Modeling and Controller Design of Superheated Steam Temperature System Based on SVM Combining Adaptive DMC

Abstract

Superheater steam temperature in power plant is the strong nonlinearity system. Sparse Least squares support vector networks (LSSVN) are proposed to model the superheated steam of power plant in this paper. The structure is obtained by equality constrained minimization. By combining the DMC with discount recursive partial least squares (DRPLS), a adaptive DMC control method based on discount recursive least square is presented. This method can reduce the effect of the old data, and tone up new data in order to improve the predictive capability of model. Model based on discounted-measurement has the better flexibility and adaptability. Simulation of a superheating system is taken in a 600 MW supercritical concurrent boiler. The result shows that the proposed model can adapt to the strong nonlinear super-heater steam temperature process, and the control system performance is better than conventional PID cascade control.