Online Fault Detection in PV Systems

This paper presents the development of a practical fault detection approach in photovoltaic (PV) systems, intended for online implementation. The approach was developed and validated using field measurements from a Canadian PV system. It has a fairly low degree of complexity, but achieves a high fault detection rate and is able to successfully cope with abnormalities present in real-life measurements. The fault detection is based on the comparison between the measured and model prediction results of the ac power production. The model estimates the ac power production using solar irradiance and PV panel temperature measurements. Prior to model development, a data analysis procedure was used to identify values not representative of a normal PV system operation. The original 10-min measurements were averaged over 1h, and both datasets were used for modeling. In order to better represent the PV system performance at different sunlight levels, models for different irradiance ranges were developed. The results reveal that the models based on hourly averages are more accurate than the models using 10-min measurements, and the models for different irradiance intervals lead to a fault detection rate greater than 90%. The PV system performance ratio (PR) was used to keep track of the system's long-term performance.