Should you clean your solar panels now?

Abstract

Photovoltaic (PV) applications worldwide are being increasingly deployed; however, the performance of PV systems is greatly affected by external factors, such as weather, environment, and terrain. These factors can form anomalies on solar panels, and lead to low performance. We present a data driven approach to identify anomalies on panels, based on the power output data and using our simplified solar irradiance model. The approach includes the in-plane solar irradiance model, disaggregation and detection of anomalies, and a decision tree to classify the types of the anomalies. The detection sensitivity is adjustable to suit different production environments. This methodology can be applied in multiple areas, such as anomaly alerts, energy loss analysis on different interval bases, and so on. The approach has been tested using real data collected in two cities in Ontario, Canada. The classification has a 85% precision rate for the detected anomalies.