摘要
为解决传统比例?积分?微分(proportional integral differential,PID)控制应用于静止无功补偿器(static var compensator,SVC)非线性控制系统存在的不足,提出了一种基于自适应动态规划(adaptive dynamic programming,ADP)的SVC自适应优化控制策略。采用小波神经网络和BP神经网络分别设计执行依赖启发式动态规划(action dependent heuristic dynamic programming,ADHDP)的执行网络和评价网络,以增强ADHDP对性能指标函数的逼近能力和控制律优化能力,然后用其设计了SVC控制系统的电压调节器。在Matlab/Simulink仿真平台对所提出的ADHDP控制方法进行了仿真,并与执行网络、评价网络均采用BP神经网络设计的经典ADHDP控制方法的控制效果进行了对比,验证了基于ADHDP的SVC电压优化控制方法的可行性和有效性。相比之下,所提出的ADHDP控制方法具有更好的电压稳定和控制效果,控制系统具有较快的响应速度、较好的动态和静态稳定性和较强的自适应能力。
In order to solve problem of control lack of static var compensator（SVC） nonlinear voltage control system based on conventional proportional integral differential（PID） controller, a method based on adaptive dynamic programming（ADP） to optimize control of SVC voltage control system adaptively was proposed. Wavelet neural network and BP neural network were used to design action network and critic network of action dependent heuristic dynamic programming（ADHDP） to enhance its approximation ability of performance index function and optimization ability of control law in this paper. Then, the proposed method was used to design voltage regulator for SVC voltage control system. Simulation of the proposed ADHDP control method was carried out in Matlab/Simulink platform. By comparing control result of two methods, i.e. classical ADHDP control method and the proposed ADHDP control method, simulation results verify effectiveness and feasibility of the proposed method. Action network and critic network of classic ADHDP was designed using BP neural network. The proposed ADHDP control method for SVC has better voltage stability control performance. The control system has faster response speed, better static and dynamic stability, and stronger adaptive performance.
出处
《电网技术》
EI
CSCD
北大核心
2017年第3期895-900,共6页
Power System Technology
基金
国家自然科学基金项目(61563006)
广西科技攻关项目(桂科攻1598008-2)
广西自然科学基金项目(2013GXNSFCA019020)~~
关键词
静止无功补偿器
自适应动态规划
电压稳定控制
小波神经网络
static var compensator
adaptive dynamic programming
voltage stability control
wavelet neural network
