基于麻雀搜索算法的BP神经网络优化技术  

Optimization technology of BP neural network based on sparrow search algorithm

作　　者：孙全 孙渊(指导)[1] SUN Quan;SUN Yuan(School of Mechanical Engineering,Shanghai Dianji University,Shanghai 201306,China)

机构地区：[1]上海电机学院机械学院,上海201306

出　　处：《上海电机学院学报》2022年第1期12-16,共5页Journal of Shanghai Dianji University

基　　金：上海市高峰高原学科资助项目(A1-5701-18-007-03);上海电机学院研究生科创资助项目(B1-0225-21-011-07)。

摘　　要：采用麻雀搜索算法(SSA)对BP神经网络的初始权值和阈值进行优化,加快BP神经网络对PID控制器参数的整定,最终获得最优参数。将正弦余弦算法和Levy飞行引入SSA的迭代过程中,扩大前期搜索范围,提高后期搜索精度。在Matlab仿真环境下,对比BP神经网络、粒子群优化的BP神经网络以及SSA优化的BP神经网络在优化PID参数方面的效果。结果表明:粒子群优化的BP神经网络PID控制器,在上升、稳定时间方面比BP神经网络PID控制器缩短了33.3%和51.9%,同时超调减小了5.81%;在稳定时间方面SSA优化的BP神经网络PID控制器比粒子群优化的BP神经网络PID控制器缩短了7.4%,同时消除了超调。Sparrow search algorithm(SSA) is used to optimize the initial weights and thresholds of the BP neural network in order to speed up the parameters tuning of PID controller by the BP neural network and finally obtain the optimal parameters. The sine-cosine algorithm and Levy flight are introduced into the iterative process of the SSA to expand the early-stage search range and improve the late-stage search accuracy. In the Matlab simulation environment, the optimization effects of the PID parameters are compared on BP neural network, BP neural network optimized by particle swarm optimization(PSO-BP) and BP neural network optimized by sparrow search optimization(SSA-BP).The results show that PSO-BP based PID controller can shorten the rise and stability time by 33.3%and 51.9%, and reduce the overshoot by 5.81% compared with the BP based PID controller. The SSABP based PID controller can further reduce the stability time by 7.4% compared with the PSO-BP based PID controller, and the overshoot is eliminated.

关 键 词：麻雀搜索算法(SSA) 正弦余弦算法 Levy飞行 反向传播(BP)神经网络 比例-积分-微分(PID)控制器 超调 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]