学习方法整定电力系统非线性控制器参数控制阀
2022-08-03 11:25:56 驰援五金网
学习方法整定电力系统非线性控制器参数
学习方法整定电力系统非线性控制器参数 2011： 中图分类号：TP181；TM71 文献标识码：A文章编号：0258-8013 (2000) 04-0001-05THE ADJUSTMENT OF THE PARAMETERS OF POWER SYSTEMNON-LINEAR CONTROLLER BY LEARNING ALGRITHMS ZHANG Cai ZHOU Xiao-xin(Electric Power Research Institute Chian，Beijing 100085,China）JIANG Lin WU Qing-hua（Department of Electrical Engineering and Electronics,The University of Liverpool, Liverpool, L69 3GJ, UK)ABSTRACT： In this paper, the iterative and increasing learning algorithms are introduced to the adjustment of the parameters of TCSC nonlinear controller in power systems. According to the characteristics of practical power systems, such as the strong nonlinearity, dynamic process and etc., the learning algorithms are improved: the off-line iterative learning is firstly modified into the on-line equal periodical learning, and then into the on-line unequal periodical learning. By changing the objective function defined in a continuous set into a function in a point, the increasing learning can be on-line. In order to make the learning algorithms process satisfied effectiveness and can learn under large disturbances of the system, the nonlinearity of the system is used in the learning algorithms. The improved learning algorithms are efficient, simple and practical, and provide new methods for the adjustment of the parameters of the controller. The digital simulation shows that under the same conditions, the performance of the unequal periodical learning algorithm is better than that of equal periodical learning , the performance of the increasing learning algorithm is better than that of the unequal periodical learning. The parameters of the controller found by the learning algorithms make the controller possess better dynamical performance,strong adaptability and robustness. KEY WORDS：on-line unequal periodical iterative learning algorithm; increasing learning algerithen; TCSC nonlinear controller; the adjustment of parameters; robustness1 引言 文［1～3］所设计的输电线路可控串补 (TCSC) 及其协调非线性控制器较少依赖或不依赖于被控系统的知识，但却具有良好的控制性能和简单的结构。调整好非线性控制器的参数是使控制器有效、可靠地实现其各项性能指标的前提。学习方法可直接设计控制器，也可作为其它设计方法的辅助工具如确定优化控制器参数等。为克服控制器参数整定的困难，本文将学习方法引入电力系统非线性控制器的参数整定，使控制器具有更好的控制特性，为控制器参数的整定提供了新方法。数字仿真结果表明：在几种运行及故障方式下, 用学习方法整定的控制器参数具有较好的动态性能、较强的智能性、鲁棒性、容错性等，控制器采用学习参数将有更好的品质特性。2 学习算法 G.N.萨里迪斯［4］将学习定义为：一个系统，如果能对一个过程或其环境的未知特征所固有的信息进行学习，并将得到的经验用于进一步的估计、分类、决策或控制，从而使系统的性能得到改善，那么就称该系统为学习系统。 具有学习功能的算法称为学习算法，学习算法有以下几种类型：(1) 基于模式识别的学习算法［4］： 针对先验知识不完全的对象和环境，将改变量进行分类，确定这种分类的决策，根据不同的决策切换对改变量的作用进行切换选择， 通过对对象性能估计来引导学习过程，从而使系统的性能得到改善；(2) 基于迭代的学习算法［4、5］：针对一类特定的系统但又不依赖系统的精确模型，通过反复训练的方式进行自学习，使系统逐步逼近期望输出；(3) 联结主义学习系统［4、6、7］：具有网络结构的形式（例如人工神经网络），由节点以及节点间的联结弧组成。每个节点可以看作一个简单的处理单元，其中包含若干可调参数。3 学习算法原理3.1 迭代学习法 设系统在一时间段［0，T］内以相同起始条件x(0)＝x0重复运行，yd(t)为t∈［0，T］事先给定的期望输出，通过以下重复迭代学习算法得到的控制策略，可使系统在这一时间段的实际输出逼近期望输出［5］：εk(t)＝yd(t)－yk(t) (1)uk＋1(t)＝uk(t)＋Dc(t)εk(t) (2)式中 t∈［0，T］；uk(t)为第k次训练时的输入；yk(t)为第k次训练时的输出；εk(t)为第k