论文部分内容阅读
针对一类具有分数传输滞后、参数时变的典型工业过程开环参数与广义预测控制器参数之间的内在联系 ,通过在目标函数中引入开环增益 ,并利用BP网络的非线性映射能力得到了自适应广义预测控制的一种直接算法。在该算法中 ,先用一个辨识器辨识过程开环参数 ,然后由一个已训练好的BP网络根据辨识结果和加权因子的值直接计算出控制器的参数 ,得到控制律。该方法不依赖于过程的精确模型 ,极大地简化了在线计算负担。同时 ,把区域控制的思想引入到预测控制之中 ,给出了两种区域预测控制方案。在一个二元精馏塔模型上与常规广义预测控制方案的对比仿真结果验证了文中所示方法的可行性
Aiming at the inherent relationship between a class of open-loop parameters of typical industrial processes and parameters of generalized predictive controllers with fractional transmission lag and time-varying parameters, the open-loop gain is introduced into the objective function and the nonlinear mapping capability of BP network is obtained A Direct Algorithm for Adaptive Generalized Predictive Control. In this algorithm, an identifier is used to identify the open-loop parameters of the process, and then a trained BP network is used to directly calculate the parameters of the controller according to the identification results and the weighting factors to obtain the control law. This method does not depend on the exact model of the process, greatly simplifying the online computational burden. At the same time, the idea of regional control is introduced into the predictive control, and two kinds of regional predictive control schemes are given. The simulation results of a binary distillation column model in comparison with the generalized generalized predictive control scheme verify the feasibility of the proposed method