在线识别同调机群的优化支持向量机算法  

On-line Identification of Coherent Generator Using Optimized LS-SVM

作　　者：文俊[1] 刘天琪[1] 李兴源[1] 任景[1] 

机构地区：[1]四川大学电气信息学院,四川省成都市610065

出　　处：《中国电机工程学报》2008年第25期80-85,共6页Proceedings of the CSEE

基　　金：国家自然科学基金项目(50595412)~~

摘　　要：对大型电力系统采用动态等值可显著降低计算量,并能突出主要特征。同调等值法作为动态等值的一种,其核心是同调机群的自动识别。为此,提出将最小二乘支持向量机(least square support vector machine,LS-SVM)应用于同调机群的在线识别,通过离线学习建立表征发电机同调性的特征参数与发电机之间同调性指标的非线性映射关系,并对比2组不同输入特征找出更能表现发电机同调性的特征参数。为进一步提高最小二乘支持向量机的学习、预测能力,提出采用多层动态自适应优化算法对其参数进行优化。最后通过对中国电力科学研究院(EPRI)36节点系统的仿真计算,结果表明该方法具有准确、快速的优点,并且适用于系统不同运行方式,能有效解决同调机群的识别问题。Dynamic equivalent can reduce the computing load and manifest the dominant characteristics. Coherency method is one approach of the dynamic equivalents. A method of on-line identification of coherent generators based on least square support vector machine （LS-SVM） is presented. By off-line learning, it establishes the relation between coherency features and coherency index of two generators, and finds out the better coherency features after comparing two sets of coherency features. In order to improve learning and forecasting ability of LS-SVM, a method called multi-layer adaptive optimization algorithm is presented to optimize the LS-SVM＇s parameters. It is tested on the China Electric Power Research Institute （EPRI） 36-bus model of power system analysis software package （PSASP）. The result demonstrates that the method has good accuracy and speed, which is also suitable for different service conditions. It can identify coherent generators effectively.

关 键 词：电力系统 同调机群 同调识别 最小二乘支持向量机 

分 类 号：TM744[电气工程—电力系统及自动化]