Photovoltaic Array Fault Detection and Classification based on T-Distributed Stochastic Neighbor Embedding and Robust Soft Learning Vector Quantization

Photovoltaic (PV) as one of the most promising energy alternatives brings a set of serious challenges in the operation of the power systems including PV system protection. Accordingly, it has become even more vital to provide reliable protection for the PV generations. To this end, this paper proposes two-stage data-driven methods. In the first stage, a feature selection method, namely t-distributed stochastic neighbor embedding (t-SNE) is implemented to select the optimal features. Then, the output of t-SNE is directly fed into the strong data-driven classification algorithm, namely robust soft learning vector quantization (RSLVQ) to detect PV array fault and identify the fault types in the second stage. The proposed method is able to detect the two different line-to-line faults (in strings and out of strings) and open circuit fault and fault type considering partial shedding effects. The results have been discussed based on simulation results and have been demonstrated the high accuracy and reliability of the proposed two-stage method in detection and fault type identification based on confusion matrix values.