In order to achieve a more efficient and accurate correction of near-field error in the semiphysical radio frequency simulation system, the precise control parameters of the three antenna elements need to be obtained. This article is based on the method of moments electromagnetic simulation, and propose corresponding improvement ideas for the problems of limited optimization accuracy and low calculation efficiency in the near-field error correction process. From the aspects of high-precision intelligent inversion algorithm and high-efficiency electromagnetic forward modeling, systematic optimization design and verification were carried out. The results prove that the control parameter filtering scheme based on PSO-GA hybrid method has better optimization efficiency and accuracy than single genetic algorithm or differential evolution algorithm, which can provide more ideal initial amplitude and phase parameters for the subsequent selection of electromagnetic simulation and forward verification. In order to solve the problem of time-consuming in the electromagnetic simulation, the multivariate vector forward model based on GA-BP network and PSO-SVM network are established, which can achieve high-precision positioning of synthetic vector target points. The neural network method has been proved to be feasible on the basis of the current sample size. The paper selects hybrid algorithms to improve the shortcomings of single algorithm and uses algorithms to optimize neural networks, thereby obtaining better optimization results and reducing the time-consuming of electromagnetic simulations, which can realize efficient correction of near-field error.

中文翻译：

---

应用于近场效应纠错的高效优化算法研究

为了在半物理射频仿真系统中实现更高效、更准确的近场误差校正，需要获取三个天线单元的精确控制参数。本文基于矩电磁仿真的方法，针对近场纠错过程中存在的优化精度有限、计算效率低的问题提出了相应的改进思路。从高精度智能反演算法和高效电磁正演建模方面，进行了系统优化设计和验证。结果证明，基于PSO-GA混合方法的控制参数滤波方案比单一遗传算法或差分进化算法具有更好的优化效率和精度，可为后续电磁仿真选型和正演验证提供更理想的初始幅相参数。为解决电磁仿真耗时问题，建立了基于GA-BP网络和PSO-SVM网络的多元矢量正演模型，实现了合成矢量目标点的高精度定位。在当前样本量的基础上，神经网络方法已被证明是可行的。本文选择混合算法来改善单一算法的缺点，并利用算法对神经网络进行优化，从而获得更好的优化结果，减少电磁仿真的耗时，可以实现对近场误差的高效修正。