Power loss from dust coverage is a common issue in the solar industry. While having a significant affect on the power output and reliability, it also requires manpower to rectify the issue. Some sophisticated solar farms have automated cleaning and washing systems, however typically this cleaning process needs to be triggered by farm maintenance personnel.
VROC conducted a study based on publicly available real experimental data, collected from two solar panels. This experimental data included the images of two adjacent solar panels captured by a fixed position digital camera at 5 seconds rate, ambient and solar irradiance and recorded power output of two panels. In this experiment, one panel was used as a reference, and the other as test panel which was subjected to various types of dust coverage over a coarse of one month.
VROC's sophisticated data ingestion and storage pipeline collected more than 45,000 images. A region of interest (ROI) covering the test panel while removing the surroundings was defined through image processing and masking techniques and multiple statistical parameters e.g. the mean and standard deviation of images pixel intensity were extracted. For each panel, the time-transient trend of image features were analysed and filtered in order to highlight anomalies.
VROC AI models trained based on the historical data (solar-irradiance-normalised power loss and ROI statistics) were able to detect a significant increase of actual power loss compared to the model prediction during multiple periods of dust coverage. .
This study reveals how artificial intelligence can assist in the performance and health monitoring of photovolataic panels, providing solar farm operators critical insights for predictive maintenance, diagnostics, as well as structural health monitoring. These AI insights can provide operators with time-critical alerts for maintenance, ensuring the panels produce energy at consistently high levels. As solar energy makes up a growing portion of the energy grid, ensuring reliability is increasingly critical.
For a technical details about this case study, please contact our team or explore our AI solutions for renewable energy.