基于高斯核模糊C均值聚类的光伏阵列故障诊断方法

PHOTOVOLTAIC ARRAY FAULT DIAGNOSIS BASED ON GKFCM

光伏阵列故障诊断过程中传统的故障特征量难以区分特征相似的单故障和多重故障情况,而实际诊断中外场实验采集到的数据也含有较强的噪声,从而导致故障诊断准确率下降。针对这一问题,提出由一个新的特征向量对不同故障进行表征,该特征向量包含:归一化电压V_(norm)、归一化电流I_(norm)、填充因子FF共3个特征量,并利用这3个特征量采用高斯核模糊C均值聚类(GKFCM)方法对光伏阵列中8种故障进行故障识别的方法。这3种故障特征量的结合可有效减少外界气象条件对故障识别的影响;由于GKFCM对复杂数据集具有良好的聚类性能,在复杂环境下不同故障类的识别过程中可有效提高识别准确率。该算法分为训练和测试阶段,在训练阶段对训练集中故障数据利用3个特征量构成的特征向量进行表征并聚类获取类心,在故障识别阶段新故障数据利用同样的方法获得聚类类心并与训练阶段获得的各类故障类心进行相似度计算,从而实现故障分类和新故障的识别。该方法不仅可诊断单故障情况,也可识别多重故障情况,具有较强的抗干扰能力。最后通过仿真及实验证明该方法可有效诊断光伏阵列中的常见故障。

In the process of PV array fault diagnosis,the traditional characteristics are difficult to identify the single and some multifaults with similar features,and the fault datasets collected in the operating condition of the field experiment are disturbed by the external environment in the meantime,these two reasons lead to the failure rate of the diagnostic accuracy.In order to conquer the problem,a new eigenvector including V_(norm),I_(normand)FF is proposed to characterize these fault conditions.This feature vector,combined with Gaussian kernel fuzzy C-means clustering method,is used to classify eight common faults in PV array is proposed in this paper.The combination of these three fault characteristics effectively reduces the impact of environmental conditions in the process of fault identification.The GKFCM has good performance for clustering the complex datasets,so it can effectively improve the recognition accuracy in the process of identify different fault classes with noise interference.The algorithm contains two phases:training and testing process.In the training phase,the fault datasets are clustered to obtain the center points and then the new fault data is classified as an existing fault type or a new fault kind by calculating the similarity between the existing center points and new fault data.This method can not only identify single fault condition but also detect multiple failure conditions.Finally,the Simulink simulation and the field experiment showed that the method can effectively detect several main types of PV array faults.