Feedback-compensated neural network inverse control for Superheated Steam Temperature of a 600MW supercritical boiler unit

To improve the control quality of the Superheated Steam Temperature (SST) of a supercritical boiler unit, in view of the multi-stage water-spray desuperheating system, this paper presents an error feedback compensation control scheme based on neural network (NN) inverse process models. A time-delay BP neural network is used to model the superheater system. With analysis of the boiler construction and operation characteristic, the inputs and outputs of the NN models are determined. Two NN inverse models are established, trained and validated with abundant historical operation data over wide-range load-changing condition. The trained models are then employed as NN controllers to improve the SST control effect by providing real-time supplementary signals to the original cascade PID controllers. Real-time SST feedback signal is introduced to automatically adjust the reference values of the NN controllers. The control scheme is programmed with MATLAB, which communicates real-time with a full-scope simulator of the 600MW supercritical coal-fired power unit. Comprehensive control simulation tests are carried out, which shows that the new NN inverse compensation control scheme can dramatically improve the SST control quality of the supercritical boiler.