Compound fault diagnosis for photovoltaic arrays based on multi-label learning considering multiple faults coupling

For photovoltaic (PV) systems with complex operating environment and long operation time, there are multiple faults coupled simultaneously. However, most of the existing fault diagnosis methods for PV systems can only diagnose single faults. In this paper, a composite fault diagnosis schema based on multi-label classification for PV systems with multi-fault coupling is proposed. In order to effectively distinguish between various faults, new effective features extracted from the preprocessed current–voltage (I–V) curves are used. Then, for realizing the diagnosis of compound faults, two different types of diagnostic models are developed, that is, k-Nearest Neighbor for multi-label learning (ML-KNN) combined with Random Forest (ML-RFKNN), and simply residual network for multi-label learning (ML-SResNet). Besides, a variety of simulations and experiments are performed to obtain enough PV fault datasets, and verify the performance of the compound fault diagnosis models, indicating that they have more excellent results at different irradiance and temperature levels compared with the existing models. Moreover, preprocessing of I–V curves and extraction of features are also analyzed and compared with the other literatures.