CX-102056: Robust PV Performance Loss Rate Prediction: Using Spatiotemporal Graph Neural Network Models in a Reliable System-Topology-Aware Learning Framework

Award Number: DE-EE0009353
CX(s) Applied: A9, A11
Solar Energy Technologies Office
Location(s): OH
Office(s): Golden Field Office

The U.S. Department of Energy (DOE) is proposing to provide federal funding to Case Western Reserve University to use machine learning, deep learning, and artificial intelligence approaches to develop a computer model which would predict the Performance Loss Rate (PLR) of photovoltaic (PV) plants. Rather than studying the performance of PV power plants individually, models will use data from various PV plants in a system’s fleet to illustrate how plants correlate in space and time and address questions of long-term performance for fleets of plants. Models would be used to understand PV challenges related to gaps in datasets or uncertain values, performance loss rate of PV plants over lifetimes, and inferring factors affecting PV plant performance loss rate.