Elman网络在Smith预测控制中的应用

Smith预测控制在实际应用中的难点在于很难得到实际系统精确的数学模型. 通过Elman网络拟合传统Smith估计器的模型误差, 并对其进行补偿. 实验结果表明, 这种基于Elman网络补偿模型的Smith预测控制充分利用了神经网络的非线性拟合能力, 只要对纯滞后环节精确建模, 就可以完全抵消纯滞后环节对控制品质及系统稳定性的不利影响. 这种方法使得Smith预测控制可以用于模型不易精确确定的系统.

Application of Elman neural network in Smith predictive control

The difficulty of applying Smith predictive control to practice lies in the fact that it is very hard to formulate a precise mathematical model of the practical system. This problem was solved here by using of Elman network to approximate the modeling error of the common Smith predictor, and to compensate it. The experimental results proved that the Smith predictive control algorithm based on Elman network compensatory model took good advantage of the nonlinear modeling capability of the neural network, and that the harm from the time delay to the performance and stability of the system could be counteracted completely Fonly if the time delay is precisely known. This method makes it possible for the Smith predictive control to be applied to the system whose mathematical model is difficult to determine precisely.