基于RS-SVM算法的配电网故障诊断方法

Fault Diagnosis Method Based on RS-SVM Algorithm for Power Distribution Network

为提高配电网故障诊断工作效率,针对故障发生时故障信息数据冗杂并且夹杂着设备元件误动、拒动、损坏等信息不完备的情况,提出一种基于RS-SVM算法的故障诊断方法。该方法将故障告警信息中断路器、过流保护动作信息作为故障诊断模型的输入量,将系统中线路或区域(如线路、母线、配电变压器、配电区域等)是否发生故障作为诊断输出量。首先,根据配电网拓扑结构关系建立原始决策表,利用粗糙集(rough set,RS)对原始决策表进行基于依赖度和重要度的属性约简,求得最小决策表;然后,建立支持向量机(support vector machine,SVM)故障诊断模型,通过优化参数完成训练学习,利用约简后的决策数据进行故障诊断;最后,充分考虑单次和多重故障的情况,结合多种评价指标,通过算例对比验证了所提方法的精确性和高效性。

To improve working efficiency of fault diagnosis for the power distribution network,this paper proposes a kind of fault diagnosis method based on RS-SVM to solve problems of lengthy and jumbled faulted information data and incomplete information such as maloperation,rejection and damage of equipment components.This method takes information of the circuit breaker and overcurrent protection action in the faulted alarming information as the input of the fault diagnosis model and fault occurrence of lines or regions such as system lines,buses,distribution transformers,distribution areas,and so on as the output of the model.The method firstly uses the rough set(RS)for attribute reduction based on dependency and importance for the original decision-making table established according to the topology of the power distribution network so as to obtain the minimum decision-making table.Then it establishes a support vector machine(SVM)fault diagnosis model,makes optimization on parameters to finish training and learning,and uses decision-making data after reduction for fault diagnosis.Finally,considering conditions of single fault and multiple faults and combining various evaluation indicators,it compares and verifies accuracy and high efficiency of the proposed method.