基于动量因子优化学习率的BP神经网络PID参数整定算法  

BP Neural Network PID Parameter Tuning Algorithm Based on Momentum Factor Optimized Learning Rate

作　　者：胡黄水[1] 赵思远 刘清雪 王出航[3] 王婷婷 HU Huangshui;ZHAO Siyuan;LIU Qingxue;WANG Chuhang;WANG Tingting(College of Computer Science and Engineering,Changchun University of Technology,Changchun 130012,China;College of Computer Science and Engineering,Jilin University of Architecture and Technology,Changchun 130114,China;College of Computer Science and Technology,Changchun Normal University,Changchun 130032,China)

机构地区：[1]长春工业大学计算机科学与工程学院,长春130012 [2]吉林建筑科技学院计算机科学与工程学院,长春130114 [3]长春师范大学计算机科学与技术学院,长春130032

出　　处：《吉林大学学报（理学版）》2020年第6期1415-1420,共6页Journal of Jilin University:Science Edition

基　　金：吉林省教育厅“十三五”科学技术研究项目(批准号:JJKH20181166KJ);吉林省科技发展计划技术攻关项目(批准号:20190302026GX);吉林省发改委产业技术研究与开发项目(批准号:2019C054-4)。

摘　　要：针对传统BP神经网络学习过程中学习率选取过大导致振荡的问题,提出一种新的BP神经网络PID(比例-积分-微分)参数自适应整定算法.采用BP神经网络对PID参数进行自适应调节和优化,并利用动量因子优化学习率和增加动量项抑制BP神经网络训练中出现的振荡现象,以加快收敛速度.实验结果表明,该算法有效缓解了振荡现象,加快了算法的收敛速度.Aiming at the problem of oscillation caused by excessive selection of learning rate in the learning process of traditional BP neural network,we proposed a new adaptive tuning algorithm for PID(proportional-integral-differential)parameters of BP neural network.BP neural network was used to adjust and optimize PID parameters adaptively,and momentum factor was used to optimize learning rate and increase momentum term to restrain oscillation phenomenon in BP neural network training,so as to accelerate convergence speed.The experimental results show that the proposed algorithm can effectively alleviate the oscillation phenomenon and accelerate the convergence speed of the algorithm.

关 键 词：PID参数自整定 神经网络 学习率 动量因子 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]