Multi-model predictive function control based on neural network and its application to the coordinated control system of power plants

The coordinated control system of boiler-turbine unit in power plants is a complicated multivariable system with nonlinear, uncertainty and strong coupling. In this paper the algorithm of multi-model predictive function based on neural network is proposed and it is applied in a 500 MW unit. Firstly, several linearized models of the unit on different working conditions are obtained with small deviation linearized method and the global predictive model is gained by the method of neural network weights. Then, the control variables are calculated by predictive function controller. Finally, the simulation results testify the validity of this control algorithm.