光伏电站由数量庞大的光伏组件构成，因复杂的生产工艺及艰苦的工作环境，光伏系统直流侧故障频发，直接影响到光伏系统的发电效益。如何从光伏阵列的运行数据中提取有效的故障样本，并对其进行识别，是建立故障模型的重要步骤。因此提出一种基于模糊C均值(fuzzy C-means，FCM)聚类算法对故障样本进行划分及标识的方法。首先对故障条件下光伏阵列的输出特性进行分析，提取出故障特征向量(开路电压Uoc，短路电流Isc，最大工作点电压Um，最大工作点电流Im)。为排除外部激励条件对电气参数的影响，将故障特征向量统一转换至标准测试条件(standard test condition，STC)。最后根据FCM算法良好的模糊信息处理功能，对开路故障、短路故障、阴影故障、异常老化故障的样本进行聚类划分。实际运行数据证明，该方法可以精确、可靠地对光伏系统直流侧典型故障的样本进行智能聚类，并有效地描述不同故障下光伏阵列电气参数的分布特征。
School of Renewable Energy, North China Electric Power University, Changping District, Beijing 102206, China
Abstract:
The photovoltaic (PV) power station is composed of a large number of photovoltaic modules.Due to the complicated production technology and hard working environment, photovoltaic system DC-side faults occur frequently, directly affecting the photovoltaic system's power generation efficiency.How to extract valid fault sample from the PV array's operating data and identify the fault is an important step to establish a fault model.Therefore, a method based on fuzzy C-means (FCM) clustering to divide and identify the fault samples was proposed.Firstly, the output features of PV array under fault conditions were analyzed and the fault eigenvectors were put forward (open circuit voltage, short circuit current, maximum power point voltage and current).In order to exclude the influence of external excitation conditions on the electrical parameters, the fault eigenvectors were uniformly converted to the standard test condition (STC).Finally, according to the good fuzzy information processing function of FCM, the fault samples of open fault, short fault, shadow fault and abnormal aging fault were clustered.By using the actual operation data, it was proved that this method could accurately and reliably classify the fault samples of the typical fault on the DC side of the PV system and could effectively describe the distribution characteristics of the PV array's electrical parameters in different faults.
