基于多代理技术的分布式模型预测长期电压稳定紧急控制  

Distributed Model Predictive Emergency Control for Long-Term Voltage Stability Based on Multi Agents

作　　者：喻振帆[1] 谢敏[1] 刘明波[1] 

机构地区：[1]华南理工大学电力学院,广东省广州市510640

出　　处：《电网技术》2012年第4期108-115,共8页Power System Technology

基　　金：国家自然科学基金项目(50907023;50777021);广东省绿色能源技术重点实验室资助项目(2008A060301002);广东省自然科学基金博士启动项目(9451064101003157)~~

摘　　要：将分布式模型预测控制方法应用于电力系统长期电压稳定紧急控制器设计,重点解决如何将大系统的在线滚动优化问题转化为若干个子系统的在线协同优化问题。在寻优过程中,利用各子系统优化目标函数的凸组合构造全局优化目标函数,通过迭代和网络通信共同求解该分布式优化问题,使系统达到纳什均衡点的收敛性得到提高。此外,根据分布式模型预测控制算法和多代理技术之间的共同性,利用Matlab和JADE在优化计算和面对代理编程方面的优势,提出并构建一种适用于分布式模型预测控制算法的多代理系统,在各子系统间实现了对滚动优化问题的并行求解。新英格兰10机39节点系统上的仿真计算验证了所提方法的有效性。The distributed model predictive control is applied to the design of emergency controller for power system long-term voltage stability, in which the key problem to be solved is how to turn the on-line receding optimization of large-scale power system into on-line collaborative optimization of several subsystems. During the searching process, a global optimized objective function is constructed by convex combination of optimized objective functions of subsystems, and this distributed optimization problem is solved by iteration and network communication to improve the convergence property by which the Nash equilibrium point is achieved. Besides, according to the intercommunity between distributed model predictive control algorithm and multi-agent technology and by use of the priority of Matlab and JADE in optimization calculation and agent-oriented programming, a multi-agent system suitable to distributed model predictive control algorithm is proposed and constructed, by which the parallel solution of receding optimization among subsystems is implemented. Simulation results of New England 10-machine 39-bus system show that the proposed method is effective.

关 键 词：长期电压稳定 紧急控制 分布式模型预测控制 多代理系统 Java代理开发框架 

分 类 号：TM712[电气工程—电力系统及自动化]