The Research Of Differential Evolution Under Dynamic Environment

There are a lot of optimization problems in our daily life which are the importantelements for the life. Traditional mathematical optimization methods is the initialoptimization, but with the optimization problem becoming more complex and moredifficult, traditional methods can’t satisfy the need to solve these optimizationproblems. Researchers inspired by the biological evolution has proposed evolutionaryalgorithm to solve these complex optimization problems.Differential Evolution (DE) is proposed by Storn and Price in1995which is arelatively new evolutionary algorithm and attracts a large number of scholars’attention for its simple algorithm, less control parameters and good robustness. DE isan optimized method as all evolutionary algorithms, whose optimization process isstarted from initializing multiple initial points randomly in the search space. DE iscoded by real number, uses differential vector among populations to decide thedirection of individual disturbance, realizes the mutation of individual and reaches theaim of decreasing the function value for every individual through crossover andselection operation.The study of DE in static environment is numerous and mature, but many realproblems are changing dynamically and changed with the change of objectivefunction, environmental parameters or constraint conditions. To deal with thedynamic problems, not only the global optimum needs to be found, but also thechange of optimum needs to be tracked with time. It brings new challenge forevolutionary algorithm.According to the feature of dynamic optimization problems, an improveddifferential evolution is proposed based on the Fitness Euclidean-Distance Ratio(FER) and the change of standard differential evolution, and then it is used to testsome dynamic functions. The work in this thesis contains the following aspects:It mainly presents the basic concept of optimization problems and mathematicalmodel which are classified according to the characteristics in the first part, and theconclusion presents the feature of optimization problems currently. Evolution Algorithm is introduced at the same time, including originals, basic concepts, thebasic idea of the basic concepts and their characteristics.Differential Evolution is presented mainly in the second section. A simpledescription about Differential Evolution comes first, and then a detail information isabout standard Differential Evolution, following it is the study and research ofDifferential Evolution, including the improvement of control parameter, method ofmutation, selection of strategy and others, after that the application of DifferentialEvolution is in the last.Dynamic optimization problems is described in third part, where the feature of adynamic test functions are presented firstly, then dynamic test functions are classifiedand type of test functions in this task are listed, following it is the introduction ofdifferent methods of performance evaluation for dynamic optimization algorithm andtheir advantage and disadvantage, after that are the methods of solving dynamicoptimization algorithm and advantage and disadvantage for different methods.Dynamic test functions are firstly introduced in detail in the fourth part, thenstandard Differential Evolution is improved according to the characteristics of thesefunctions, and dynamic test functions are tested through improved DifferentialEvolution, and effectiveness and feasibility for the proposed algorithm to deal withdynamic optimization algorithm are tested and verified.