EAGER: EPCN: Computational Imaging of the Sky for Precise Prediction of Solar Variability

Carnegie Mellon University, Pittsburgh PA

Investigators

Abstract

Renewable energy sources such as the sun are naturally intermittent. This uncertainty increases the operational cost of systems and processes that are reliant on solar power, thus hindering the widespread adoption of solar photovoltaics. This NSF EAGER project aims to develop a system for predicting the distribution of solar power for time horizons spanning minutes to an hour. The project will bring transformative change to energy dispatch and scheduling systems that affect a broad range of residential and commercial solar producers and consumers. This will be achieved by imaging the sky to monitor the spatial distribution of clouds as well as their movement. The intellectual merits of the project include the design of a sky imaging system that can acquire high-resolution images of clouds especially near the horizon, learning-based techniques for predicting the evolution of clouds, and the evaluation of the prediction against a solar photovoltaic testbed. The broader impacts of the project include workshops for K-12 students as well as dissemination efforts that will allow the broader public to visualize and utilize the solar energy forecasts. To achieve the goal of solar power forecasting, this project develops new tools in computational imaging as well as physics and data-aware modeling for predicting the evolution of the sky image over long-time horizons. To this end, the research will build a reconfigurable sky imager that mitigates the severe loss of resolution near the periphery of the field of view (or the horizon). The captured imagery will allow us to model the dynamics of clouds using physics-aware learning that considers wind flow as well as data-driven models. We will validate the forecasts made by our system against actual measurements made with a co-located photovoltaic system. Extensive analysis and experimentation will be performed with setups involving multiple distributed imaging and solar units to assess and enhance the scalability of the proposed prediction solutions. The impact of the core technology developed in this project will be amplified via outreach and education efforts. We will organize workshops for middle and high-school students through established programs at Carnegie Mellon University. We will also engage in citizen science projects using a web portal that will allow access to the solar energy forecasts made by our system. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.