IE 11 Not Supported

For optimal browsing, we recommend Chrome, Firefox or Safari browsers.

Machine Learning Helps the Metropolitan Water District of Southern California Maximize Solar Power Investments

Automated analysis of thermal imaging helps the district rapidly spot problems in its solar farms.


The Metropolitan Water District of Southern California (MWD) — which provides water to 19 million people in a 5,200-square-mile service area — uses machine learning technology to improve the efficiency of its solar energy arrays and quickly spot failing solar panels. MWD, which has invested heavily in solar energy to power its facilities, developed a deep learning model that automatically reviews and analyzes thermal imagery of solar panels collected by drones flying over department solar farms. This analysis can show areas of lower efficiency within solar farms and spot failing solar panels while they are still under warranty.


Automating the analysis of thermal imagery has dramatically reduced turn-around time and improved accuracy of this analysis, according to the district. MWD staff initially reviewed thermal imagery manually, digitizing anomalies by hand and comparing these areas to natural color high-resolution images. The deep learning model can recognize the solar panels themselves and identify any thermal anomalies within the solar panels at the tolerances specified during the model's training. In an early proof of concept, the deep learning algorithm recognized verified solar anomalies at a rate nearly double that of a human analyst, MWD says.

MWD operations staff use this data to quickly identify panel failures, obstructions or other issues that may reduce efficiency, thus protecting and maximizing the district’s investment in the renewable energy. Now MWD plans to expand its use of this technology into dam surface analysis, road pavement maintenance, crop fallowing, facility inspections and irrigated turf water usage.


Districts interested in leveraging this technology will need access to high-resolution natural color and thermal imagery, which can be captured by a drone or aerial photography. If captured by an in-house drone program, the images must be processed by software that can provide a spatially enabled mosaic that can be imported into a GIS software platform. A GIS analyst can use this mosaic to identify features from the imagery that will be used to train the model. As with any machine learning algorithm, the results will improve with additional training and repetition.