基于BP神经网络辨识的预测滤波PID控制  

Predictive Filtering PID Control Based on BP Neural Network Identiftication

作　　者：侯小秋 李丽华 Hou Xiaoqiu;Li Lihua(School of Electronics and Controlling Engineering,Heilongjiang University of Science and Technology,Harbin City,Heilongjiang Province 150022)

机构地区：[1]黑龙江科技大学电气与控制工程学院,黑龙江哈尔滨150022

出　　处：《黄河科技学院学报》2023年第5期26-31,共6页Journal of Huanghe S&T College

摘　　要：针对复杂非线性系统采用常规PID控制性能不好的问题,通过BP神经网络构建系统输出量的辨识器,在目标函数中加入优化参数的增量约束项,提出可克服算法病态的牛顿法,并用其在线学习BP神经网络的连接权,根据增量式预测滤波PID控制,推导了PID控制参数应满足的线性不等式约束条件,将PID控制参数的优化归结为线性不等式约束的最优化问题。综上研究提出基于BP神经网络辨识的在线优化参数预测滤波PID控制。仿真研究说明,因算法具有在线优化PID控制参数和预测控制性能,故提出的智能PID控制算法与常规PID控制相比,具有更优的性能。To address the problem of poor performance of complex nonlinear system using conventional PID control,this paper constructs the discriminator of the system output quantity by BP neural network structures identifier of system output variable,adds the incremental constraints set into optimization parameter of object function,proposes the Newtons method with overcoming algorithmic-ill was developed,and uses it’s on line learning connection weight of BP neural network,derives the linear inequality constraints that its PID control parameters should satisfy according to the incremental predictive filtering PID control,and reduces the optimization of its PID control parameters to the optimization problem of linear inequality constraints.Based on BP neural network,an on line optimization parameter predictive filtering PID control algorithm was obtained.Simulation results indicate that the control performance of the algorithm is better than that of the conventional PID control due to the algorithm with functions of on line optimization parameter and predictive-control.

关 键 词：BP神经网络 复杂非线性系统 预测滤波PID控制 牛顿法 PID控制参数优化 在线学习算法 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]