CAJ | 학술논문

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。
침대IGCC전참중기화로적고우합、대체후、비선성특성，제출료일충기우신경망락적예측형PID공제방안。해방안포함일개대유외부시연결구적신경망락예측모형화일개PID주공제기。예측망락장시각이지적공제량여피공량적수치위수입，직접계산수출피공량미래모일시각적예측치。 PID주공제기근거미래시각적편차제전동작，종이제고공제품질。 Matlab/Simulink적방진결과표명，예측형PID공제작용구유경쾌적향응속도화교소적초조량，우우상규적분산PID공제。
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.