基于多个神经网络模型进行过程变量的向前多步预测，采用Ｄａｖｉｄｏｎ最小二乘法学习网络的权值，并用二次逼近的方法求解预测控制律，构成了一种非线性自适应预测控制器。 Based on multiple neural network models, the forward and multi-step prediction of process variables is carried out. The weights of the network are studied by using Davidon least squares method. The predictive control law is solved by the quadratic approximation method to form a nonlinear adaptive predictive controller.