Design And Research Of Multi-objective Evolutionary Algorithms For EED Problem

Environmental Economic Dispatch(EED)has attracted wide attention,because it can take into account both environmental protection and economic benefits.EED problem is a non-linear and non-convex multi-objective optimization problem.Early methods are difficult to trade off among multiple objectives.These methods use constraints method or weight coefficient method to transform multiple objectives into single objective problems.In recent years,evolutionary algorithms have performed well in solving multi-objective problems.Research on EED problems has gradually turned to using evolutionary algorithms to obtain faster and better dispatching plans.Therefore,this paper designs two learning strategies.In order to improve the uniformity,diversity and convergence of Pareto front,we put these two strategies into the multi-objective algorithm,and propose three multi-objective optimization algorithms to solve EED problems.The main research work is as follows:1.In order to improve the uniformity and diversity of Pareto front,two specific learning strategies are proposed.One is the leader updating strategy.In this strategy,a concept of maximum distance is designed to measure the sparsity of solution in external archive set;another is the leader wandering strategy.In this strategy,a concept of sparse direction is designed to guide leaders to search for sparse directions around themselves.2.Combining these two strategies with backtracking search algorithm(BSA),this paper proposes a multi-objective learning backtracking search algorithm(MOLBSA).In MOLBSA,the mutation operator which learns from the leader is added.The original mutation operator is controlled by the history population(oldP),while the improved mutation operator is guided by the leader and oldP.The experimental results of MOLBSA on IEEE30-bus 6 unit test system are compared with those of other algorithms,which verifies the excellent performance of MOLBSA.3.Combining these two strategies with collective decision optimization algorithm(CDOA),this paper proposes a multi-objective collective decision optimization algorithm(MOCDOA).In MOCDOA,a geometric center update strategy is designed to increase the probability of searching the two endpoints of the current Pareto front.In order to verify the performance of MOCDOA,experiments on IEEE30-bus 6 unit show that MOCDOA has better performance than other algorithms.4.Combining these two strategies with a new grey prediction optimization algorithm(GPEA),this paper proposes a multi-objective grey prediction optimization algorithm(MOGPEA).The algorithm uses three successive populations(calledinformation evolutionary chain)to form time series and predicts the next population by GM(1,1).The application of MOGPEA in IEEE30-bus 6 unit test system proves the superiority of MOGPEA in uniformity,diversity and convergence.In order to further verify the performance of the combination of the two strategies and algorithms,the results of MOLBSA,MOCDOA and MOGPEA in IEEE30-bus 6unit test system are compared.The results show that MOGPEA is superior to other multi-objective algorithms in uniformity and convergence.MOCDOA is superior to the other two multi-objective algorithms in diversity,and the three multi-objective algorithms have potential to solve other multi-objective optimization problems in power system.