基于支持向量机的光伏系统电弧检测方法

The arc fault detection method for photovoltaic system based on support vector machine

光伏发电设备在长期使用中受外力侵袭等因素影响,可能产生电弧故障而引发火灾,准确检测故障电弧是提高光伏系统安全性的关键。文章基于Cassie电弧模型,搭建了光伏系统电弧故障仿真模型,对数种典型环境下的不同故障和正常工况进行仿真,获取了正常工况和电弧故障的电流波形数据,并基于支持向量机对电弧故障特征值进行了分类。结果表明:在正常状态与故障状态下,小波高频分量能量值和模极大值相差两个以上的数量级,可作为电弧检测的故障判据;建立的分类器可以准确检测光伏系统在多种环境条件下不同位置、不同类型的电弧故障,具有较好的泛化能力。

During the long-term utilization of photovoltaic equipment,affected by external forces and other factors,some hidden dangers may lead to arc faults and fire disaster.Accurate detection of arc fault is the key to improve the photovoltaic(PV)system security.Based on Cassie arc model,the simulation model of PV system arc fault is established in this paper,and the current waveform data in normal and various arc faults condition is obtained.Two indexes,the wavelet high frequency component energy value and modulus maximum value,are found to have better discrimination through comparison.The sample data set is formed and the support vector machine(SVM)classifier is trained and tested.The results show that the difference between normal state and fault state is more than two orders of magnitude,which can be used as fault criterion for arc detection.The SVM classifier has the ability to detect arc faults in different positions and various environmental conditions accurately.Simulation results shows generalization ability of the proposed method.