration of oil reserves is a great disadvantage for te energy security of oil-importing countries. By contrast, coal is an abundant and relatively ceap fuel, wose price is typically $1-2 per million Btu, compared to $6-12 per million Btu for natural gas and oil [20]. Coal resources are also widely distributed around te world, including some large energy consuming countries suc as United States, Cina and India [2, 20]. Hence coal will be an alternative to crude oil in te new century, used for power generation, syntetic liquid fuels and cemicals. Te Energy 23
Information Administration (EIA) projects tat coal will account for about 20% of primary energy usage in te United States up to te year 2035 [52]. However, a significant problem tat may obstruct wide utillization of coal is air pollution from coal conversion processes. Coal-fired plants generate large amounts of particulates, sulpur oxides and nitrogen oxides. Coal is also te largest contributor to global carbon dioxide (CO 2 ) emissions for energy use (41%) [20]. More concerns about global warming, wic is partially caused by increasing CO 2 levels in te atmospere, ave led to efforts to reduce CO 2 emissions all over te world. CO 2 capture and sequestration tecnologies must be applied to coal conversion processes in te future greenouse gas constrained world [54]. And, coal conversion processes wit iger efficiency sould be utilized to acieve lower CO 2 emissions for te same amount of energy produced. Several coal-based conversion processes wit ig energy efficiency and low CO 2 emissions, suc as Integrated Gasification Combined Cycle (IGCC) and Coal-to- Liquids (CTL) processes wit carbon capture and sequestration (CCS), are being developed at present [108, 182, 173, 174, 8], serving as potential supplements for current oil-based processes. Figure 1-1 sows te flowseet of a typical IGCC process wit CCS [182]. In te IGCC process, coal is converted to syntesis gas (or syngas), wic primarily contains carbon monoxide (CO), ydrogen (H 2 ), CO 2 and water, by gasification. Hig-temperature oxygen-blown entrained-flow gasifiers are selected to acieve ig conversion of coal and low metane content in te syngas. Coal can be slurry-fed (wit water) or dry-fed (wit nitrogen or CO 2 ) depending on te gasification tecnology. An air separation unit (ASU) is installed to produce pure oxygen for gasification. After gasification, syngas is cooled and passes troug te scrubber to remove particulate, ammonia and clorine species. Ten syngas is sent to a water gas sift (WGS) reactor converting CO and water to CO 2 and H 2. Te sulfur species (wic is primarily ydrogen sulfide (H 2 S)) and CO 2 in te syngas are removed in te absorption unit, in wic cemical solvents (amine) or pysical solvents (Selexol, Rectisol, Purisol, etc.) are used. Pysical solvents are currently more economically affordable for large- 24
scale CO 2 capture. In recent years, several advanced separation processes, including adsorption and membrane tecnologies, ave been developed for igly efficient CO 2 removal, wic can be possibly incorporated into te IGCC process in te future. Te clean H 2 -ric syngas wit very low sulfur content and low CO 2 content is sent to te power generation unit (te gas turbine or fuel cell) to produce electricity. Te captured H 2 S is converted to elemental sulfur in te sulfur recovery unit (e.g., te Claus process), and te captured CO 2 is compressed and sent to some geological storage sites. All ig-quality eat generated in te process is recovered in steam cycles and used for additional electricity generation by steam turbines. Note tat it is optional to install te CO 2 capture units (including WGS reactors and te acid gas absorption unit) in te IGCC process, depending on te economics and policy. For example, te IGCC plant witout CCS is suggested to be built first, and wen Cost and Performance Comparison of Fossil Energy Power Plants carbon capture becomes profitable, CO 2 capture units can be ten installed. Exibit 3-32 Case 2 Process Flow Diagram, GEE IGCC wit CO 2 Capture SHIFT REACTORS GAS COOLING MERCURY 11 BFW HEATING 12 13 REMOVAL & KNOCKOUT DUAL STAGE SELEXOL UNIT CO2 STREAMS CO2 COMPRESSION CO2 PRODUCT 16 SHIFT STEAM 10 QUENCH AND SYNGAS SCRUBBER GEE GASIFIER SECTION 6 7 (RADIANT SLAG COAL SLURRY COOLER) 9 8 SOUR WATER STRIPPER WATER RECYCLE TO PROCESS DEMAND 6.3 MWe NOTE: WATER FROM TAIL GAS COOLER MODELED, BUT NOT SHOWN CLEAN GAS 14 SYNGAS H/P REHEAT SYNGAS EXPANDER 17 CO2 PURIFICATION OFF-GAS CLAUS PLANT OXIDANT CLAUS PLANT 18 SULFUR AIR TO ASU 1 VENT GAS 5 ELEVATED PRESSURE ASU 2 3 CLAUS PLANT OXIDANT NITROGEN DILUENT 4 2X ADVANCED F CLASS GAS TURBINE GAS TURBINE COMBUSTOR 15 SYNGAS FLUE GAS NOTE: TAIL GAS BOOST COMPRESSOR MODELED, BUT NOT SHOWN HYDROGENATION REACTOR AND GAS COOLER 19 21 HRSG 22 TAIL GAS RECYCLE TO SELEXOL STACK GAS 20 TURBINE COOLING AIR 464.0 MWe STEAM TURBINE 274.7 MWe AMBIENT AIR Figure 1-1: Te flowseet of an example 119 IGCC process wit CCS. [182] Te IGCC power plant acieves ig energy conversion efficiency (up to 45% 25
(HHV) for te plant witout CCS), wic is muc iger tan most of currently operated pulverized coal (PC) fired power plants (33-37%, HHV) [108]. Te environmental benefit of te IGCC process is also significant. In te IGCC process, most of te air pollutants, including particulates, mercury, sulfur and nitrogen species, can be removed before combustion at relatively ig concentrations. On te oter and, tese pollutants ave to be removed from muc more diluted flue gas in PC plants. Terefore, pollutant control will be a easier task for IGCC plants compared to conventional PC plants [108]. A similar situation is encountered for te control of CO 2 emissions. Te implementation of pre-combustion carbon capture in IGCC plants will correspond to 5-8% of energy efficiency loss, wile post-combustion carbon capture in PC plants will cause about 12% of energy efficiency loss [182]. Te CTL process contains some similar unit operations as tose in te IGCC process, including gasification, air separation, WGS reaction, acid gas removal, sulfur recovery and electricity generation. Figure 1-2 sows te flowseet of a typical CTL process [174]. Coal is first converted to raw syngas by gasification, and syngas is ten cleaned and upgraded by scrubber, WGS reactors and acid gas absorption units (e.g., Selexol or Rectisol units). Te clean syngas tat is ideal for liquid fuels and cemicals production sould possess a H 2 /CO mole ratio equal to 2, be free of sulfur species, and ave low concentrations of all oter species, especially CO 2 and water. In order to protect te catalyst for liquids production, very low sulfur content in te syngas is required. Te clean syngas can be syntesized to liquid products by two different patways: te metanol process and te Fiscer-Tropsc (FT) process. In te metanol process, clean syngas is converted to metanol by te metanol syntesis reaction, followed by a separation unit removing unreacted syngas, water and iger alcools from te metanol product. Unreacted syngas is recycled back to te reactor or sent to te gas turbine to produce electricity. Te metanol product can be directly sold to te market or furter upgraded to oter products, suc as dimetyl eter (DME), gasoline (by te MTG process) or olefines (by te MTO process). In te FT process, clean syngas is converted to ydrocarbons wit a wide range of carbon numbers by te FT syntesis reaction. Te composition of te FT product is igly 26
dependent on te catalyst and operating conditions (e.g., temperature and pressure). A complicated separation system is required to obtain qualified products. Typically, five different streams are produced from te separation system: ligt ends (including unreacted syngas and ydrocarbons wit small carbon numbers), napta, diesel, wax and water. Napta and diesel can be directly sold or furter upgraded. Wax is usually converted to napta and diesel by catalytic cracking or ydrocracking. Ligt ends are sent to te gas turbine to produce electricity, or converted back to syngas Figure 3-2 Concept 1 - Process Block Flow Diagram GE Gasifier-Based FT Liquid Production Plant by steam reforming or partial oxidization for te FT reaction. Final Report Figure 1-2: Te flowseet of an example CTL process. [174] 32 Te liquid fuels produced by te CTL process are considered as alternatives to current petroleum-derived fuels, especially for tose oil-importing countries. Te FT products ave nearly no sulfur and very low content of aromatics, wic can be sold as ig-quality fuels or blended wit ig-sulfur fuels. Te economic performance of te CTL process as been studied, and it is indicated tat te CTL plant will be profitable if te crude oil price stays above $37 per barrel [173]. Te CTL process wit CCS will not generate more CO 2 emissions tan te oil refinery process. It is 27
estimated tat CTL-derived diesel will result in 5-12% less life cycle CO 2 emissions tan te average petroleum-derived diesel [167]. At present, several small-scale CTL demonstration plants are being built worldwide. 1.1.2 Biomass Conversion Processes Biomass is a promising energy source wit its abundant reserves and renewable supplies, low air pollution and very low lifecycle CO 2 emissions [81]. Te EIA predicts tat biomass will lead to te growt of renewable electricity generation and biofuel will lead to te growt of te liquid fuel supply in te next 25 years [52]. Biomass-derived transportation fuels are produced via several approaces, including fermentation, gasification and pyrolysis. At present, fermentation is te only commercialized biofuel production tecnology. Compared to te oter two ways, fermentation as advantages of lower capital cost and flexible operation. However, feedstocks for te fermentation approac are quite limited so far, e.g., only grains and sugar can be used as feedstock based on te current tecnology. Development of grain-based biofuel will eventually treaten global food supply. In contrast to fermentation, te gasification approac is able to utilize a wide range of non-grain biomass feedstocks, including wood, grass and crop residues. Combined wit te FT process, te gasification approac produces liquid fuels (napta and diesel) tat can be directly used by veicles and is compatible wit te current infrastructure. Biomassto-liquids (BTL) processes (via gasification) are terefore of increasing interest to te energy industry. Te BTL process as a similar structure to te CTL process, and bot of tem include te gasifier, scrubber, sulfur removal unit, FT system, gas turbine and steam turbine. Since biomass is a carbon neutral feedstock, CO 2 sequestration is not needed for te BTL process. Te air-blown circulating fluidized bed (CFB) gasifier, wic is operated at relatively low temperature and low pressure, is usually selected for biomass gasification. CFB gasifiers suffer from incomplete conversion of feedstock and formation of certain amounts of ydrocarbons. Larger ydrocarbons generated in te CFB gasifier, including BTX (benzene, toluene and xylene) and tars, must be 28
removed before te FT process. Several metods are available for tar removal, e.g., termal cracking, catalytic cracking and scrubbing. A typical BTL process flowseet wit tree tar removal alternatives is sown in Figure 1-3 [82]. In order to acieve iger conversions, oxygen-blown entrained-flow gasifiers are also considered for te BTL process, wit some additional feedstock pre-treatment steps suc as drying and torrefaction before gasification. C.N. Hamelinck et al. / Energy 29 (2004) 1743 1771 1747 Fig. 2. Tree gas cleaning trains applied in tis study. Top: tar cracking and conventional wet gas cleaning; middle: Figure tar scrubbing 1-3: and Te conventional flowseet wet gas ofcleaning; example and bottom: BTL Tar cracking process andwit dry gas tree cleaning. tar removal alternatives. [82] temperature by advanced scrubbing wit an oil based medium [12]. Te tar is subsequently stripped from te oil and reburned in te gasifier. At atmosperic pressures, BTX are only partially removed, from about 6 bar BTX are fully removed. Te gas enters te scrubber at about 400 v C, wic allows ig temperature eat excange before te scrubber. 1.1.3 Wen te Energy tars and BTX Polygeneration are removed, te oter Processes impurities are removed by standard wet gas cleaning tecnologies or advanced dry gas cleaning tecnologies. Maximal acceptable contaminantaforementioned concentrations forcoal te and cobalt biomass FT catalyst conversion are summarised processes in Table ave 1, advantages togeter wit of te ig Te effectiveness oftwo gas-cleaning metods. energy In dry efficiency, gas cleaning, lowresidual toxic pollutants contaminations andare lowremoved CO 2 emissions. by cemical However, absorbentseveral at elevated problems operates are encountered at 200 250 v C, before especially tese wen processes precedingbecome additional applicable. compression One is required major drawback (efficient temperature. In te FT situation, ot gas cleaning as little energy advantages as te FT reactor compression requires a cold inlet gas). However, dry gas cleaning may ave lower operational ofcosts IGCC, tan CTL wet gasand cleaning BTL[16]. is te ig capital cost per unit of product. For example, Early compression reduces te size ofgas cleaning equipment, owever, sulpur and cloride te capital cost for an IGCC plant wit CCS can be as ig as $2390/kW based on compounds condense wen compressed and tey may corrode te compressor. Terefore, inter- estimation compression in 2007 to 6 [182]. bar takes Wit placete afterincrease bulk removal of construction and 60 bar compression material just prices before and anmediate te guard bed. more Teunderstanding syngas can contain of tecnical a considerable difficulties, amount ofmetane te estimated and oter capital ligtcost ydrocarbons, for IGCCrep- resenting a significant part ofte eating value ofte gas. Reforming to convert tese com- and CTL pounds will tobe COeven and Higer 2 is optional, in tedriven future. by steam Anoter addition drawback over a nickel of tese catalyst. processes Autotermal is tat reforming is applied in te present study. Compared wit steam reforming, it is of simpler 29
a fixed production rate must be maintained due to rigorous operational requirements for te gasifier. Tese single-product processes cannot easily adapt to te fluctuation of product market prices, especially liquid fuels prices, and teir profitability cannot be guaranteed under all economic conditions. Te availability of feedstocks for te BTL process migt also be a problem. Biomass is usually arvested in certain seasons, wile te gasifier requires continuous operation during te wole year. Te BTL plant will suffer from eiter sortages of feedstock during some times or ig feedstock storage costs. Hig capital costs, uncertainties in te product market and te feedstock supply result in ig investment risks for potential application of tese single-product processes. Energy polygeneration could be a plausible way to address te above issues. Polygeneration, or cogeneration, is a concept in wic multiple products are generated in a single plant from multiple feedstocks by tigtly integrating multiple processes into one system. Polygeneration is attractive for te above advanced energy conversion processes. Note tat IGCC, CTL and BTL processes sare some common unit operations, including gasification, scrubbing, acid gas removal and power generation. It is possible to design an energy polygeneration process by integrating IGCC, CTL and BTL processes togeter, wic uses coal and biomass as feedstocks and co-produces electricity, liquid fuels, cemicals, ydrogen and eat in one plant. Compared to single-product energy processes, energy polygeneration processes ave many economical and environmental advantages. Wit polygeneration, te capital cost and production cost per unit of product will be possibly reduced since some equipment included in te IGCC, CTL and BTL can be sared in one process [183, 118, 119]. For example, in a polygeneration plant co-producing DME and electricity, te production cost of DME will be $6-6.5/GJ, wic is comparable wit conventional fuel prices [118, 51]. Moreover, in a polygeneration process, economic risks can be reduced by diversification of product portfolios, and potentially iger profits can be acieved compared to te single-product plants by optimization of te portfolios. Higer energy efficiency may also be attained in polygeneration processes due to te tigt eat integration of te system [118], e.g., eat generated in exoter- 30
mic reactors in te FT or metanol syntesis process can be recovered by steam generation systems for additional power production. Polygeneration is a promising process tat facilitates te usage of biomass. In polygeneration processes, biomass and coal can be co-gasified in ig-temperature entrained-flow gasifiers wit ig conversions, and te biomass pre-treatment becomes unnecessary. A stable supply of biomass is not required in polygeneration because coal can be used as feedstock for gasifiers wen biomass is unavailable. Liquid fuels produced by co-gasification of biomass and coal wit CCS will lead to muc lower life cycle CO 2 emissions tan petroleum-based fuels. Biomass and coal polygeneration processes wit CCS will even ave negative process CO 2 emissions, wic can be sold as carbon credits or compensate for CO 2 emissions from oter processes. Design and operation of energy polygeneration processes is a callenging task, in wic knowledge and information in different disciplines suc as cemical engineering, mecanical engineering, termal engineering, biocemical engineering and electrical engineering are needed. Because of te ig system complexity, engineering experience and experimental metods, wic are frequently used for traditional process design, are not enoug for te design of polygeneration processes. Hence, advanced simulation and optimization tecnologies need to be developed and applied to te optimal design and operation of energy polygeneration systems. Matematical programming is an effective metod for tis purpose. By formulating design and operational problems as typical optimization problems, suc as nonlinear programming (NLP) problems or mixed-integer nonlinear programming (MINLP) problems, te mass and energy integration of te wole process is systematically studied, and all design and operational variables are optimized to acieve best economic performance or lowest pollutant emissions. Global optimization algoritms can be applied in order to ensure global optimal solutions for tese problems. In tis tesis, a polygeneration system co-producing electricity, liquid fuels (napta and diesel) and cemicals (metanol) from coal and biomass as feedstock is investigated. Te detailed process is described in Capter 2. Te optimal design and operation of (static) polygeneration systems under different economic and policy sce- 31
narios is studied. Optimal product portfolios are obtained under different product price scenarios. Te influence of different carbon tax policies on te optimal production strategy, suc as te implementation of CCS or biomass usage, is also explored. Te case study results are presented in Capter 3. 1.1.4 Flexible Energy Polygeneration Processes Conventional energy and industrial processes attempt to maintain operations at teir maximum capacities during te wole operational period, wic are called static processes. Static processes are relatively easy to operate and control, and most of equipment are most efficient wen operated at teir design capacity. However, static processes may not be economically optimal. In reality, market prices and demands fluctuate frequently. For example, prices of liquid fuels (i.e., gasoline and diesel) vary seasonally; power prices fluctuate during te course of te day due to te difficulty of storage, and bot are affected by unpredictable uman beavior. Static plants may suffer from ig inventory levels or lack of stock under some unpredictable market conditions, resulting in significant profit loss. More significant problems are encountered for static power plants, suc as coal-fired power plants, nuclear power plants and even IGCC power plants (due to te inflexible operation of te gasifier). Power prices and demands at peak times can be several times iger tan tose at off-peak times, and power demands can be extremely ig under some bad weater conditions suc as ig temperatures. Hig dependence on static power plants will lead to serious power sortages at some peak times and significant energy wastage in off-peak times. Te concept of a flexible polygeneration process, wic allows variable product mixes during te project lifetime according to market prices and demands, is terefore proposed. A flexible polygeneration plant alters te production rates of individual products in response to canging market conditions by oversizing equipment. In oter words, te flexible plant focuses on power generation during peak times wen te power price and demand is ig, and is switced to liquids production during offpeak times wen te power price and demand drop significantly. Liquids can be stored 32