Research Of Antenna Optimization Technology Based On Quasi-Opposition Grey Wolf Optimization Algorithm

In the field of antenna design and optimization,with the continuous emergence of antennas with sophisticated structures,the traditional comprehensive design method has shortcomings such as high-time cost,limited application range of approximate model formulas,and the selection of optimization parameters heavily dependent on the design experience of engineers,which gradually gets into trouble.In recent years,the emergence of intelligent algorithms has effectively alleviated the above problems,making the intelligent design of antennas possible.Grey Wolf Optimization(GWO),as a new type of swarm intelligent optimization algorithm,has better performance than other algorithms.Therefore,in this paper,we will carry out research on GWO algorithm for optimizing antenna designs.Although the GWO algorithm has many advantages,it also has many limitations in some engineering problems.On the one hand,when the GWO algorithm solves complex problems,its optimization accuracy will not be guaranteed,the convergence speed in the later iterations is relatively slow,and it is easy to fall into the local optimal solution.On the other hand,the fitness values of individuals in the gray wolf population need to be computed through the electromagnetic simulation software such as HFSS.The time cost of performance optimization will increase dramatically with the complexity of the antenna structure,even to an unacceptable level.This paper focuses on the two problems above in the GWO algorithm.The specific research contents are following:(1)Aiming at the shortcomings of the standard GWO algorithm,which contains easily falling into the local optimal solution,the low solution accuracy of complex problems,and the slow convergence speed,this paper adopts a quasi-opposition population initialization strategy,a nonlinear reduction of the convergence factor strategy and the update strategy of individual positions according to weights.The strategies improve the corresponding links of the standard GWO algorithm,and propose the Quasi-Opposition Grey Wolf Optimization(QOGWO)algorithm.Its comparison with other optimization algorithms in six unimodal benchmark functions shows that the QOGWO algorithm has higher solution accuracy and stronger robustness.The optimization simulation of a broadband dual-polarized magnetoelectric dipole antenna verifies the efficiency of the proposed algorithm for antenna design and performance optimization.(2)In order to reduce the time costed by calling the electromagnetic simulation software to obtain the individual fitness value,this paper adopts the Gaussian Process Regression(GPR)model to calculate the fitness value of each individual in the population.Because different kernel functions will affect the fitting ability and prediction ability of GPR model to different degrees,and the conjugate gradient method(CG)commonly used for hyper parameter optimization in kernel functions has excessive dependence on the initial value and is easy to fall into local optimum.In this paper,the CG algorithm is replaced by the QOGWO algorithm and the QOGWO-GPR prediction model is developed.Taking a broadband unidirectional antenna as an example,this paper uses the QOGWO algorithm to optimize hyper parameters among four GPR models with different kernel functions,and selects the model with the best prediction ability as the QOGWO-GPR prediction model of the antenna.The simulation results show that,compared with the CG algorithm,the QOGWO-GPR model has higher prediction accuracy for the performance parameters of the broadband unidirectional antenna.(3)This paper combines the above two techniques and develops the QOGWO-GPR hybrid model algorithm.The algorithm not only optimizes the antenna with better performance,but also greatly reduces the design optimization time.The efficiency of the algorithm is validated by a broadband dual-polarized magnetoelectric dipole antenna and a miniaturized stepped ultra-wideband printed monopole antenna.