Publication Date
Spring 2026
Degree Type
Master's Project
Degree Name
Master of Science in Computer Science (MSCS)
Department
Computer Science
First Advisor
Katerina Potika
Second Advisor
Robert Chun
Third Advisor
Amith Kamath Belman
Keywords
Forecasting, Global Horizontal Irradiance, Graph Neural Network, Temporal Graph, Edge Features
Abstract
Solar energy has emerged as a promising alternative to traditional sources due to its many environmental and economic advantages, especially in places with abundant sunlight. It is renewable, and lowers the carbon footprint, but is hard to store and to transfer over long distances. Power grid stability is crucial for the continuous heat and electricity that consumers require. However, the unpredictable and intermittent nature of solar irradiance can threaten the reliability of the power grid when solar PV (Photovoltaic) systems are integrated alongside other sources. Solar irradiance forecasting for PV estimation is necessary to maintain grid resilience, as it aids in power system scheduling and power ramp-rate control to maintain a smooth output. The aim of this project is to take advantage of the spatiotemporal nature of solar irradiance across multiple geographical sites (nodes) to predict Global Horizontal Irradiance (GHI) on these sites. The proposed model, WindCast-Spatio-Temporal Advection Routed Network (hereafter referred to as WindCast-STARNet), leverages Graph Convolutional Networks and Temporal Convolutional Networks with dynamic edge features. The model incorporates static geographic edge features and dynamic wind and irradiance-aware edge features to refine inter-site message passing. Exper- imental results on the National Solar Radiation Database show that the proposed model outperforms temporal-only and graph-based baseline models, demonstrating the benefit of combining graph-based spatiotemporal learning with sparse advection-aware message passing for multi-site solar irradiance forecasting.
Recommended Citation
Challa, Charita, "WindCast-STARNet: A SpatioTemporal Advection-Routed Network for Solar Forecasting" (2026). Master's Projects. 1796.
DOI: https://doi.org/10.31979/etd.nm5j-x42g
https://scholarworks.sjsu.edu/etd_projects/1796
