Electrical and Computer Engineering (ECE) Department doctoral students Akansha Singh Bansal and Noman Bashir, both advised by ECE Professor David Irwin, won prizes at the “Lightning Talks Competition” during the Energy Data Analytics Symposium, hosted by the Duke University Data Analytics Lab on December 8 and 9. Bansal won a second-place prize of $250 for her talk, titled “See the Light: Modeling Solar Performance using Multispectral Satellite Data.” Meanwhile, Bashir earned an honorable mention for a talk titled, “Solar-TK: A Data-driven Toolkit for Solar PV Performance Modeling and Forecasting."

The competition highlighted research by emerging scholars in energy data analytics, attracting 21 entries from 12 universities and organizations. Judges assessed participants’ five-minute “lightning talks” on: compelling communication of the core ideas and outcomes of the project to an interdisciplinary audience; and innovation and potential for impact of the energy application and data science methodology.

As Bansal described her talk, “Developing accurate solar performance models, which infer solar output from widely available external data sources, is increasingly important as the grid's solar capacity rises. These models are important for a wide range of solar analytics, including solar forecasting, resource estimation, and fault detection.”

But the most significant error in existing models, said Bansal, is inaccurate estimates of the effects of clouds on solar output, especially because cloud formations and their impact on solar radiation are highly complex.

According to Bansal, in 2018 and 2019 the National Oceanic and Atmospheric Administration in the U.S. began releasing multispectral data from 16 different light wavelengths from the GOES-16 and GOES-17 satellites every 5 minutes. Enough channel data is now available to learn solar performance models using machine learning.

“In this work,” said Bansal, “we show how to develop both local and global solar performance models using machine learning on multispectral data and compare their accuracy to existing physical models based on ground-level weather readings and on NOAA's estimates of downward shortwave radiation (DSR), which also derive from multispectral data but using a physical model.”

Bansal concluded that “We show that machine-learning-based solar performance models based on multispectral data are much more accurate than weather- or DSR-based models, improving the average MAPE across 29 solar sites by over 50 percent for local models and 25 percent for global models.”

As Bashir said about his talk, "Solar energy capacity is continuing to increase. The key challenge with integrating solar into buildings and the electric grid is its high-power generation variability, which is a function of many factors, including a site’s location, time, weather, and numerous physical attributes.

Unfortunately, as Bashir explained, much of the prior work on solar performance modeling and forecasting is not accessible to researchers, either because it has not been released as open-source, is time-consuming to re-implement, or requires access to proprietary data sources.

“To address the problem,” said Bashir, “we present Solar-TK, a data-driven toolkit for solar performance modeling and forecasting that is simple, extensible, and publicly accessible. Solar-TK’s simple approach models and forecasts a site’s solar output given only its location and a small amount of historical generation data.”

As Bashir explained, Solar-TK’s extensible design includes a small collection of independent modules that connect together to implement basic modeling and forecasting, while also enabling users to implement new energy analytics.

“We have released Solar-TK as open-source to enable research that requires realistic solar models and forecasts,” said Bashir, “and to serve as a baseline for comparing new solar modeling and forecasting techniques.” (January 2021)

Full story available on the College of Engineering Electrical and Computer Engineering website.