Adaptive inverse control based on parallel self-learning neural networks and its applications

This work presents an adaptive inverse control based on parallel self-learning neural networks that aims at the main steam temperature control system which has a large inertia, a long time-delay, and is time-varying in the thermal power plant. It recurs to the strong, complex, and nonlinear system identification ability of the neural networks that identifies the model and as well as the system´s inverse plant model The model of the plant is identified by NNM and its inverse model by NNC. The NNC is trained online in a parallel self-learning system. The whole controller is made up of the inverse controller NNC and a robust controller RC in order to improve the robustness of the control system. Simulation results show that this strategy has strong robustness and self-adaptive ability, and adapts to the parameters changing in the plant and puts on a good control performance as compared with the general PID controller.