Multi-objective optimization for the design and operation of energy efficient chemical processes and power generation

Multi-objective optimization for the design and operation of energy efficient chemical processes and power generation

Available online at www.sciencedirect.com ScienceDirect Multi-objective optimization for the design and operation of energy efficient chemical proces...

630KB Sizes 0 Downloads 55 Views

Available online at www.sciencedirect.com

ScienceDirect Multi-objective optimization for the design and operation of energy efficient chemical processes and power generation GP Rangaiah, Shivom Sharma and Bhargava Krishna Sreepathi Process optimization for two or more objectives is increasing. Since objectives are often conflicting, multi-objective optimization (MOO) provides many Pareto-optimal solutions, quantitative trade-off between objectives, optimal values of decision variables and their trends. These results give greater insight about the process and are useful for selecting one of the optimal solutions. In this paper, MOO is introduced and compared with single objective optimization, and then MOO studies on design of energy efficient processes, published from January 2013 to February 2015, are reviewed. MOO applications in 65 papers of interest to chemical engineers, are classified into four groups: energy efficient processes, biofuels, power generation/CO2 capture, and fuel cell/hydrogen production. Main features of these studies are summarized in four tables. Finally, concluding remarks are given for future studies on MOO applications. Address Department of Chemical & Biomolecular Engineering, National University of Singapore, Singapore 117585, Singapore Corresponding author: Rangaiah, GP ([email protected])

Current Opinion in Chemical Engineering 2015, 10:49–62 This review comes from a themed issue on Process systems engineering Edited by Mahmoud El-Halwagi and Ka Ming Ng

http://dx.doi.org/10.1016/j.coche.2015.08.006 2211-3398/# 2015 Elsevier Ltd. All rights reserved.

Introduction Process industries require raw materials, energy and water for the production of a variety of products such as liquid/ gaseous fuels, polymers/plastics, fertilizers, food & drinks, pharmaceuticals, metals, paper and industrial/fine chemicals. For producing these valuable products for the society’s benefit, a sustainable process efficiently fulfills the demands of the present generation without compromising the needs of future generations. Production of fuels and power for industry, transportation and domestic use has some adverse environmental impact, and so clean energy production is one of the biggest challenges to the engineering community, particularly www.sciencedirect.com

to chemical engineers. Hence, both reduction in energy required by process industries and production of clean energy/fuel with lower carbon footprint are important for sustainability. Energy integration, use of process operations/routes requiring lower energy and optimal process design can reduce the energy required by the industry. They lead to energy efficient chemical processes. On the other hand, energy production with higher conversion efficiency, production of renewable energy and power generation with CO2 reduction and/or capture are useful in reducing the environmental impact of power/fuel production for transportation, industrial and household purposes. These processes too should be optimized for their energy efficiency. In general, there will be several performance criteria (e.g., capital cost, profit, selectivity, controllability and safety) for a process besides the energy-related criterion (e.g., thermal efficiency and utility cost). Considering all these criteria and knowledge on the quantitative tradeoff among them are useful for analyzing the process performance and for convincing managers, which is essential for implementing any project. Optimization for a single objective (i.e., performance criterion) has been used for many decades in a variety of fields such as engineering, science and business. Numerical techniques for single objective optimization (SOO) have been developed and applied to improve the process performance [1–3]. Some of these are readily available in the Solver tool of MS Excel and in process simulators such as Aspen Plus. Optimization techniques for dealing with two or more objectives have been studied and developed, particularly in the last two decades [4]. Such multi-objective optimization (MOO) techniques have found numerous applications in chemical engineering and related areas [5,6,7]. They are now finding applications in process retrofitting and revamping as well [8]. This paper provides an overview of MOO and its applications to the design and operation of energy efficient processes and power generation in the period: January 2013 and February 2015. The next section introduces MOO problems, their solutions, comparison with SOO and techniques for solving MOO problems. The third section outlines journal papers describing MOO applications to the design and operation of energy efficient chemical processes, biofuels, power generation, carbon dioxide capture, fuel cell and hydrogen production. We searched Current Opinion in Chemical Engineering 2015, 10:49–62

50 Process systems engineering

the Scopus database and then screened to choose the journal papers published from January 2013 to February 2015, which are of interest to chemical engineers, particularly to researchers in process systems engineering. Conference presentations are not considered to keep this review concise. Journal papers in each of the areas are then briefly summarized in third section itself. Finally, concluding remarks are given in the final section.

Multi-objective optimization (MOO) An SOO problem for a process application has only one objective function, as follows: Min: or max: F 1 ðxÞ (1) Subject to

xL ¼ x ¼ x U

(2)

gðxÞ  0

(3)

hðxÞ ¼ 0

(4)

In this, x is the vector of decision variables, and xL and xU are respectively vectors of lower and upper bounds on the decision variables. g and h are respectively the sets of inequality and equality constraints, which arise from the governing equations and design limitations. Both deterministic and stochastic techniques have been developed for solving SOO problems. Usually, only one optimal solution is obtained from solving a SOO problem; it can be either a local or a global solution depending on problem characteristics and the optimization technique employed. Some SOO problems may have two or more global solutions, having same objective function value but different values of decision variables. An MOO problem has M number of objectives: F1(x), F2(x), . . ., FM(x), unlike only one objective in an SOO problem; all these objectives are assumed to be of minimization type in the rest of this paper. Decision variables and equality/inequality constraints in the MOO problem are same as those in the SOO problem (Equations 2–4). MOO techniques are used to solve optimization problems having two or more objective functions, which are often conflicting. MOO problems can be solved using either classical methods or stochastic methods. Solution of a MOO problem involving conflicting objectives gives many optimal (trade-off) solutions, known as non-dominated solutions and also called as Pareto-optimal front (after the Italian economist Vilfred Pareto). The set: xP, F1(xP), F2(xP), . . ., FM(xP) is said to be a Pareto-optimal or non-dominated solution if and only if no other feasible x exists such that F1(x)  F2(xP), F2(x)  F2(xP), . . ., FM(x)  FM(xP) with strict inequality valid for at least one objective function. Among the nondominated solutions, one of them is better in at least one objective and also worse in at least one other objective compared to the remaining non-dominated solutions. Current Opinion in Chemical Engineering 2015, 10:49–62

Some MOO problems may have many Pareto-optimal fronts, one of which is the global Pareto-optimal front; in such cases, one of these Pareto-optimal fronts will be obtained depending on the MOO technique employed. Figure 1 shows non-dominated solutions and optimum values of decision variables for a MOO problem having two objective functions and two decision variables. Plot (a) is the objective function space showing the trade-off between the two objectives, whereas plot (b) is the decision variable space showing optimum values of the two decision variables. It is desirable to set lower/upper limits of axes in plot (b) same as the lower/upper bounds on the respective decision variable, instead of zooming into a narrow range; one can then easily observe the extent and practical significance of variation in optimum decision variables. Plot (b) is not helpful in relating the variation of decision variables with the objective functions. Further, there will be usually more than two decision variables. Hence, plots such as (c) and (d) are often presented for discussing the variation of objective functions with the decision variables. It is sufficient to prepare these two plots for one of the objective functions. There will be many plots such as (c) and (d) if there are more decision variables. Graphical representation of more than two objective functions requires three-dimensional plots, spider plots etc. [9,10], which are understandably more difficult to comprehend, and particularly in black and white figures in papers. As mentioned above, solution of a SOO problem usually gives only one optimum solution, whereas solution of a MOO problem gives many alternate optimal/trade-off solutions. So, in case of MOO, selection of one optimal solution from the set of alternate solutions is important and required for implementation in the plant. This can be done based on the experience of senior engineers/managers (i.e., decision makers) or using a Pareto ranking approach, which often requires certain preferences about the objective functions and their changes. Table 1 summarizes main differences and similarities between SOO and MOO approaches. MOO techniques

According to the classification in [4], MOO techniques can be grouped into two main categories: generating methods and preference based methods. Generating methods find Pareto-optimal solutions without any inputs from the decision maker (DM) whereas preference based methods utilize inputs from the DM in obtaining solutions of the MOO problem. Generating methods can be further grouped into no-preference methods and a posteriori methods. No-preference methods do not prioritize objectives, and provide only one Pareto-optimal solution; however, a few optimal solutions can be obtained by using different no-preference methods. A posteriori methods can be further divided into www.sciencedirect.com

Multi-objective optimization for process design Rangaiah, Sharma and Sreepathi 51

Figure 1

(a)

(b) Area of Membrane 2 (m2)

Max. Methane Purity

0.998 0.996 0.994 0.992

10000 8000 6000 4000 2000 0

0.99 0.8 (c)

0.85 0.9 0.95 Max. Methane Recovery

2000

4000

6000

8000 10000

Area of Membrane 1 (m2) (d) Max. Methane Purity

0.998 Max. Methane Purity

0

1

0.996 0.994 0.992

0.998 0.996 0.994 0.992 0.99

0.99 0

2000

4000

6000

Area of Membrane 1

8000 10000 (m2)

0

2500

5000

7500

Area of Membrane 2

10000

(m2)

Current Opinion in Chemical Engineering

Non-dominated solutions for optimizing CO2 removal from natural gas using a two-stage membrane separation process: (a) objective space showing the trade-off between two objectives, (b) decision variable space showing optimal values of decision variables, and (c) and (d) variation of one objective function (methane purity) with decision variables (membrane areas).

Table 1 Main differences and similarities between SOO and MOO approaches Feature Number of objectives Decision variables Constraints Type of optimal solutions Multi-dimensional spaces Techniques for finding solutions Development of techniques Number of solutions obtained Computational time for finding solutions

Solution selection for implementation

www.sciencedirect.com

Single objective optimization (SOO)

Multi-objective optimization (MOO)

Only one objective function. Two or more objective functions. Continuous and/or integer decision variables with bounds on them. Equality and/or inequality constraints; some problems may have no constraints. Local or global solution. Local or global Pareto-optimal front. Only one space for decision variables. Two spaces, one for objective functions and another for decision variables. Deterministic and stochastic optimization techniques. Developed for solving SOO problems. SOO techniques are adapted/extended for solving MOO problems. Usually, only one solution. Usually, many non-dominated solutions (Pareto-optimal front). Deterministic techniques are generally faster Deterministic techniques are faster if only compared to stochastic techniques but the a few non-dominated solutions are required, latter are more likely to find the global solution. whereas stochastic techniques are faster for finding many non-dominated solutions. Straightforward since only one or a few Additional technique/preferences are required optimal solutions are found. to select one of the many non-dominated solutions obtained.

Current Opinion in Chemical Engineering 2015, 10:49–62

52 Process systems engineering

two categories: those using scalarization approach and the others using multi-objective approach. Scalarization approach is intuitive and simple; it transforms an MOO problem into an SOO problem followed by solution using a suitable SOO technique to obtain one optimal (i.e., non-dominated) solution. The transformed SOO problem has to be solved many times, each time with different parameter values, to obtain other nondominated solutions. Weighted sum, e-constraint and normalized normal constraint (NNC) techniques are based on the scalarization approach. The resulting SOO problems can be solved using deterministic techniques such as generalized reduced gradient, successive quadratic programming, branch and bound, and branchand-reduce optimization navigator (BARON). They can also be solved using stochastic techniques like simulated annealing (SA), genetic algorithm (GA), differential evolution (DE) and particle swarm optimization (PSO). In the weighted sum technique, a MOO problem is converted into a SOO problem by constructing a weighted sum of all the objectives. The transformed objective function by this technique is as follows: FðxÞ ¼

M X wi Fi ðxÞ

(5)

i¼1

Here, wi is the weight of objective function ‘i’, given by the user. The difficulty here is in giving the weight to each of the objectives since weights do not necessarily

correspond directly to the relative importance of the objectives or allow trade-offs among the objectives to be expressed. In the e-constraint technique, one of the objectives is chosen as the primary objective function to be optimized while expressing the other objectives as additional inequality constraints. That is, minimize Fp(x) subject to Fi(x)  ei for i = 1, 2, . . ., M and i 6¼ p, where ‘p’ is the primary objective function. Solution of the transformed SOO problem by either weighted sum or e-constraint technique gives only one non-dominated solution, and so the SOO problem with different weights or e values has to be solved repeatedly for finding more non-dominated solutions. A posteriori techniques using multi-objective approach are mostly population-based stochastic optimization techniques. In each iteration, they rank the solutions based on objective values, to obtain many Pareto-optimal solutions at the end of the given stopping criterion (usually, maximum number of iterations), all in one run of the optimizer/program. These include NSGA-II (elitist nondominated sorting GA), SPEA2 (strength Pareto evolutionary algorithm-2), MOEA (multi-objective evolutionary algorithm) and MOPSO (multi-objective PSO). Simplified flow chart of NSGA-II is presented Figure 2. See [9,11] for details on different stochastic MOO techniques. Preference-based techniques can be subdivided into two groups: a priori and interactive type. In the former techniques, the DM is involved in the initial problem formulation of the SOO problem; some examples of this type

Figure 2

Initialize population (size N)

Start

Evaluate Objective Functions and Constraints

Rank Initial Population

Selection for Reproduction No

Yes Stop

Stopping Criteria Met?

Generation of Child Population

Crossover and Mutation

Evaluate Objective Functions and Constraints

Select N Individuals using Nondominated Sorting and Crowding Distance Calculations (if required)

Combine Parent and Child Populations (size 2N) Current Opinion in Chemical Engineering

A flowchart of NSGA-II for MOO. Current Opinion in Chemical Engineering 2015, 10:49–62

www.sciencedirect.com

Multi-objective optimization for process design Rangaiah, Sharma and Sreepathi 53

are lexicographic ordering, value function methods and goal programming. In the interactive techniques, the DM is involved not only in the initial formulation but also in every iteration of the optimization process; s/he reviews the trade-off solutions obtained in each iteration and provides suggestions for the next iteration. NIMBUS [12] is one example of interactive techniques. One optimal solution from the Pareto-optimal front has to be selected for industrial implementation. Some methods for this are the technique for order of preference by similarity to ideal solution (TOPSIS), linear programming technique for multidimensional analysis of preference (LINMAP), fuzzy and pseudo weight approaches. In the applications reviewed in the next section, TOPSIS is often used to choose one solution from the Paretooptimal front. It is based on the premise that the chosen solution should have the shortest Euclidean distance from the ideal solution and the longest Euclidean distance from the negative-ideal solution, both in the objective function space. The ideal solution is composed of best values of all objective functions whereas the negativeideal solution consists of worst values of all objectives. To account for different magnitude of objectives, each objective value can be normalized using Euclidean norm of all its values; weights to account for relative importance of objective functions can also be included. An assumption of TOPSIS is that the objective functions are monotonically increasing or decreasing. See Section 2.3.5 in [13] for details on TOPSIS including a numerical example.

MOO for energy efficient process design and operation Recently, many researchers have used MOO techniques for analysis of processes to produce clean fuel or energy with lower environmental impact, and to improve process design and operation for energy efficiency. There are several databases of peer-reviewed literature such as Scopus and Web of Science. We searched Scopus database with the keyword ‘multi-objective optimization’ in the article title or abstract; limiting the search based on three criteria: Firstly, language — English; secondly, document type — journal article, review article or book chapter; and finally, category — energy. This search gave 585 articles published between January 2013 and February 2015. These articles were published in 40 journals in different engineering disciplines; of these; 20 journals related to chemical engineering are selected. The number of articles published in these 20 journals is 373 out of 585. We browsed through the short-listed 373 articles, and chose 65 articles of interest to chemical engineers, for summarizing in this section. It is hoped this search and screening captured most, if not all, of the papers of interest to readers of this journal. The selected articles are divided into four areas, and Table 2 provides the number of papers in each of these four areas. In the period www.sciencedirect.com

Table 2 Number of papers based on area, journal and country, published between January 2013 and February 2015 No. of papers Area Energy efficient processes Biofuels production Power generation & CO2 capture Fuel cell & hydrogen production

27 9 17 12

Journal Applied Energy Applied Thermal Engineering Biomass and Bioenergy Chemical Engineering and Processing: PI Energy Energy Conversion and Management Fuel International Journal of Energy Research International Journal of Greenhouse Gas Control International Journal of Hydrogen Energy International Journal of Sustainable Energy Journal of Energy Engineering Journal of Power Sources Journal of Renewable and Sustainable Energy Renewable Energy Solar Energy Thermal Engineering Thermal Science

11 10 1 3 13 9 1 2 1 2 1 1 4 1 2 1 1 1

Country Australia Brazil Canada China France Germany India Iran Iraq Italy Malaysia Mexico Qatar South Korea Romania Singapore Slovenia South Africa Spain Sri Lanka Sweden Taiwan Turkey USA

2 1 4 12 3 2 1 10 1 3 4 3 1 1 1 3 1 1 1 2 1 1 1 5

considered (January 2013 and February 2015), there are 27 and nine papers on MOO applications in energy efficient processes and biofuels, respectively. This is understandable considering the relative scope of these two areas. The chosen 65 papers were published in 18 different journals; of these, Applied Energy, Applied Thermal Engineering, Energy and Energy Conversion and Management journals have published about 10 articles each (Table 2). These papers came from researchers Current Opinion in Chemical Engineering 2015, 10:49–62

54 Process systems engineering

Table 3 MOO for the design and operation of energy efficient chemical processes Application

Objectives

Techniques

Air cooled heat exchanger (ACHE) design [14]

Total annual cost and approach temperature

NSGA-II

Fin-and-tube heat exchanger [15]

Heat transfer rate and total pressure drop

Multi-objective GA

Heat pipe heat exchanger [16]

Total cost and effectiveness

NSGA-II

Heat exchanger tube arrangement problem [17]

Heat exchange rate and pressure loss

NSGA-II

Absorption refrigeration system integrated with a heat exchanger network (two examples) [18]

Total annual cost and greenhouse gas emissions

e-Constraint with DICOPT/ CONOPT/CPLEX

Serpentine-channel heat sink [19]

Total thermal resistance and pressure drop

Plate-fin heat sink [20]

Entropy generation rate and material cost

Multi-objective artificial swarm fish algorithm with variable population size and non-dominated sorting (MOAFNS) Teaching-learning-based optimization (TLBO)

Heat exchanger network retrofitting [21]

Investment cost and utility cost

Heat exchanger network (HEN) retrofitting [22]

Investment cost and utility cost

Fuel gas (energy) network in an integrated iron and steel plant [23] Off-gas management in an integrated steel plant [24]

Total cost and CO2 emission

Profit and CO2 emissions

e-Constraint with linear programming and SPEA2

Batch water networks (two cases) [25]

Total annual cost and total storage

e-Constraint with DICOPT/ CONOPT/CPLEX

Reverse osmosis plant [26]

Specific water production cost and specific environmental impact Energy consumption and quality of effluent Heat duty and number of stages

e-Constraint with CONOPT

Activated sludge process [27] Hybrid distillation/melt crystallization process for separation of ortho, meta and para xylene (four hybrid configurations) [28] Hybrid distillation/melt crystallization process for separation of long-chain isomeric aldehydes [29]

Total energy demand and packing height

Current Opinion in Chemical Engineering 2015, 10:49–62

NSGA-II for MOO and integrated differential evolution for SOO NSGA-II for MOO and integrated differential evolution for SOO SPEA2

MOPSO NSGA-II

Evolutionary MO algorithm (SMS-EMOA)

Remarks Change in each objective was studied by varying optimal value of each design parameter. From the Pareto-optimal front, closest solution to the ideal point was selected. This study improves heat transfer and reduces pressure drop in the heat exchanger. A correlation between optimum values of both objectives was established. TOPSIS and LINMAP were used for selecting one of the non-dominated solutions. Tube arrangement is optimized to provide an understanding of Paretooptimal front and optimal geometries. Computational fluid dynamics simulation was used. Heat exchanger network was explicitly optimized. Social impact, measured by number of jobs created by the project, was calculated for each Pareto-optimal solution. Pareto risk index was proposed to handle the optimal trade-off between the two objectives. Selected optimal solutions were verified through experimentation. Dynamic heat dissipation of plate-fin heat sink is investigated using finite element software ANSYS. Several exchanger reassignment strategies were developed. SOO and MOO results are compared. A continuous approach for handling variable heat capacity streams in HEN retrofitting is proposed and tested. Three scenarios of plant operation were optimized and compared. Relative merits of two MOO techniques are discussed in the context of this application. This study considered process intensification via minimizing total quantity of stored materials. Environmental indicator includes overall contribution to global warming based on life cycle assessment. Neural network and industrial data were used to develop the process model. Total annual cost and CO2 emissions were calculated for the optimal solutions. The best design is the hybrid system using a thermally coupled distillation column. A four-step method for the design of hybrid separations was proposed and applied. Effect of model and cost parameters on the hybrid process performance was investigated.

www.sciencedirect.com

Multi-objective optimization for process design Rangaiah, Sharma and Sreepathi 55

Table 3 (Continued ) Application Benzene-toluene-xylene separation in a large aromatics plant (four production rate scenarios) [30] Hydrodealkylation of toluene (two scenarios) [31]

Objectives

Techniques

Exergy destruction and environmental impact; and annual cost and environmental impact

Multi-objective GA

Annual cost and five other objectives for environmental impact (six objectives in total)

NSGA-II with some modifications

Synthesis gas production by CO2 reforming and partial oxidation of CH4 [32] Design of preflash drums for the crude distillation unit in petroleum refineries [33]

CH4 conversion, CO selectivity and H2/CO ratio (three objectives in total) CO2 emissions and residue yield

Normal Boundary Intersection with gravitational search method (GSM) and PSO NSGA-II

Cogeneration system using CO2 as working fluid [34]

Exergy efficiency and total length of the heat exchangers

Multi-objective GA in Matlab optimization toolbox (gamultiobj)

Low grade waste heat recovery using Rankine cycle [35]

Exergy efficiency and total capital cost

NSGA-II

Ammonia-water power system [36]

Exergy efficiency, heat transfer capability and turbine size (three objectives in total)

NSGA-II

Solar assisted heat pump (SAHP) [37]

Coefficient of performance and total annual cost

MOPSO

Parabolic trough receiver with perforated plate inserts [38]

Nusselt number for heat transfer performance and pressure drop for fluid friction

A variant of NSGA-II

Exhaust gas recirculation (EGR) in a diesel engine [39]

Brake specific fuel consumption and brake specific nitrogen oxide

Multi-objective GA

Underwater compressed air energy storage system [40]

Round-trip efficiency, cumulative cost rate function and operating profit (three objectives in total)

Controlled elitist GA, a variant of NSGA-II

in 24 countries; of these, researchers from China and Iran have contributed 12 and 10 articles respectively; number of papers from other countries is five or less. This is an indication of number of researchers currently active in MOO for energy-related applications, in these countries. Applications for energy efficient processes

A total of 27 articles were published on MOO of energy efficient processes. Applications in these articles are summarized in Table 3; related papers are placed together in www.sciencedirect.com

Remarks If environmental impact of the plant is due to utility consumption, then exergy and economic optimization is sufficient; no need to consider environmental impact objective. For the HDA process, natural gas turbine was found to be less economical but more environmental friendly compared to fuel oil turbine. Radar graph (spider plot) is used for comparing values of objectives of a few selected solutions. TOPSIS and FUCA were used for solution selection from the set of Pareto-optimal solutions. For this application, PSO was found to be computationally efficient compared to GSM. For the system studied, crude preflash reduces CO2 emissions at high residue yields, and has no benefit at low residue yields. A new cycle based on Brayton power cycle and ejector-expansion refrigeration cycle was proposed and analyzed. It can be driven by low temperature heat sources, and provides heating, cooling and power simultaneously. An increase in exergy efficiency increases total capital cost. From the Pareto-optimal front, closest solution to the ideal point was selected. An ammonia-water power system with LNG as heat sink is studied for recovering low grade waste heat from exhaust gas. SAHP was analyzed for five working fluids, of which the best is R245fa for both performance criteria. Optimization procedure includes computational fluid dynamics simulation, design of experiments and response surface methodology. Several optimal solutions were selected based on the single weighted objective function. EGR system combines high and low pressure systems to improve combustion of diesel engines Pseudo weight vector multi-criteria decision making approach was used to select a single preferred solution

this and subsequent tables. Studies in Table 3 cover a wide range of applications such as heat exchangers and their networks, water networks, reverse osmosis, distillation, absorption, crystallization, solar assisted heat pump, energy storage system, benzene–toluene–xylene separation and hydrodealkylation (HDA) processes. In these applications, process economics (e.g., annual cost, production cost, capital cost, energy consumption and utility cost) is often one objective function. Apart from the economic criterion, another objective such as approach temperature, pressure Current Opinion in Chemical Engineering 2015, 10:49–62

56 Process systems engineering

Table 4 MOO for the design and operation of biofuel processes Application

Objectives

Techniques

Oleic acid esterification to bio-diesel [41]

Oleic acid conversion and methyl oleate yield

MOO using artificial neural network and GA

Esterification of oleic acid to biodiesel [42]

Methyl oleate yield and oleic acid conversion

MATLAB: gatool for SOO and gamultiobj (a variant of NSGA-II) for MOO

Biodiesel from waste cooking oil (involving both esterification and transesterification) [43]

Profit and heat duty or organic duty (two sets)

NSGA-II

Biodiesel production from soybean oil [44]

VOC emissions, profit and number of operators/jobs (two sets)

MATLAB: gamultiobj (a variant of NSGA-II)

Thermochemical conversion of biomass into methanol (two scenarios) [45]

Capital investment and equivalent efficiency

Multi-objective EA

Integrated energy system based on bio-mass (pine sawdust) [46]

Exergy efficiency and total cost rate (which includes cost rate of environmental impact)

NSGA-II

Biofeedstocks-to-biofuels system design [47]

Cost, CO2 emissions and energy recovery (three objectives in total)

Minimax approach using Solver tool in Excel

Biogas production supply chain [48]

Profit and relative sustainability index

e-Constraint

Biofuel supply chain (BSC) [49]

Profit, greenhouse gas emission and fossil energy input (three objectives in total)

e-Constraint with IBM ILOG CPLEX

Current Opinion in Chemical Engineering 2015, 10:49–62

Remarks Ionic liquid catalyst was used and evaluated for esterification of oleic acid. MOO using ANN-GA was concluded to be more effective than SOO using response surface methodology. Artificial neural network model was developed for experimental results, and then used for optimization. This kinetic study indicates that the magnetic ionic liquid is a promising new catalyst for oleic acid esterification. Two alternate processes were simulated in Aspen Plus and then optimized. The process with three transesterification reactors and product separation before washing is better. SuperPro Designer was used for process simulation. It was linked with MATLAB via MATLAB graphical user interface based on COM technology. Process design with energy integration was performed. Cogeneration can improve the overall efficiency of the process. The integrated system has a biomass combustor, an organic Rankine cycle to produce electricity, an absorption chiller for cooling, heat exchanger, a proton exchange membrane electrolyzer to produce hydrogen, a domestic water heater and a reverse osmosis desalination unit. A superstructure of 17 production paths was considered. Optimal results for each of the three feedstocks (rapeseed, corn and switch grass) were discussed. Both direct and indirect effects on the environment are considered. Life cycle assessment based model with multi-conversion pathways for different biomass, was used to design a BSC for China. Effect of price change on the optimal solutions was studied.

www.sciencedirect.com

Multi-objective optimization for process design Rangaiah, Sharma and Sreepathi 57

Table 5 MOO applications in power generation and CO2 capture Application

Objectives

Techniques

Multi-generation energy system [50]

Total cost rate of the system and exergy efficiency

NSGA-II

Hybrid energy system (Photovoltaic, wind, battery bank and internal combustion generator) [51]

Levelized energy cost, unmet load fraction, wasted renewable energy and fuel consumption (four objectives in total) Initial capital cost, CO2 equivalent and levelized energy cost (three objectives in total) Total cost, unmet load and fuel emission (three objectives in total)

e-state evolutionary algorithm

Environmental/economic dispatch in power system operation [54]

Total cost of generation and atmospheric pollutants from thermal units

Multi-objective scatter search

Short-term economic & environmental hydrothermal scheduling (SEEHTS) for one day operation of 4 hydro and three thermal plants [55]

Fuel cost and pollutants emitted

Non-dominated sorting gravitational search with chaotic mutation (NSGSACM)

Short-term economic & environmental hydrothermal scheduling for two case studies [56]

Fuel cost and pollutants emitted

Improved MO gravitational search algorithm (IMOGSA)

Energy quality management for a district in China [57]

Life cycle CO2 equivalent, life cycle cost and exergy efficiency (three objectives in total)

Multi-objective GA

Energy generation portfolio under uncertainty [58]

Total risk and cost per kWh

MOPSO

Heat and power supply system (two cases) [59]

Fuel cost and gas (Sox, NOx and CO2) emissions

Multi-objective line-up competition algorithm

CO2 capture from a power plant [60]

Capital cost and net power output

NSGA-II

Coal fired power plants [61]

Energetic efficiency and cost of electricity

MODE

Hybrid energy system: photovoltaic, wind, battery bank and internal combustion generator [52] Hybrid renewable energy system [53]

www.sciencedirect.com

e-state evolutionary algorithm

e-Constraint, PSO

Remarks Total cost rate is the sum of environmental and economic costs. Closest point to the ideal point was selected from the Pareto-optimal front. Fuzzy TOPSIS and level diagrams were used to select one solution from the set of alternate solutions. The proposed approach determines relative weights of objectives. Integration of renewable and conventional energy sources is beneficial. The system includes wind, photovoltaic, diesel generator, batteries, fuel cell, electrolyzer and H2 tank. The obtained results were compared with those obtained using SPEA. Clustering approach was used to select one of the Paretooptimal solutions. MODE and MOPSO algorithms were also used in this study. Improved scatter search finds better minimum fuel cost than other algorithms, whereas MODE and MOPSO gave better results for emissions. NSGSA-CM performance was compared with other MOO techniques on 6 test functions and for SEEHTS application. Strategies for handling constraints in this application are presented. A method based on feasible space was proposed and used for constraints in this application. IMOGSA performed better than NSGA-II and HMOCA. Three typical days in winter, midseason and summer were considered. Effects of economic, environmental and energy performance parameters were analyzed. Viability of solar photovoltaic projects was evaluated for 4 large investor-owned utilities in Florida. Performance of the proposed MOO algorithm was compared with NSGA-II, SPEA, niched Pareto GA and fuzzy clustering based PSO. Performance of different amine solvents for CO2 absorption was compared. Di-ethanolamine found to be better than monoethanolamine. Optimal results indicate that the current industrial design can be improved for both objectives. Effect of model parameters on the Paretooptimal front was analyzed.

Current Opinion in Chemical Engineering 2015, 10:49–62

58 Process systems engineering

Table 5 (Continued ) Application

Objectives

Techniques

Geothermal power plant [62]

Specific work output and specific heat exchanger area

MOPSO

Photo voltaic-wind turbine (two case studies) [63]

Emissions, cost per year and social acceptability (three objectives)

Multi-objective GA

Lithium-ion battery model [64]

Sum of squared errors for terminal voltage and for surface temperature, with each of these at 15 and 308C (four objectives in total)

NSGA-II

Utility system design with emission abatement [65]

Economic cost, environmental effect and exergy efficiency (three objectives in total) Exergy efficiency and total cost rate

Augmented e-constraint with CPLEX

Ice thermal energy storage system for cooling inlet air to a gas turbine cycle [66]

drop, environmental impact, conversion, selectivity, greenhouse gas emissions and exergy efficiency is considered depending upon the application. As shown in Table 3, NSGA-II, MOPSO and e-constraint are the commonly used MOO techniques for optimizing energy efficient processes. In order to select one optimal solution from the set of non-dominated solutions, TOPSIS, LINMAP, FUCA (Faire Un Choix Adequat), closest solution to the ideal point and pseudo weight vector multi-criteria decision making approach are employed in six studies. Applications in production of biofuels

The nine articles published on optimization of biofuel processes for multiple objectives, cover a range of applications such as biofuel supply chain, biogas, bio-diesel, methanol and biofuel production. In these studies on biofuels, process economics (e.g., profit and capital investment) is often one objective function. Apart from process economics, other objectives such as conversion, yield, sustainability index, exergy efficiency and emissions are considered as appropriate to the application. Further, e-constraint, MOGA and NSGA-II are commonly used MOO techniques for optimizing biofuel production processes (Table 4). Applications in power generation and CO2 capture

A total of 17 articles were published on optimization of power generation and CO2 capture for multiple objectives. These studies cover applications such as multi-generation energy system, hybrid energy system, coal fired/geothermal power plants and hydrothermal scheduling. In the Current Opinion in Chemical Engineering 2015, 10:49–62

Multi-objective GA

Remarks A method for optimizing superheater and recuperator in a binary geothermal power plant was developed and tested. The proposed approach is effective and flexible for optimal design of hybrid power generation system. MCDM-PROMETHEE II was used as the decision making tool. A multi-objective parameter estimation using NSGA-II and TOPSIS was proposed and applied to experimental data for two types of real batteries. Optimal results obtained are found to be in good agreement with the experimental data. The MINLP problem is transformed into a MILP problem, and then solved by CPLEX. Total cost rate includes capital cost, operating cost and social cost of emissions. TOPSIS was used to choose one solution from the obtained Pareto-optimal front

MOO studies on power generation, exergy efficiency is often one performance criterion whereas environmental effect (e.g., CO2 emission) is the performance criterion in the study on CO2 capture. Apart from this, another objective such as total cost rate (which is equal to capital cost, fuel cost and emission cost), fuel cost, unmet (power) load, fuel emission, social acceptability, specific work output and total risk is considered as appropriate to the application. The e-constraint and multi-objective GA including NSGAII are commonly used MOO techniques for optimizing power generation and CO2 capture processes. Further, TOPSIS, MCDM-PROMETHEE II, clustering and closest solution to the ideal point are employed in six studies to select one optimal solution from the set of alternate solutions (Table 5). Applications in fuel cell and hydrogen production

There are a total of 12 journal papers on optimization of fuel cells and hydrogen production for multiple objectives. These studies cover applications such as solid oxide fuel cell, methanol fuel cell, polymer electrolyte membrane fuel cell, integrated gasification fuel cell cycle (IGFC), hydrogen supply chain, hydrogen networks and hydrogen production from power plants. In the MOO studies on fuel cells, exergy efficiency is often one objective function, and second objective such as power output, breakeven energy cost, power density, total cost rate and total energy is considered as appropriate to the application. In the MOO studies on hydrogen production, process economics (e.g., capital cost, operating cost and total cost rate) is often one performance www.sciencedirect.com

Multi-objective optimization for process design Rangaiah, Sharma and Sreepathi 59

criterion, and the second objective such as total risk, global warming potential and total pollutant emissions is considered depending on the application. The e-constraint, weighted sum method, multi-objective GA and self-adaptive gravitational search algorithm are commonly

used MOO techniques for optimizing fuel cell and hydrogen production processes. For selecting one optimal solution from the Pareto-optimal front, TOPSIS and fuzzy membership function are used in seven studies (Table 6).

Table 6 MOO applications in fuel cell and hydrogen production Application

Objectives

Techniques

Polymer electrolyte membrane fuel cell [67]

Power output and system cost

Weighted sum method with GA

Polymer electrolyte membrane fuel cell [68]

Power density, exergetic efficiency and unit cost (three objectives in total)

NSGA-II

Solid oxide fuel cell stacks [69]

Breakeven energy cost and power output or electrical efficiency (two sets) Power output, energy and exergy efficiencies (three objectives in total) Total exergy efficiency and total cost rate of the system

Weighted sum method with GA

Direct methanol fuel cell [70]

Hybrid solid oxide fuel cell and micro gas turbine system for heat and power [71]

Weighted sum method with GA Multi-objective GA in Matlab optimization toolbox (gamultiobj)

Integrated gasification fuel cell cycle (IGFC) [72]

Energy generated by fuel cells and energy generated by gas turbines

Multi-objective GA

Combined molten carbonate fuel cell-gas turbine system [73]

Exergetic efficiency and total cost rate (= sum of economic and environmental costs) of the system

Multi-objective GA

Heat, power and H2 production planning of fuel cell power plants [74]

Operating cost, emission and voltage deviation (three objectives)

u-Self adaptive gravitational search algorithm

Heat, power and H2 production planning of molten carbonate fuel cell power plants [75]

Total cost, emission and voltage deviation (three objectives)

Self-adaptive learning batinspired algorithm

Heat, power and H2 production from PEM fuel cell power plants (two scenarios) [76] Hydrogen supply chain [77]

Total operating cost and total pollutant emissions

Modified gravitational search algorithm

Total cost, global warming potential and safety risk (three objectives in total)

e-Constraint with CPLEX

Hydrogen network in a petroleum refinery [78]

Total annual cost and CO2 emission

Adaptive weighted sum method with DICOPT

www.sciencedirect.com

Remarks Effect of membrane types and hydrogen cost on the non-dominated solutions obtained was analyzed. Model was validated before using in optimization. MOO results are compared with SOO results. The Pareto-optimal solution selected by fuzzy decision making is better than SOO solution. Sensitivity analysis of optimization results was performed for H2 cost and fuel utilization. The study concluded that fuel cell operating parameters have great influence on the system performance. Total cost rate includes investment, operating and emissions cost. Effect of fuel cost, investment cost and power output on optimum design was studied. TOPSIS was used as the decision making tool. IGFC was optimized using a superstructure model, and MOO results for biomass and natural gas as fuel were analyzed and compared. Sensitivity analysis of the Paretooptimal solutions to change in fuel unit cost and interest rate was studied. TOPSIS was used to choose one optimal solution from the Paretooptimal front. Adaptive mutation was included to improve gravitation search. Uncertainty was considered to obtain realistic results. Fuzzy membership function was used to determine the best compromised solution. Uncertainties in electrical and thermal loads, oxygen, H2 and CO2 partial pressures, and operational temperature were considered. Fuzzy decision making was used for the final solution selection. This study considered uncertainties in loads, production and cost. Fuzzy approach was used to find the best compromised solution. Both SOO and MOO were performed, and optimal results are compared. Pareto-optimal solutions were ranked by both TOPSIS and modified synthetic evaluation method (M-TOPSIS). Both SOO and MOO of H2 network in a refinery were studied. Purification technologies and fuel types are key issues in this optimization.

Current Opinion in Chemical Engineering 2015, 10:49–62

60 Process systems engineering

Concluding remarks This article provides an overview to MOO and its techniques, and then summarizes MOO applications for the design and operation of energy efficient chemical processes, biofuel processes, power generation, CO2 capture, fuel cells and hydrogen production, reported in 65 journal papers between January 2013 and February 2015. Researchers from China and Iran contributed 22 of these articles, and the remaining 43 papers came from 22 different countries. Four journals, namely, Applied Energy, Applied Thermal Engineering, Energy and Energy Conversion and Management, have published about 10 papers each, accounting for 43 of the 65 papers of interest to chemical engineers. Majority of the application problems reviewed in this paper were optimized for two objectives. In total, 15 problems were optimized for three objectives, two applications were optimized for four objectives [51,64] and one application considered six objectives [31]. Both stochastic and deterministic MOO techniques have been used to solve MOO application problems reviewed; in particular, genetic algorithm based MOO and e-constraint techniques are popular. Further, 20 of 65 studies have used one or more selection techniques for choosing one solution from the Pareto-optimal front. In future, optimization of chemical engineering applications including those on energy efficiency of chemical processes should consider two or more relevant objectives such as economic, environmental and safety criteria. This seems to be already happening, as can be seen from recent papers on process optimization in chemical engineering journals. In addition to obtaining Pareto-optimal solutions, it is desirable to apply a suitable technique to choose one or more optimal solutions and then discuss the practicality of the corresponding values of decision variables for implementation. In general, MOO application studies should present values of decision variables corresponding to the Pareto-optimal solutions and discuss their trends for completeness.

References and recommended reading1 Papers of particular interest, published within the period of review, have been highlighted as:  of special interest  of outstanding interest 1.

Edgar TF, Himmelblau DM, Lasdon LS: Optimization of Chemical Processes. 2nd ed.. McGraw-Hill; 2001.

2.

Ravindran A, Raqsdell KM, Reklaitis GV: Engineering Optimization: Methods and Applications. 2nd ed.. Chichester: John Wiley & Sons; 2006.

1 The first 13 references are books or book chapters, which will be of interest to researchers considering MOO for their applications. References 14 to 78 are journal papers describing the applications summarized in Tables 3 to 6. Readers are referred to Tables 3 to 6 for the application(s) of their interest.

Current Opinion in Chemical Engineering 2015, 10:49–62

3.

Rangaiah GP: Stochastic Global Optimization: Techniques and Applications in Chemical Engineering. Singapore: World Scientific; 2010.

Rangaiah GP: Multi-Objective Optimization: Techniques and Applications in Chemical Engineering. Singapore: World Scientific; 2009. This book has chapters describing selected MOO techniques and many applications of MOO in chemical engineering.

4. 

5.

Masuduzzaman, Rangaiah GP: Multi-objective optimization applications in chemical engineering. In Multi-Objective Optimization: Techniques and Applications in Chemical Engineering. Edited by Rangaiah . Singapore: World Scientific; 2009.

6. 

Sharma S, Rangaiah GP: Multi-objective optimization applications in chemical engineering. In Multi-Objective Optimization in Chemical Engineering: Developments and Applications. Edited by Rangaiah GP, Bonilla-Petriciolet A. Chichester: John Wiley & Sons; 2013. This book chapter reviews about 230 articles on MOO in chemical engineering and related areas, published from the year 2007 to June 2012. It is a convenient and useful reference for readers interested in MOO applications in chemical engineering.

7. Valadi J, Siarry P: Applications of Metaheuristics in Process Engineering. Heidelberg: Springer; 2014.  This recent book has several chapters on MOO techniques and their applications. 8.

Rangaiah GP: Chemical Process Retrofitting and Revamping: Techniques and Applications. John Wiley & Sons: Chichester; 2015:. (in press).

9. Deb K: Multi-Objective Optimization Using Evolutionary Algorithm.  Chichester, UK: John Wiley & Sons; 2001. This is an excellent book covering the basic concepts of multi-objective optimization and describes evolutionary algorithms extensively. 10. Miettinen K: Graphical illustration of Pareto optimal solutions. In Multi-Objective Programming and Goal Programming: Theory and Applications. Edited by Tanino T, Tanaka T, Inuiguchi M. Berlin, Heidelberg: Springer-Verlag; 2003. 11. Coello Coello CA, Lamont GB, Van Veldhuizen DA: Evolutionary Algorithms for Solving Multi-Objective Problems. 2nd ed.. Berlin, Heidelberg: Springer; 2007. 12. Miettinen K: Non-Linear Multi-Objective Optimization. Kluwer Academic Publishers: Boston; 1999. 13. Hwang CL, Yoon K: Multiple Attribute Decision Making: Methods and Applications — A State-of-the-Art Survey, vol 186 of Lecture  Notes in Economics and Mathematical Systems. Berlin: SpringerVerlag; 1981. In this book, Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) is described with an example. 14. Alinia Kashani AH, Maddahi A, Hajabdollahi H: Thermaleconomic optimization of an air-cooled heat exchanger unit. Appl Therm Eng 2013, 54:43-55. 15. Juan D, Qin QZ: Multi-objective optimization of a plain fin-andtube heat exchanger using genetic algorithm. Therm Eng 2014, 61:309-317. 16. Sanaye S, Modarrespoor D: Thermal-economic multi-objective optimization of heat pipe heat exchanger for energy recovery in HVAC application using genetic algorithm. Therm Sci 2014, 18:375-391. 17. Daroczy L, Janiga G, Thevenin D: Systematic analysis of the heat exchanger arrangement problem using multi-objective genetic optimization. Energy 2014, 65:364-373. 18. Lira-Barraga´n LF, Ponce-Ortega JM, Serna-Gonza´lez M, ElHalwagi MM: Synthesis of integrated absorption refrigeration systems involving economic and environmental objectives and quantifying social benefits. Appl Therm Eng 2013, 52:402-419. 19. Chen Y, Peng B, Hao X, Xie G: Fast approach of Pareto-optimal solution recommendation to multi-objective optimal design of serpentine-channel heat sink. Appl Therm Eng 2014, 70:263-273. www.sciencedirect.com

Multi-objective optimization for process design Rangaiah, Sharma and Sreepathi 61

20. Rao RV, Waghmare GG: Multi-objective design optimization of a plate-fin heat sink using a teaching-learning-based optimization algorithm. Appl Therm Eng 2015, 76:521-529.

38. Mwesigye A, Bello-Ochende T, Meyer JP: Multi-objective and thermodynamic optimization of a parabolic trough receiver with perforated plate inserts. Appl Therm Eng 2015, 77:42-56.

21. Sreepathi BK, Rangaiah GP: Improved heat exchanger network retrofitting using exchanger reassignment strategies and multi-objective optimization. Energy 2014, 67:584-594.

39. Park J, Song S, Lee KS: Numerical investigation of a dual-loop EGR split strategy using a split index and a multi-objective Pareto optimization. Appl Energy 2015, 142:21-32.

22. Sreepathi BK, Rangaiah GP: Retrofitting of heat exchanger networks involving streams with variable heat capacity: application of single and multi-objective optimization. Appl Therm Eng 2015, 72:677-684.

40. Cheung BC, Carriveau R, Ting DSK: Multi-objective optimization of an underwater compressed air energy storage system using genetic algorithm. Energy 2014, 74:396-404.

23. Porzio GF, Colla V, Matarese N, Nastasi G, Branca TA, Amato A, Fornai B, Vannucci M, Bergamasco M: Process integration in energy and carbon intensive industries: an example of exploitation of optimization techniques and decision support. Appl Therm Eng 2013, 70:1148-1155. 24. Porzio GF, Nastasi G, Colla V, Vannucci M, Branca TA: Comparison of multi-objective optimization techniques applied to off-gas management within an integrated steelwork. Appl Energy 2014, 136:1085-1097. 25. Va´zquez-Castillo JA, Ponce-Ortega JM, Segovia-Herna´ndez JG, El-Halwagi MM: A multi-objective approach for property-based synthesis of batch water networks. Chem Eng Process: Process Intens 2013, 65:83-96. 26. Antipova E, Boer D, Cabeza LF, Guille´n-Gosa´lbez G, Jime´nez L: Multi-objective design of reverse osmosis plants integrated with solar Rankine cycles and thermal energy storage. Appl Energy 2013, 102:1137-1147. 27. Kusiak A, Wei X: Optimization of the activated sludge process. J Energy Eng 2013, 139:12-17. 28. Bravo-Bravo C, Segovia-Herna´ndez JG, Herna´ndez S, Go´mezCastro FI, Gutie´rrez-Antonio C, Briones-Ramı´rez A: Hybrid distillation/melt crystallization process using thermally coupled arrangements: optimization with evolutive algorithms. Chem Eng Process: Process Intens 2013, 67:25-38. 29. Micovic J, Beierling T, Lutze P, Sadowski G, Go´rak A: Design of hybrid distillation/melt crystallisation processes for separation of close boiling mixtures. Chem Eng Process: Process Intens 2013, 67:16-24. 30. Sahraei MH, Farhadi F, Boozarjomehry RB: Analysis and interaction of exergy, environmental and economic in multiobjective optimization of BTX process based on evolutionary algorithm. Energy 2013, 59:147-156. 31. Ouattara A, Pibouleau L, Azzaro-Pantel C, Domenech S: Economic and environmental impacts of the energy source for the utility production system in the HDA process. Energy Convers Manage 2013, 74:129-139. 32. Ganesan T, Elamvazuthi I, Ku Shaari KZ, Vasant P: Swarm intelligence and gravitational search algorithm for multiobjective optimization of synthesis gas production. Appl Energy 2013, 103:368-374. 33. Al-Mayyahi MA, Hoadley AFA, Rangaiah GP: Energy optimization of crude oil distillation using different designs of pre-flash drums. Appl Therm Eng 2014, 73:1204-1210. 34. Jamali A, Ahmadi P, Jaafar MNM: Optimization of a novel carbon dioxide cogeneration system using artificial neural network and multi-objective genetic algorithm. Appl Therm Eng 2014, 64:293-306. 35. Wang J, Yan Z, Wang M, Li M, Dai Y: Multi-objective optimization of an organic Rankine cycle (ORC) for low grade waste heat recovery using evolutionary algorithm. Energy Convers Manage 2013, 71:146-158.

41. Fauzi AHM, Saidina Amin NA: Optimization of oleic acid esterification catalyzed by ionic liquid for green biodiesel synthesis. Energy Convers Manage 2013, 76:818-827. 42. Fauzi AHM, Amin NAS, Mat R: Esterification of oleic acid to biodiesel using magnetic ionic liquid: multi-objective optimization and kinetic study. Appl Energy 2014, 114:809-818. 43. Patle DS, Sharma S, Ahmad Z, Rangaiah GP: Multi-objective optimization of two alkali catalyzed processes for biodiesel from waste cooking oil. Energy Convers Manage 2014, 85:361-372. 44. Woinaroschy A: Multi-objective optimal design for biodiesel sustainable production. Fuel 2014, 135:393-405. 45. Peduzzi E, Tock L, Boissonnet G, Mare´chal F: Thermo-economic evaluation and optimization of the thermo-chemical conversion of biomass into methanol. Energy 2013, 58:9-16. 46. Ahmadi P, Dincer I, Rosen MA: Thermoeconomic multiobjective optimization of a novel biomass-based integrated energy system. Energy 2014, 68:958-970. 47. Eason JP, Cremaschi S: A multi-objective superstructure optimization approach to biofeedstocks-to-biofuels systems design. Biomass Bioenergy 2014, 63:64-75. 48. Kravanja Z, Cˇucˇek L: Multi-objective optimisation for generating sustainable solutions considering total effects on the environment. Appl Energy 2013, 101:67-80. 49. Liu Z, Qiu T, Chen B: A study of the LCA based biofuel supply chain multi-objective optimization model with multiconversion paths in China. Appl Energy 2014, 126:221-234. 50. Ahmadi P, Dincer I, Rosen MA: Thermodynamic modeling and multi-objective evolutionary-based optimization of a new multigeneration energy system. Energy Convers Manage 2013, 76:561-570. 51. Perera ATD, Attalage RA, Perera KKCK, Dassanayake VPC: A hybrid tool to combine multi-objective optimization and multicriterion decision making in designing standalone hybrid energy systems. Appl Energy 2013, 107:412-425. 52. Perera ATD, Attalage RA, Perera KKCK, Dassanayake VPC: Designing standalone hybrid energy systems minimizing initial investment, life cycle cost and pollutant emission. Energy 2013, 54:220-230. 53. Sharafi M, ELMekkawy TY: Multi-objective optimal design of hybrid renewable energy systems using PSO-simulation based approach. Renew Energy 2014, 68:67-79. 54. de Athayde Costa e Silva M, Klein CE, Mariani VC, Dos Santos Coelho L: Multiobjective scatter search approach with new combination scheme applied to solve environmental/ economic dispatch problem. Energy 2013, 53:14-21. 55. Tian H, Yuan X, Ji B, Chen Z: Multi-objective optimization of short term hydrothermal scheduling using non-dominated sorting gravitational search algorithm with chaotic mutation. Energy Convers Manage 2014, 81:504-519.

36. Wang J, Yan Z, Wang M, Dai Y: Thermodynamic analysis and optimization of an ammonia-water power system with LNG (liquefied natural gas) as its heat sink. Energy 2013, 50:513-522.

56. Li C, Zhou J, Lu P, Wang C: Short-term economic environmental hydrothermal scheduling using improved multi-objective gravitational search algorithm. Energy Convers Manage 2015, 89:127-136.

37. Khorasaninejad E, Hajabdollahi H: Thermo-economic and environmental optimization of solar assisted heat pump by using multi-objective particle swarm algorithm. Energy 2014, 72:680-690.

57. Lu H, Yu Z, Alanne K, Xu X, Fan L, Yu H, Zhang L, Martinac I: Parametric analysis of energy quality management for district in China using multi-objective optimization approach. Energy Convers Manage 2014, 87:636-646.

www.sciencedirect.com

Current Opinion in Chemical Engineering 2015, 10:49–62

62 Process systems engineering

58. Zeng Z, Nasri E, Chini A, Ries R, Xu J: A multiple objective decision making model for energy generation portfolio under fuzzy uncertainty: case study of large scale investor-owned utilities in Florida. Renew Energy 2015, 75:224-242.

69. Forough AB, Roshandel R: Multi objective optimization of solid oxide fuel cell stacks considering parameter effects: fuel utilization and hydrogen cost. J Renew Sustain Energy 2013, 5.

59. Shi B, Yan L-X, Wu W: Multi-objective optimization for combined heat and power economic dispatch with power transmission loss and emission reduction. Energy 2013, 56:135-143.

70. Mert SO, O¨zc¸elik Z: Multi-objective optimization of a direct methanol fuel cell system using a genetic-based algorithm. Int J Energy Res 2013, 37:1256-1264.

60. Lee AS, Eslick JC, Miller DC, Kitchin JR: Comparisons of amine solvents for post-combustion CO2 capture: a multiobjective analysis approach. Int J Greenh Gas Control 2013, 18:68-74. 61. Wang L, Yang Y, Dong C, Morosuk T, Tsatsaronis G: Multiobjective optimization of coal-fired power plants using differential evolution. Appl Energy 2014, 115:254-264. 62. Clarke J, Mcleskey JT: Multi-objective particle swarm optimization of binary geothermal power plants. Appl Energy 2015, 138:302-314. 63. Alsayed M, Cacciato M, Scarcella G, Scelba G: Design of hybrid power generation systems based on multi criteria decision analysis. Solar Energy 2014, 105:548-560. 64. Zhang L, Wang L, Hinds G, Lyu C, Zheng J, Li J: Multi-objective optimization of lithium-ion battery model using genetic algorithm approach. J Power Sources 2014, 270:367-378. 65. Luo X, Hu J, Zhao J, Zhang B, Chen Y, Mo S: Multi-objective optimization for the design and synthesis of utility systems with emission abatement technology concerns. Appl Energy 2013, 136:1110-1131. 66. Shirazi A, Najafi B, Aminyavari M, Rinaldi F, Taylor RA: Thermaleconomic-environmental analysis and multi-objective optimization of an ice thermal energy storage system for gas turbine cycle inlet air cooling. Energy 2014, 69:212-226. 67. Abbas Tahmasbi A, Hoseini A, Roshandel R: A new approach to multi-objective optimisation method in PEM fuel cell. Int J Sustain Energy 2013, 5:283-297. 68. Sayyaadi H, Esmaeilzadeh H: Determination of optimal operating conditions for a polymer electrolyte membrane fuel cell stack: optimal operating condition based on multiple criteria. Int J Energy Res 2013, 37:1872-1888.

Current Opinion in Chemical Engineering 2015, 10:49–62

71. Sanaye S, Katebi A: 4E analysis and multi objective optimization of a micro gas turbine and solid oxide fuel cell hybrid combined heat and power system. J Power Sources 2014, 247:294-306. 72. Naraharisetti PK, Lakshminarayanan S, Karimi IA: Design of biomass and natural gas based IGFC using multi-objective optimization. Energy 2014, 73:635-652. 73. Mamaghani AH, Najafi B, Shirazi A, Rinaldi F: Exegetic, economic and environmental evaluations and multi-objective optimization of a combined molten carbonate fuel cell-gas turbine system. Appl Therm Eng 2015, 77:1-11. 74. Niknam T, Bornapour M, Gheisari A: Combined heat, power and hydrogen production optimal planning of fuel cell power plants in distribution networks. Energy Convers Manage 2013, 66:11-25. 75. Niknam T, Bornapour M, Ostadi A, Gheisari A: Optimal planning of molten carbonate fuel cell power plants at distribution networks considering combined heat, power and hydrogen production. J Power Sources 2013, 239:513-526. 76. Niknam T, Golestaneh F, Malekpour AR: Probabilistic model of polymer exchange fuel cell power plants for hydrogen, thermal and electrical energy management. J Power Sources 2013, 229:285-298. 77. De-Leo´n Almaraz S, Azzaro-Pantel C, Montastruc L, Pibouleau L, Senties OB: Assessment of mono and multi-objective optimization to design a hydrogen supply chain. Int J Hydrogen Energy 2013, 38:14121-14145. 78. Zhou L, Liao Z, Wang J, Jiang B, Yang Y, Hui D: Optimal design of sustainable hydrogen networks. Int J Hydrogen Energy 2013, 38:2937-2950.

www.sciencedirect.com