AI for Good

Chapter 13Forecasting and Explaining Degradation of Solar Panels with AI

—Felipe Oviedo and Tonio Buonassisi

Executive Summary

Solar energy is an increasingly important way to address climate change. Solar energy panels use photovoltaics to transform light energy from the sun into stored energy. Degradation of photovoltaics is when their performance suffers over time and under rigorous environmental conditions: degradation means that the panel is not operating at maximal efficiency or effectiveness, and it presents a technical and market hurdle to novel photovoltaic development.

Understanding degradation requires complex measurements on multiple samples over time; therefore, we developed DeepDeg, a machine learning model that combines deep learning, explainable machine learning, and physical modeling that can forecast and explain degradation. Using a large and diverse dataset of organic solar cell stability tests and validating the model with a large held-out dataset, we found that, using only a few initial hours of degradation, DeepDeg can accurately and quickly predict degradation dynamics, characterize the degradation, and explain the physiochemical factors driving degradation.

Why Is This Important?

Solar energy is an important arrow in the quiver of fighting climate change. Solar energy incorporates photovoltaics into panels that capture energy from the sun and transform it into energy that can be stored and used by humans. However, degradation—meaning a reduction ...