Research And Application Of Interval Optimization Algorithm Based On Evolutionary Strategy

This thesis mainly studies a deterministic global optimization algorithm which called interval optimization algorithm.Compared with the point optimization algorithm,the interval can represent the uncertainty of the data,and the result of the strict operation in the mathematical sense can be obtained.The interval optimization algorithm can provides a wider feasible domain in the complex industry,and it can effectively reduce the impact of disturbance and noise.So the interval optimization algorithm is suitable for solving the optimization problems in complex industrial process control.The traditional interval optimization algorithm based on the branch and bound method,which makes the algorithm have the problem of correlation and dimension disaster.The relevance problem is due to the characteristics of the interval operation and the non-linear objective function.The relevance problem can only be minimized and never be completely avoided.Dimension of disaster problems need to construct a reasonable acceleration tool,which can effectively guide the algorithm branch and remove the unreliable interval.In this thesis,we focus on the evolutionary strategy as an acceleration tool to compare with interval optimization algorithm.In addition,the proposed new algorithms are applied to the parameter estimation of the cell growth model,and a set of parameters with high fitting degree is obtained.The main work of this thesis is as follows:(1)A brief introduction to the optimization algorithm,focusing on the interval optimization algorithm and summarizing the current research on interval algorithm at home and abroad;Introduce the basic knowledge of interval analysis theory such as interval concept,interval operation rule,interval expansion and so on.Explain and analyze the existing interval optimization methods that are the traditional interval algorithm and interval genetic algorithm.(2)Propose the interval optimization algorithm-I based on the evolutionary strategy.This algorithm select a point in each sub-interval to branch or pruning the sub-interval branch or pruning by the evolutionary strategy of these points.A reliable upper bound facilitates pruning and reduces the computational complexity of the algorithm.At the same time,a new interval rule is proposed to further improve the efficiency of algorithm search.(3)Propose the interval optimization algorithm-II based on the evolutionary strategy.This algorithm select multiple points in each subinterval to represent the subinterval information,and make a few evolution strategies for all points.The purpose is to make these points close in the optimal solution.Then separate the reliable sub-intervals with most points and prune the subintervals that do not contain any point.(4)Doing the numerical simulation to compare the above mentioned algorithms with several typical test functions.The proposed two algorithms are applied to the unknown but bounded parameter estimation problem.Compared with the conventional interval parameter identification algorithm,the parameters obtained by the proposed algorithms are more fit degrees than those of the original model.The identification of the transfer function model also satisfies the uncertainty requirements.