Superheated steam temperature predictive optimal control based on external time-delay BP neural network and a simpler PSO algorithm

Power industry is facing a rapid development toward larger-capacity supercritical and ultra-supercritical coal-fired power generating units. The big inertia, large time delay and nonlinear characteristics of the superheater system for a supercritical boiler are becoming more and more obvious. The conventional cascaded PID control scheme is often difficult to obtain satisfactory steam temperature control effect over wide-range loading conditions. In this paper, an intelligent predictive optimal controller based on neural network (NN) modeling and Particle Swarm Optimization (PSO) is presented for superheated steam temperature control. Aiming at the known drawbacks of the NN predictive optimal controller, a time-delay BP neural network is used to establish the nonlinear dynamic model of the superheater system to improve the model´s real-time prediction accuracy and generalization ability. A new simplified high-efficiency PSO algorithm discarding the concept of velocity is adopted to search the optimal controls with faster convergence rate to meet the real-time control demand. An elastic search space is updated dynamically based on the real-time steam temperature control error to prevent oscillation and to enhance stability. By taking a full-scope simulator of a 600MW supercritical power unit as the object investigated, the neural network model of the superheater system is built and trained with historical wide-range operating data. The new control scheme is programmed with MATLAB software and it communicates two-way with the simulator. Extensive control simulation tests are made. It is shown the proposed intelligent predictive optimal control scheme can greatly improve the superheated steam temperature control with good application prospect.