论文部分内容阅读
在配电网出现孤岛、故障等异常情况时,并网保护需要及时断开分布式电源系统与主电网的连接。文中首先分析了并网保护的功能与特点,然后提出了一种基于多分类支持向量机的智能型并网保护方案。该保护方案在传统二分类支持向量机的基础上,建立了并网保护的概率型多分类器模型,实现对扰动、孤岛和故障事件的检测。为了提高模型的泛化能力,采用SVM-RFE方法对特征进行选择。最后,从保护的可信赖性与安全性两个方面,利用仿真对所提智能型并网保护与常规保护的性能进行了对比。
In the distribution network appears island, fault and other anomalies, the need for timely protection of grid interconnection distributed power system and the main grid connection. Firstly, the paper analyzes the functions and features of grid-connected protection, and then proposes an intelligent grid-connected protection scheme based on multi-class support vector machines. Based on the traditional two-class SVM, this protection scheme establishes a probabilistic multi-classifier model for grid-connected protection, and realizes the detection of disturbance, islanding and fault events. In order to improve the generalization ability of the model, SVM-RFE method is used to select the features. Finally, from the aspects of trustworthiness and security of protection, we compared the performance of the proposed intelligent grid-connected protection with that of conventional protection.