Online photovoltaic fault detection method based on data stream clustering

Abstract

As the core component of solar power station, PV array is particularly important for safe and stable operation of the entire system. The existence of PV array faults for a long time can lead to potential danger of the entire PV system. Since the PV data is greatly affected by the environment, the continuous data stream generated during the operation of PV arrays can form clusters of arbitrary shape. When a PV fault occurs, new data streams can form a cluster that is different from the one under normal operation. Accordingly, this paper presents a model for online fault detection of PV arrays faults using data stream clustering approach. The real-time data stream of the PV arrays is transmitted to the diagnostic system through RabbitMQ server for online detection and data storage. The online density-based spatial clustering of applications with noise (DBSCAN) algorithm is used for clustering the data. Then, the faults are detected by judging whether new clusters are formed. The experiment result shows the effectiveness of the proposed method in grid-connected PV system.