采用BP神经网络求解CUEP的实用计算  

Practical calculation of CUEP based on BP neural network

作　　者：刘彦呈[1] 王川[1] 魏一[1] 

机构地区：[1]大连海事大学轮机工程学院,辽宁大连116026

出　　处：《电力系统保护与控制》2011年第22期144-149,共6页Power System Protection and Control

基　　金：辽宁省自然科学基金(20092148)~~

摘　　要：主导不稳定平衡点(Controlling Unstable Equilibrium Point,CUEP)的计算在电力系统暂态稳定分析直接法中具有重要意义。提出一种计算CUEP的新的实用算法:由系统受扰后各切除时刻状态及达到稳态后系统状态可分别作为输入和输出样本,训练BP神经网络;基于Thévenin定理及转矩-滑差曲线以及转子暂态特性,估算出等值系统中负荷母线发生三相短路故障下的极限切除时间,并以此作为大扰动故障临界切除依据,输入训练好的人工神经网络即可输出CUEP。数值仿真采用IEEE-39节点系统来构造算例,以C#.NET2010为编程语言,与时域仿真法对比证明了该方法的有效性及较高的精度。该方法无需牛顿法迭代计算、节省机时、误差小,具有一定的实用价值。The calculation of Controlling Unstable Equilibrium Point （CUEP） is very significant in the direct analysis method of electric power system transient stability. This paper presents a novel practical algorithm to calculate CUEP. Back Propagation （BP） neural network is trained by using variables at different fault cut times as input samples and variables at stable state as output samples separately. Then, according to Th6venin Theorem, torque-slip curve and rotor transient characteristic, the algorithm calculates CUEP by inputting variables at critical cut time of weak load bus into the trained BP neural network. Numerical simulation uses IEEE-39 system as examples, and C#.NET 2010 as programming language. Simulation results demonstrate that the proposed method is effective and precise by comparing it with time domain simulation. The method is practical because of its low time cost and error, and it does not need to calculate by iteration.

关 键 词：电压稳定 暂态稳定 大扰动 主导不稳定平衡点 BP神经网络 Thévenin定理 暂态电压稳定指标 

分 类 号：TM712[电气工程—电力系统及自动化]