The Decoupling Research Of Combustion Control System Of Circulating Fluidized Bed Boiler

The combustion control system of Circulating Fluidized Bed Boiler(CFBB) is an object that has many features.including distributed parameters, nonlinear, time-varying, big lagging and a lot of tightly coupled variables. It is difficult to control and the automation level is low, mainly based on manual control in actual operation.This dissertation focus on decoupling control based on neural network. In connection with the strong coupling characteristic of control system in the domestic CFBB.Two kind of decoupling control designs are presented including a decoupling control system composed of self-adaptive Back Propagation Neural Network PID controller(BPPID) and Radial Basic Function Neural Network(RBFNN) on-line identification network which is applied in control system of Circulating Fluidized Bed Boiler combustion control system to decouple these system. The algorithm solves the problem of unknown information of the controlled plant when using self-adaptive BPPID control algorithm individually. Another decoupling control system composed of self-adaptive BPPID and dispersive network decoupling control is presented which is applied in control system of CFBB combustion control system to decouple these system. The algorithm not only avoids the problem of poor quality control with multivariable strong coupling plant when using self-adaptive BPPID control algorithm individually. But also solves the problem of original parameters of the controller can not adapt to changes in the controlled plant.The simulation results show that the two designed system dispel the coupling between the systematic variables basically that it is improved greatly the dynamic and static responsible performance of the system. Compared with the first design and the difference between the stable time of the second design and it of the first design is small and control error of the second design is smaller than it of the first design. And two kinds of designs have a certain degree of robustness.