Differential Evolution Algorithm For Solving Stochastic Expected Value Models

There are many uncertainties in the fields of operation research, informationscience, computer science, military science, reliability engineering and space technology.Thus, setting up and solving of the optimization model under uncertainties is greatacademic value and extensive application prospect. However, the most common one isstochastic expected value model which optimizes some objective function of themathematical expectation subject to some expected constraints. But the objectivefunction and constraint function of stochastic expected value model are alsomathematical expectation, and the acquisition of objective function is stochasticsimulation, it may require large numbers of computations for solving such models. Sothe intelligent optimization algorithm that using fitness estimation strategy solvesstochastic expected value models was appeared. In this thesis, the research involvesoptimization algorithm for solving stochastic expected value model. The main results asfollows:1. In view of stochastic expected value model, the thesis proposes a DifferentialEvolution algorithm based on the reliability and similarity estimation. In the mutationand crossover operation of Differential Evolution algorithm, the method using reliabilityand similarity constructs predictive value of fitness function by weighting method,thereby reducing the computational of the objective function and improving thecalculation efficiency. Finally, the example shows the feasibility and effectiveness of thealgorithm.2. For stochastic expected value model, the paper proposes another DifferentialEvolution algorithm based on fitness estimation. In the mutation operation of thealgorithm, compression factor of Differential Evolution algorithm was used to constructpredictive value of fitness function, while the distance weighted fitness estimationstrategy was employed to construct predictive value of fitness function in crossoveroperation. And the examples show the feasibility and effectiveness of the algorithm.