优化的BP神经网络在水轮机调节系统中的应用研究

Research on the Application of Optimized BP Neural Network in the Hydraulic Turbine Governing System

针对常规PID难以取得期望的控制效果和单一神经网络在线整定存在结构复杂、学习时间过长、收敛速度慢等问题,提出了一种基于主成份优化的BP神经网络PID控制。利用主成份分析对网络的输入层单元进行降维分析,以此简化网络的结构,提高网络的收敛速度及泛化能力,进而提升智能控制系统的控制品质。基于上述理论在MATLAB/Simulink平台进行水轮机调节系统实例仿真分析,阐述并分析试验结果,说明了经主成份分析优化后的BP神经网络PID控制效果较优化前有了显著改善。

According to the control effect of conventional PID is difficult to achieve the desired and single neural network on-line setting problems of complex structure, long training time, slow convergence speed, this paper proposes a BP neural network PID control based on principal component optimization. In this scenario, an analysis is made of the network input layer element analysis dimensionality reduction by using principal components to simplify the structure of the network, improve the network convergence speed and generalization ability, and enhance the quality control of intelligent control system. Based on the above theory, a simulation analysis of turbine governing system is carried out on the MATLAB/Simulink platform, and an analysis of the test results shows that BP neural network PID control based on principal component analysis control effect optimization has been significantly improved.