Research On Evolutionary Constrained Optimization And Evolutionary Dynamic Optimization Algorithms

Optimization problems widely exist in real-world applications.In this dissertation,we focus on the constraint handling method,time-linkage feature,dynamic optimiza-tion method,and two real-world applications(i.e.,the short-term hydrothermal schedul-ing problem and the dynamic optimal power flow problem).Evolutionary Algorithms(EAs),a class of population-based bio-inspired stochastic algorithms,are adopted to solve the above mentioned problems.EAs have been widely used because they have advantages in solving complex problems which are discontinuous,multi-modal,non-differentiable and they have good global search ability and adaptivity.The contributions are given as follows:First,for constrained optimization problems,the species-based Repair Strategy(SRS)is proposed to select representative infeasible individuals instead of the random selection for gradient-based repair.Although gradient-based repair is effective in han-dling constraints,it is known that it would be time-consuming if all infeasible individu-als are repaired.Therefore,so far the infeasible individuals to be repaired are randomly selected from the population and the strategy of choosing individuals to be repaired has not been studied yet.The basic idea of the proposed SRS strategy is as follows.First,divide the population into species using a clustering algorithm.Then,for every species,small number of infeasible individuals with best objective values are repaired using the gradient-based repair.And the number of individuals to be repaired is decided by the proportion of the feasible solutions of each species.Because diverse individuals are selected for gradient-based repair,redundant repair can be reduced,and the diversity of post-repair individuals are encouraged,as a result,the probability of falling into local optima can be reduced.The experimental results show that the improved algorithm out-performs the original algorithm in most benchmarks.Meanwhile the number of repaired individuals is reduced markedly.Second,for dynamic time-linkage optimization problems(DTPs),we concentrate on how to deal with the situation where the predictor is unreliable.Dynamic Time-linkage Optimization Problems(DTPs)are a special class of Dynamic Optimization Problems(DOPs)with the feature of time-linkage.Time-linkage means that the deci-sions taken now could influence the problem states in future.Although DTPs are com-mon in practice,attention from the field of evolutionary optimization is little.To date,the prediction method is the major approach to solve DTPs in the field of evolutionary optimization.However,in existing studies,the method of how to deal with the situation where the prediction is unreliable has not been studied yet for the complete Black-Box Optimization(BBO)case.Therefore,we propose a inversion-based method to measure the prediction accuracy.Meanwhile,a stochastic-ranking selection scheme based on the prediction accuracy is designed.The experimental results on existing and the newly proposed tested problems show that the performance of our algorithm is competitive.Third,for dynamic constrained optimization problems(DCOPs),we propose to apply the idea of multi-modal optimization to design the strategies of locating and track-ing feasible regions,i.e.,simultaneously locating and tracking multiple feasible regions.Based on this idea,an ensemble of three locating and tracking feasible regions strategies is proposed to handle different types of dynamics in constraints.Meanwhile,we propose an adaptive local search strategy and a species-based change detection strategy.More-over,two benchmarks are designed to simulate the DCOPs with dynamic constraints.The first benchmark,including two variants of G24(called G24v and G24w),could control the size of feasible regions.The second benchmark,named Moving Feasible Regions Benchmark(MFRB),is highly configurable.The global optimum of MFR-B is calculated mathematically for experimental comparisons.Experimental results on G24,G24v,G24w and MFRB show that the proposed algorithms performs significantly better than the compared algorithms.Moreover,our algorithms have advantage in lo-cating multiple disconnected feasible regions,including the situation when the regions are very small.Fourth,for short-term hydrothermal scheduling problems,we formulate the un-certain single-objective short-term hydrothermal scheduling problem for the first time,by modifying the classical problem model to take into account the uncertainties of load demands and transmission losses.And then we propose a novel Hybrid Particle Swar-m Optimizer(HPSO)to solve this uncertain problem.Moreover,a special encoding scheme is introduced to handle two kinds of constraints.The results of HPSO is com-pared with the typical lbest PSO,gbest PSO,and two typical gradient-based algorithms including interior point algorithm and series quadratic programming(SQP).The results show that HPSO performs the best for all the tested cases.Fifth,for dynamic optimal power flow problems,we concentrate on the problems with double-sided uncertainty.Most of existing works on dynamic optimal flow prob-lems only focus on the problems with uncertain loads.However,because of the increas-ing use of renewable energy sources and the randomness feature of renewable power generation,it is important to consider the uncertainty of renewable generation.So far,only small number of works focus on the dynamic optimal power flow problems under double-sided uncertainty.And to the best of our knowledge,there is no evolutionary algorithm specially designed for this type of problems.We propose a species-based dif-ferential evolutionary(DE)algorithm,including two versions.The first version adopts the nearest-better clustering(NPC)to divide the population into species.The parameter of NBC is not problem-sensitive.The second version uses a improved NBC with less time complexity to divide the population into species.An improved memory strategy is also proposed.