基于DS证据理论与多标签算法的区域电网故障诊断方法

Fault diagnosis method of regional power grid based on DS evidence theory and multi-label algorithm

摘要

为提升电网故障诊断的准确性与可靠性,确保区域电网平稳运行,提高电力服务质量,提出基于DS证据理论与多标签算法的区域电网故障诊断方法。针对常规标签增量特征变换算法算法忽略标签之间关联性的问题,标签关联对其作出改进。利用改进后的算法实现对区域电网各类故障数据的分类处理,以此建立多标签故障集合,将其作为电网故障诊断的证据源;利用Yager规则改进常规的德普斯特-谢弗证据理论,在处理证据冲突时提供不同的策略,以此来构建区域电网故障诊断识别框架;在识别框架中,利用基本概率分配对证据源中的证据进行表达,引入电网元件边际谱熵与故障度熵指标作为独立的证据体建立可信度函数,利用该函数完成证据的合成,依据合成后的证据实现对区域电网的故障诊断。实验结果表明,该方法可以有效且精准地诊断出存在于区域电网中的故障,面对不同类型的电网故障可以保持较高的识别灵敏度。

Abstract

To improve the accuracy and reliability of power grid fault diagnosis, ensure the smooth operation of regional power grids, and improve the quality of power services, a regional power grid fault diagnosis method based on DS evidence theory and multi label algorithm is proposed. Aiming at the problem of ignoring the correlation between labels in conventional incremental feature transformation algorithms, label correlation are improved. The improved algorithm is used to classify and process various types of fault data in regional power grids, and a multi label fault set is established as the evidence source for power grid fault diagnosis; Using the Yager rule to improve the conventional Dempster-Shafer theory of evidence and provide different strategies for handling evidence conflicts, in order to construct a regional power grid fault diagnosis and recognition framework; In the recognition framework, basic probability allocation is used to express the evidence in the evidence source, and the marginal spectral entropy and fault degree entropy indicators of power grid components are introduced as independent evidence bodies to establish a credibility function. This function is used to complete the synthesis of evidence, and based on the synthesized evidence, fault diagnosis of regional power grids is achieved. Experiments show that this method can effectively and accurately diagnose faults existing in regional power grids, and maintain high recognition sensitivity in the face of different types of power grid faults.

关键词

德普斯特-谢弗证据理论 / 多标签算法 / 区域电网 / 故障诊断 / 标签增量 / 特征变换算法 / 可信度函数

Key words

Dempster-Shafer theory of evidence / multi-label algorithm / regional power grid / fault diagnosis / label incremental / feature transformation algorithm / confidence function