基于种群分级遗传算法的受限广义预测控制

Constrained Generalized Predictive Control based on GA of Classified Population

存在约束的控制过程限制了传统GPC方法在工业领域中的广泛应用。本文利用遗传算法来解决受限GPC算法的优化问题。当控制作用突破受限条件时,启用遗传算法来处理带约束的非线性优化问题,并以此作为滚动优化策略,求得最优控制律。遗传算法采用种群分级制度,充分利用所获信息对高级种群进行初始化,大大提高了算法的局部搜索能力。同时低级种群的引入,可以更好地均衡算法的全局探索能力,提高算法寻优的精度。最后对一个二阶纯滞后系统进行仿真控制,结果表明种群分级的遗传算法能够很好地处理存在约束的广义预测控制问题。

The constrained control process limits the widespread use of traditional GPC. In this article, the constrained GPC problem is solved by using GA. Once the control action breaks the constrained condition, GA is adopted to handle the constrained nonlinear optimization problem, and use it as the rolling optimization strategy to obtain the optimal control law. This article classifies the population and makes full use of the obtained information to initialize the senior population, it can improve the local searching ability and enhance the search capability of genetic algorithm. Besides, the low population can balance the global exploration ability. Finally, select the second-order time delay system commonly used in industry, it is simulated in MATLAB, the simulation result shows that the GA of classified population can handle the constrained GPC problem well.