多目标执行依赖启发式动态规划励磁控制  

Application of Multi-objective Action-dependent Heuristic Dynamic Programming in Excitation Control System

作　　者：林小峰[1] 谢树平[1] 

机构地区：[1]广西大学电气工程学院,南宁530004

出　　处：《电力系统及其自动化学报》2012年第3期28-34,共7页Proceedings of the CSU-EPSA

基　　金：国家自然科学基金项目(60964002);国家自然科学基金重点项目(61034002)

摘　　要：针对工程中大量出现的多目标最优控制问题,提出一种多目标自适应动态规划方法。通过向量2-范数形式将向量型性能指标函数处理成标量,结合自适应评价设计方法和人工神经网络推导出适合实时控制的参数递推更新、性能递进优化的多目标执行依赖启发式动态规划方法。该方法无需对象精确数学模型,并且通过递推式优化的方法来求解问题,能够实时根据模型变化优化控制律。在电力系统励磁控制系统仿真中验证了该算法的有效性及控制器的控制性能。A multi-object adaptive dynamic programming method is proposed in this paper to solve the optimal control with multi-objective performance indices. The multi-objective performance indices are commonly used in engineering fields. The classical multi-objective optimal control problem requires accurate model of the plant, which is difficult to solve. The control law is sensitive to the model parameters. In the paper, vector- valued performance indices are transformed into scalar by vector 2-norm. Multi-object adaptive dynamic pro- gramming is derived from adaptive critic design and artificial neural network, which fits to use in real-time con- trol. The parameter is updated iteratively, and the performance indices are optimal as time goes on. The algo- rithm does not require system dynamics. It solves the problem step by step and optimizes the multi-objective performance indices in real time. The simulation results show the effectiveness of the method in the excitation control system of power system.

关 键 词：自适应动态规划 同步发电机 励磁系统 多目标 神经网络 

分 类 号：TM76[电气工程—电力系统及自动化]