基于改进的BP网络误差修正的广义预测控制

该文利用预测误差的历史数据，基于改进的BP神经网络，对系统的建模误差进行预测。该网络采用了修正激励函数的BP算法，预测性能好，能够克服标准BP算法中Sigmoid函数的不足，加快了网络的学习速度。并将其与模型预测相结合构成广义预测控制算法，有效地克服了模型失配的影响，提高了控制的速度，同时引入控制增量增益，利用这个自由度来提高闭环的稳定鲁棒性。仿真结果表明了该算法的有效性。

Generalized Predictive Control Based on Error Correction Using Improved Neural Network

This paper presents an error prediction based on improved BP neural network to obtain the error prediction model according to the past error data.The network adopts Back Propagation Algorithm with a new activation function.The capability of forecast is well.And it can overcome the shortage of the Sigmoid function in the standard BP algorithm.At the same time it quickens the speed of the network.This prediction is combined with the model prediction to form the generalized predictive control (GPC). It affectively offsets the effect of the model mismatch, increases the pace of the network.At the same time we introduce a factor to multiply the control increment.Using this convertibility to improve the stability and robustness of the closed loop.The simulation results show the effectiveness of this control method.