【作者】 李月英； 申东日； 陈义俊； 李素杰；

【Author】 LI Yue-ying, SHEN Dong-ri, CHEN Yi-jun, LI Su-jie (College of Information and Engineering, Liaoning University of Petroleum and Chemical Technology, Fushun 113001, China.

【机构】 辽宁石油化工大学信息工程学院；

【摘要】 滚动优化是预测控制的核心,应用RBF神经网络对预测控制系统进行优化设计．为避免依据于复杂的数学运算的常规滚动优化控制策略,提出采用黄金分割优化的方法．黄金分割法不用求解偏微分方程以及大规模动态规划问题,而且能保证较快的收敛速度,计算简便．该方法可以避免神经网络预测拉制中复杂的下降梯度计算问题,而且通过设定非导数优化的搜索范围,可以解决控制量的优化问题．仿真结果验证了优化方法的有效性．更多还原

【Abstract】 Optimal operating model is core of predictive control. RBF neural networks is used for optimal operating model of predictive control. In order to avoid normal optimal operating model control strategy of complicated mathematics, the method of golden section is advanced. The method of golden section doesn’t solve the partial differential equation and large-scale dynamic programming problem, and it assures quicker convergent speed, and simple computing. By this method, on the one hand, it can avoid complicated computing of descendent grads in the predictive control of neural networks, on the other hand by setting non-differentially optimized manhunt scope, it solves the optimization of control. Finally simulation verifies that the method of golden section is efficient.更多还原