神经网络预测PID控制在气化炉中的应用

Application of Neural Network Predictive PID Control in Gasifier

针对IGCC电站中气化炉的高耦合、大滞后、非线性特性,提出了一种基于神经网络的预测型PID控制方案。该方案包含一个带有外部时延结构的神经网络预测模型和一个PID主控制器。预测网络将时刻已知的控制量与被控量的数值为输入,直接计算输出被控量未来某一时刻的预测值。PID主控制器根据未来时刻的偏差提前动作,从而提高控制品质。Matlab/Simulink的仿真结果表明,预测型PID控制作用具有更快的响应速度和较小的超调量,优于常规的分散PID控制。

In accordance with the features of gasifier in IGCC power station, e.g. , large time lag, closed coupling and non-linearity, the predictive PID control strategy based on neural network is proposed. The control scheme contains the neural network predictive model with external time delay structure, and a PID main controller. With the known values of manipulating and controlled variables at previous moment as the input of network, the predictive value of controlled output at certain future time is calculated directly by the predictive network. The PID main controller acts in advance in accordance with the deviation of future time, thus the control quality is enhanced. The results of simulation based on Matlab/Simulink show that this predictive PID method possesses faster response speed and lower overshoot; it is better than conventional distributed PID control.