广义预测控制在时滞系统中的应用

Application of the Generalized Predictive Control in Time Delay System

对于时滞系统，首要解决的问题是系统的输出预测。广义预测控制利用预测模型和当前和过去的偏差预估过程未来的输出．本文推导了基于CARIMA对象模型的广义预测控制器，利用渐消记忆的递推最小二乘法在线辨识广义预测模型，利用二次型性能指标对广义预测控制器进行滚动优化，通过在温度控制中的实验表明：广义预测控制可以很好的克服时滞，系统无超调、无稳态误差。

Solving the output prediction is the chief work aimming at the object with time delay. The generalized predictive control algrithm predicts the future output using the current and past system error and predictive model. In the paper, the generalized predictive controller is deduced based on the controlled auto-regression integral moving average model. The generalized predictive model is identified by the on-line least square method of weaked memory recursion. The controller is optimized by rolling window under the quadratic form performance index： Experiments in the temperature control system show that the generalized predictive controller can overcome the side effects caused by time delay and satisfy the performance of no overshoot and no steady-state error.