基于优化LS-SVM的电力系统暂态稳定域分析

Analysis of power system transient stability domain based on optimized LS-SVM

针对电力系统暂态稳定域无法准确获取的问题,提出将最小二乘支持向量机映射高维空间中的点到分界超平面的距离作为稳定裕度,积分持续故障轨线直到该稳定裕度为0时的时间作为临界切除时间,并且将故障模式进行分类,通过最小二乘支持向量机分类器对故障模式进行识别。针对传统遗传算法收敛速度慢的问题,采用改进的遗传算法进行参数优化,以提高其识别率及运算速度。仿真结果表明,该方法可以有效地对故障系统临界切除时间进行估计,误差在允许范围内,并能准确判断故障模式。

In order to solve the problem that power system＇s transient domain can＇t be obtained accurately, this paper uses least squares support vector machine＇s （LS-SVM＇s） distance between the mapping point of the higher-dimensional space and the hyperplanes which calculated by LS-SVM as stability margin, and uses the time when stability margin meets zero on integral continuous fault rail line as critical clearance time （CCT）. Meanwhile, it uses LS-SVM＇s classifiers to identify the fault models which are determined by fault scenarios. In view of that the ordinary genetic algorithm converges slowly, the improved genetic algorithm is used to optimize LS-SVM＇s parameters in order to improve its recognition rate and speed. The simulation results show that this method can effectively estimate fault system＇s CCT, and can accurately judge the instability modes.