本文提出一种神经网络预测偏差补偿自校正控制器 ,用线性模型的预测控制去控制非线性系统 ,其预测偏差用神经网络进行补偿 .线性模型的辨识和神经网络的学习均采用阻尼最小二乘法 .仿真结果表明 ,用这种控制器能有效地控制非线性系统 ,并具有超调小 ,鲁棒性好的特点 . In this paper, a neural network prediction error compensation self-tuning controller is proposed, which uses the linear model predictive control to control the nonlinear system, and the prediction error is compensated by the neural network. The linear model identification and the neural network learning are both damped least squares The simulation results show that this kind of controller can effectively control the nonlinear system and has the advantages of small overshoot and good robustness.