Research On The Coordinated Control Of Unit Plant Based On Neural Network Inverse System

As the main control system of unit plant, an effective thermal power plant unit coordinated control system is the key to raise the level of thermal automation and the economic efficiency of power plant. For boiler-turbine unit, a complex object with the characteristic of multivariable, nonlinear, large delay, and uncertainty, the traditional coordinated control system sacrifice the dynamic performance of the system to ensure steady-state decoupling performance in practice. Though the structure is relatively simple and easy to be built, the actual operating performances are not satisfactory. As the controlled object, plant unit meets with a large number of typical problems of automation field, such as non-linear, multivariable, strong coupling, large delay, and uncertainty interference, so it becomes a research focus of advanced control strategy and almost every kind of popular control strategy has been used to improve the performance of coordinated control system.According to boiler-turbine unit, a complex non-linear MIMO system, this paper uses neural network inverse control method, which has already been successfully used in robot arm control and multi-motor control, to realize the decoupling control of this system. Bell-Astrom model is introduced in this paper, and the invertibiliy of this model is proved in arithmetic. Aiming at unit plant, this MIMO nonlinear system, we combine the fitting ability of neural network and inverse system method together, and design a PID controller based on neural networkα-th order inverse system. Then we construct Bell-Astrom model in Matlab / Simulink platform, and study on the neural networkα-th order inverse system controller. Decoupling effect of the system, following performance, load disturbance in different operation points, and large-scale variable working condition performance are tested in Matlab/Simulink platform. The decoupling control target of the unit plant is achieved. In addition, the differences between two results of the relative degree of Bell-Astrom model have been discussed. The simulation experiment shows that the coordinated control system based on the method of neural network inverse control has strong robustness and good prospect of application. And this method provides a new way of thinking of the use of advanced control algorithm in unit plant coordinated control.