Application Of Particle Swarm Optimization Algorithm Based On Adaptive Surrogate Model In Antenna Optimization

With the substantial improvement of computer performance and the rapid development of science and technology,the optimization design of industrial products by evolutionary algorithms has been widely concerned by scientific researchers.However,the optimization of complex industrial products is not only complicated in calculation,but also time-consuming in simulation process,which makes the algorithm less efficient.In recent years,many scholars have found that surrogate model is suitable for the optimization of complex problems such as small sample size,high dimensionality and nonlinearity.The accurate prediction of surrogate model is used to replace the optimization algorithm for the real evaluation of the product parameters.Therefore,it can effectively reduce the computational complexity,save the solution time,and achieve the purpose of optimal design.Nevertheless,there are still many problems in solving complex computing problems by combining surrogate model and evolutionary algorithm.Such as,for different types of optimization problems,how to choose the model that is suitable for solving this type of problem,which ensemble strategy is used to build the ensemble model,and how to choose the number of meta-models in the integrated model can achieve a good balance between the forecasting accuracy and the solution time of the surrogate model.All of these issues are topics that require further study.In this thesis,the particle swarm optimization(PSO)algorithm based on adaptive surrogate model is proposed to solve these questions.This method enhances the efficiency and generalization performance of PSO in solving expensive computational problems.The work of this thesis mainly includes the following three aspects:(1)This thesis proposes a method of adaptive selection surrogate model and ensemble strategy.This method can choose the appropriate model adaptively according to the complexity of the problem to be optimized and the requirement of the solution.If the selected model is an ensemble model,then all meta-models are ranked according to cross-validation root mean square error(RMSE),and the top few ‘elite models' are selected to build the ensemble surrogate model.Meanwhile,inspired by the idea of self-adaptive differential evolution(SADE),an optimal strategy is selected adaptively from the five strategies in optimal weighted surrogate(OWS).The standard test function and an engineering example are used to test the proposed method.The experimental results show that the surrogate model based on adaptive selection strategy has better generalization performance and prediction accuracy.(2)This thesis proposes a particle swarm optimization algorithm based on adaptive surrogate model(ASMPSO).For each generation of the optimization process,the algorithm selects the model with the smallest root mean square error(RMSE)of contemporary cross-validation to update the model of the last generation,and updates all surrogate models by adding each generation's true optimal solution to the database.Finally,the surrogate model is used to predict the particle fitness value instead of the accurate calculation of the particle fitness value(ie,reducing the number of simulations in the optimization process),thereby reducing the computational complexity and saving optimization design time.The experimental results show that the ASMPSO can not only achieve a certain precision,but also can get the global optimal solution faster.(3)The effectiveness of the proposed algorithm is verified by the practical industrial problem of time-consuming antenna optimization design.Through the data interaction of high frequency simulation software(HFSS)and Matlab,the process of antenna modeling,simulation analysis,parameter prediction and iterative optimization are completed automatically.The experimental results show that although the algorithm proposed in this thesis can not achieve the solution precision of antenna structure parameter optimization only by using PSO algorithm,the algorithm needs shorter time to complete the optimization of antenna parameters,and the optimization efficiency is higher.In this thesis,antenna optimization design as a research background,a method of adaptive selection surrogate model and ensemble strategy is proposed.This method is combined with PSO algorithm to improve the algorithm's optimization efficiency and generalization ability for complex computational problems.Ultimately,we are expected that this method will provide technical support for the development and performance optimization of complex products.