带神经网络补偿的极点配置广义最小方差自校正控制

GENERALIZED POLE PLACEMENT SELF TUNING CONTROL WITH NEURAL NETWORK COMPENSATION

首先用一个常规线性模型对被控对象进行辨识，再对线性模型辨识的余差用一个神经网络进行补偿．线性模型和神经网络共同构成对象的辨识模型，并基于这一模型提出了一种显式极点配置广义最小方差自校正控制．该方法适用于非线性对象，且具有较高精度和较快的收敛速度，具有较强的鲁棒性．

The Controlled plant is identified using normal linear model, and then the deviation identified by linear model is compensated via a neural network. The identification model is composed of a linear model and a neural network. Based on this model, an explicit generalized pole placement self tuning control algorithm with neural network compensation is proposed. This algorithm is suitable for nonlinear system, and has higher precision, faster convergent speed and stronger robustness.