Optimal control for boiler combustion system based on DHP method and generalized RBF network

The boiler combustion process of power plant is a typical process with the features of multi-input, multi-output, strong non-linearity, strong jamming and close coupling. The coupling relationship between its parameters correlated with combustion process is very anfractuous, so it is very hard to solve its optimal control problem with conventional control methods. And Adaptive Critic Designs (ACDs) is a good way to deal with the approximate optimal control problems over time in complex nonlinear systems. But most of the ACD structures were designed by BP (Back Propagation) neural network, the controllers are easy to fall into local minimum, so usually the learning efficiency is very low even fail to training. In order to speed up the learning process of the controllers, this paper try to design an optimal controller based on Dual Heuristic Programming (DHP) by Generalized Radial Basis Function Neural Network (GRBFNN) and applied it to the simulation control of the boiler combustion process. The results indicate that the designed controller is effective.