You are on page 1of 12
Asian Journal of Seientifie Research, 2011 ISSN 1992-1454 | DOI: 10.3923ajsr.2011 © 2011 Asian Notwork for Scientific Information Reservoir Operation using Multi-objective Evolutionary Algorithms-A Review JA. Adeyemo Durban University of Technology, P.O. Box 1884, Durban 4000, South Africa ABSTRACT ‘Tho review of some applications of evolutionary algorithms to reservoir operations is presented in this study, Reservoir operation is a complex multi-objective optimization problem with many conflicting objectives and constraints. Evolutionary algorithms are stochastic search algorithms which have a lot of applications in water resources management, They have the ability to generate non-dominated solutions which converge to Pareto optimal front to multi-objective problems in one simulation run, The results presented in the literatures reviewed show that evolutionary algorithms are good algorithms for solving complex, non-linear, convex and multidimensional reservoir problems. They produce tradeoffs to reservoir operation problems from which a reservoir operator ean choose a solution applicable to him, Key words: Multi-objective, evolutionary algorithm, reservoir operation, pareto front INTRODUCTION Multi-objective optimization techniques are widely used in water resources management. They are especially good optimizers with few control parameters. They are easy to apply once the objective funetions and constraints are formulated. Water resources management problems are mostly formulated as multi-objective with many conflicting objectives and constraints. Certainly, many Multi-Objective Evolutionary Algorithms (MOEAs) are developed to solve these water resources multicobjective problems. A number of improvements are made daily in the world of evolutionary research to address the shortcomings of previous algorithms, Nowadays, we have several MOFAs producing Pareto optimal solutions in few minutes, The results produced by these algorithms are useful for decision makers in ensuring equitable distribution of resourees among competing users. Evolutionary Algorithms (EAs) as robust optimization techniques have the ability to find multiple Pareto optimal solutions in one single simulation run because of their population-approach, They are general purpose stochastic search methods simulating natural selection and biological evolution (Galman et al., 2007). They maintain a population of potential (or candidate) solutions to a problem, They are biologically-inspired optimization algorithms, imitating the process of natural evolution and are becoming important optimization tool for several real world applications. They use a set of solutions ealled population to converge to the optimal solution. They are less susceptible to problem dependent characteristics, such as the shape of the Pareto front (convex, concave or even discontinuous) and the mathematical properties of the search space, whereas these issues are of concerns for mathematical programming techniques for mathematical tractability (Rakesh and Babu, 2005). Many applications of evolutionary algorithm in water resources are widely available in the literatures. The application of Differential Evolution (DE) algorithm to a Asian J. Sei. Rea., 2011 constrained optimization problem of irrigation water use is presented by Adeyemo and Otieno (20094). Invigation water, amounting to 10 Mm’ is optimized to irrigate 2500 ha of land where 16 different erops are planted on different areas of land. Also a comparative study using linear programming (LP) is performed. The results are compared to study the effectiveness of the DE algorithm. The different areas of land where the crops are to be planted to maximize the Total Income (TI) in monetary terms (South African Rand, ZAR) are optimized. Ten strategies of DE are tested with this problem varying the population size (NP), crossover constant (CR) and weighting factor (F), Ibis found that strategy 1, DE/rand-1-bin, with values of NP, CR and F of 160, 0.95 and 0.5, respectively, obtains the best solution most efficiently. DE is considered comparable to LP in solving this optimization problem, Also, DE techniques are applied to determine the optimum cropping pattern in the Lower Orange catchments of South Africa (Otieno and Adeyemo, 2010) Three different cropping patterns are studied to determine the best cropping pattern for the study area in terms of maximum total income generated from farming, irrigation water use and total man-days required on the farm, The objective of the study is to explore the ten different strategies of DE to determine the optimum cropping pattern in each of the farmlands studied. Each farmland is studied to determine the best DE strategy for this model, From the analysis of the results, DE strategies with binomial erossover method outperform DE strategies with exponential crossover method. It is suggested that DE/best/I/bin is the best strategy for this model taking the minimum number of iterations before convergence and the maximum total net benefit generated as criteria, It can be concluded that DE can be used for determining optimum cropping pattern in a water searce environment. with necessary modifications to this model. A study by Otieno and Adeyemo (2009) presents a mathematical model of cropping patterns solved using novel Differential Evolution (DE) algorithm that. will derive the maximum total net benefit. (TNB) from farming when constrained by irrigation water. Three cropping patterns are studied to find the optimum cropping pattern for each of the farm units that will derive the maximum TNB. Multiobjective formulation for a problem is primarily due to the fact that it may not: be possible to have single solution which optimizes all objectives. Therefore an algorithm that gives a large number of alternative solutions lying on or near the Pareto-optimal front is of great practical value. Evolutionary Algorithms (EAs) are different from conventional algorithms for non-linear optimization since they use only objective function information instead of derivatives or other auxiliary information of the problems (Fan et al., 2006). In addition, they aim at finding the optima from a population of points in parallel rather than from a single point, These features make thern attractive for addressing complex engineering problems. Over the past years, a number of multi-cbjective evolutionary algorithms have been developed. Recently, multiobjective differential evolution algorithm named MDEA was suggested by Adeyemo and Otieno (2009b). The new algorithm adjusts the selection scheme of traditional DE to solve multi-objective problems. ‘The algorithm also modifies the domination criteria for the population. The offspring generated in subsequent. generations are improved hefore domination check is performed on the population in the final generation. Moreover, trial solution replaces the target solution if itis better or equal in all the objectives. The proposed algorithm has been successfully applied to five common test problems and an engineering cantilever design problem. Good spread of quality Pareto optimal solutions are achieved, The algorithm produces more Pareto optimal solutions than the previous algorithms and retains the fast convergence and diversity exhibited by DE in global optimization. The algorithm is a good choieo for solving many practical engineering problems with ease. Asian J. Sei. Rea., 2011 ‘The procedures of EAs are initialization, mutation, crossover and selection. Populations of individuals which are potential solutions are first randomly generated. Each solution is assessed by usingfitness function, A selection process is applied in each iteration to form a new population which will be better than the previous population. ‘The selection is biased towards the solution that has better fitness function, In each iteration, the solutions undergo mutation and crossover to mimic the natural evolution technique, The iteration continues until eonvergence is reached, Evolutionary ‘Algorithms (EAs) are global optimization heuristies that search for optima using a process that is analogous to Darwinian natural selection, Since their inception in the 1960s, evolutionary algorithms have been used in a tremendous array of applications. The growing popularity of evolutionary algorithms stems from their ease of implementation and robust performanes for difficult engineering and science problems. A criticism of Evolutionary Algorithms (EAs) might be the Lack of efficient. and robust generic methods to handle constraints, The most widespread approach for constrained search problems is, to use penalty methods, EAs have received increased interest during the last decade due to the ease of handling multiple objectives, A constrained optimization problem or an unconstrained multi- objective problem may, in principle, be two different, ways to pose the same underlying problem. ‘The most popular Evolutionary Algorithm (EA) is genetic algorithm. Although many genetic algorithm versions have been developed, they are still time consuming. In order to overcome this, disadvantage, the evolution strategy called DE has been recently proposed by Price and Storn (1997. It has been applied to several engineering problems in different, areas. WATER RESOURCES MANAGEMENT USING EVOLUTIONARY ALGORITHM Water resources management is a multi-objective optimization problem. It is a difficult task to estimate reservoir operating policies that maximize all the benefits provided by these reservoirs and also minimize their adverse impacts, It is a complex decision making process which will involve a number of variables, risks, uncertainties and also conflicting objectives. Reservoirs serve many purpeses. They are used to drive tnbines to generate electricity. They are used to supply water for irrigation, city and industrial uses and also for flood protection. Reservoirs may be built to satisfy a single purpose or multipurpose. Some of the purposes are conflicting in nature. For example for power generation, the reservoir should be as full as possible to increase the head, whereas for flood protection, it should be empty to provide for maximum storage of flood waters if flood oceurs Many optimization techniques have been applied to water resources management in the past. These include Linear Programming (LP); Nonlinear Programming (NLP); Dynamic Programming (DP); Stochastic Dynamic Programming (SDP); and Heuristic Programming such as Genetic Algorithms, Shuffled Complex Evolution, Fuzzy logic and Neural Networks, Differential Evolution etcetera. Sarker and Ray (2009) analyse multi-objective optimization problems and provide useful insights about solutions that are generated using population-based approached, Crop-planning problem as a multi-objective optimization model is formulated, Well-known multi-objective evolutionary algorithm called NSGAII and their proposed Multi-Objective Constrained Slgorithm (MCA) are compared, The study by Saravanan (2008) unravels the complexity of water management institutions by analysing the interactive nature of actors and rules to a particular water-related problem, using a systems approach in a hamlet in the Indian Himalayas. Azamnathulla et al. (2008) present a study to deal with the development and comparison of two models; a Genetic Algorithm (GA) and Linear Programming (LP) to be applied to real-time reservoir operation in an existing: Chiller reservoir system in India, Their performance is analysed and from Asian J. Sei. Rea., 2011 the results, the GA model is found to be superior to the LP model. Optimal water allocation and cropping patterns for the Jordan Valley, teking into consideration variations in expected incomes from agricultural production and rising water prices are studied by Doppler e¢ al. (2002). Their calculations were based on information available on water supplies, areas under irrigation and market conditions and used linear programming models for determining solutions that maximize g1oss margins and minimize potential variations in these gross margins, The results indicated that optimizing eropping patterns and the allocation of irrigation water still has a substantial potential to inerease the financial return from agriculture, Adeyemo and Otieno (20080) presents a multi-objective difforential evolution algorithm (MDEA) technique that is adapted to crop planning in a farmland in the Vaalharts irrigation scheme (VIS) in South Africa, The objectives of the model are to maximize the total net profit. (NF) in monetary terms (South African Rand, ZAR) generated on the farm by planting four different crops, maximize total planting area (mm?) and minimize the itrigation water use (m°). From the analysis of the results, it is found that the proposed MDEA can be used for solving crop planning problem and generate non-dominated solutions from where a farmer can select a solution that suits his particular situation. Further in their study, they present. four stratogies Multi-Objective Differential Evolution Algorithm (MDEA) (Adeyemo and Otieno, 2010), The four strategies namely, MDEAL, MDEA2, MDEAS and MDEA are adapted to solve the multi-objective erop planning model with multiple constraints in a farmland in South Africa, They conclude that MDEAI and MDEA2 strategies with binomial crossover method are better for solving the crop planning problem presented than MDEAS and MDEA4 strategies with exponential crossover method. Also, MDEA is proposed for irrigation planning by Adeyemo and Otieno (20090), The model in their study is adapted to Vanderkloof darn along the Orange River to determine the optimal monthly release for irrigation release requirements. It is concluded that the proposed algorithm can also be adapted to many multi- objective optimization of water resources systems in South Africa and similar area to generate optimal trade-offs. Other applications of differential evolution in water resourves management are available in the literatures (Adeyemo and Otieno, 2009a; Adeyemo et al, 2008; Janga and Nagesh, 2008, 2007. In another study, a tolerance based Fuzzy Goal Programming (FGP) and a FGP based Genetic Algorithm (GA) model for nutrient management. decision-making for rice crop planning in India are presented. In the proposed model, fuzzy goals such as fertilizer cost. and rice yield are included in the decision making process (Sharma and Jana, 2009). Montalvo et al, (2008) applied one of the variants of Particle Swarm Optimization (PSO) as one of these evolutionary algorithms to two ease studies: The Hanoi water distribution network and the New York City water supply tunnel system. Both cases occur frequently in the related literature and provide two standard networks for benchmarking studies, This allows them to present a detailed comparison of their new results with those previously obtained by other researchers. In the field of water resources engineering, particularly reservoir operations, Genetic Algorithm, (GA) has been proved to be computationally superior to traditional methods like linear programming, non linear programming and dynamic programming. Two types of genetic algorithms, real-coded and binary-coded were applied to the optimization of a flood control reservoir model (Chang et al,, 2005). Wardlaw and Sharif (1999) explored the potential of alternative GA formulations in application to real time reservoir operation, They found that: (a) GA has the potentiality to large-finite horizon multireservoir system problems where objective function is, complex; (6) GA needs no initial trial release policy; (e) easily applicable to nonlinear problems and Asian J. Sei. Rea., 2011 @) GA can generate several solutions that are close to the optimum. Several other studies have shown the application of GA to water resources management, (Akter and Simonovie, 2004; Chang et al, 2005; Nagesh et a2, 2008; Raju and Nagesh, 2004; Sharma and Jana, 2009; ‘Tospornsampan et al., 2005; Wardlaw and Sharif, 1999) RESERVOIR OPERATION MODELS Complex constrained multiobjective optimization problems are solved by evolutionary algorithms. This ability makes EA a good choieo for solving reservoir operation problems whieh present multiple conflicting objectives and constraints. The objective functions are always non- linear, con-convex and multi-modal functions. It is characterized with many decision variables, multiple conflicting objectives as well as considerable risk and uncertainty Reservoir can be operated to maximize hydropower generation while satisfying other uses like irrigation, downstream water requirement and flood control, The objective funetion ean be formulated for 12 months period as follows: Maximize SEH.RT,) a The constraints are as follows: 8, =8.,+1, “RT, -SP,-EY, @ S, Reservoir capacity @ $25, @) S,, SP,,RT,,H, 20 6) whore, Bis the hydropower energy, H, is the height of water above the turbine at the end of month t which has non-linear correlation with the volume of water in the reservoir, It is also dependent on the shape of the reservoir and the position of the turbine, 1 is the officiency of the turbines, Rt, is the volume of water released to the turbine in month t, S, is the reservoir storage at the end of month t, S, is the starting end of period storage, I, is the inflow to the reservoir in month t, Sp, is the water spill at month t. When a reservoir is a multipurpose reservoir, the multiple objectives set in, In this case, the reservoir may be required to maximize hydropower generation, maximize the irrigation water releases and minimize the flood risk, The constraints may be to satisfy the minimum downstream wator requirement, the minimum height required for hydropower generation and the height should be less than the maximum height to prevent spilling, Multiobjective Difforential Evolution (MODE) algorithm was presented by Reddy and Kumar (2007) with an application to a ease study in reservoir system optimization, The algorithm was applied to a multiobjective reservoir operation problem and provides many alternative Pareto optimal solutions with uniform coverage and convergence to true Pareto optimal fronts. The multiple objectives of the reservoir system are minimizing flood risk, maximizing hydropower 5 Asian J. Sei. Rea., 2011 production and minimizing irrigation deficits in a year. The model is subject to physical and technical constraints, ‘The model was formulated for ten daily operations with the objectives of maximizing hydropower production (f1) and minimizing the annual sum of square deficits of irrigation release from demands (f2) which are presented in Eq, 6-12, Objective functions: YR Ped © Minimize £, - $[anin(,1D, -1R, o Where: RH, fori =1, and vt ® Subject to the following constraints: 84,841 RR, -IR, “EVR, -OVE, Mt @ <8

You might also like