Design of Liquid Level Control System of Steam Generator Based on Neural Network PID Controller

Abstract

Abstract Steam generator (SG) is an important equipment of the nuclear power plant, and the stability of its liquid level affects the safe operation of the nuclear power plant. SG is a complex system with nonlinear, time-varying, nonminimum-phase, small stability margin and large time delay. In actual operation, it is difficult for classical PID control to ensure a satisfactory control performance. In this paper, the neural network methods are used to optimize the parameters of the PID controller, and a neural network controller is designed. The controller of the system consists of two components: a classical PID controller, which realizes control through a closed loop; a single-hidden-layer neural network based on the BP (back propagation) model. The neural network calculates the coefficients of the classic PID controller through matrix operations. Two weighting matrices are adjusted according to the gradient descent method to reduce the loss function and realize the training process. The control system is deployed to a SG simulation model through Simulink. The typical working conditions are simulated and investigated. The control performance is compared with that of the classical PID controller. Through analysis, it is confirmed that the neural network PID control system can meet the control requirements with fast response speed, short settling time, stable control effect under various working conditions, and strong anti-interference ability. The results prove that the neural network control has greater advantages and better application value than the classical PID controller.