The application of BPNN based on improved PSO in main steam temperature control of supercritical unit

Abstract

In a supercritical power plant, large inertia, large delay and non-linear are the big challenges for main steam temperature control. An intelligent PID cascade control system with a BP Neural Network (BPNN) is proposed in this paper to solve this issue, which is based on the algorithm of improved Particle Swarm Optimization(PSO). In this system, the parameters of PID controller are adjusted online by BPNN, whose initial weight value is optimized by PSO algorithm, meanwhile the PSO method is also improved by Simulated Annealing (SA) algorithm which can get rid of local extreme point, accelerate the convergence speed and improve precision. Simulation result shows that the control quality and robustness of the system is significantly improved comparing with the conventional PID cascade control system.