Files and links (1)
CC BY V4.0, Open Access
Thesis
Machine Learning for Hour-Ahead Solar Radiation Prediction in Seattle, WA
Washington State University
Master of Science (MS), Washington State University
05/2025
DOI:
https://doi.org/10.7273/000007361
Abstract
Solar energy is rapidly becoming a major source of electricity generating capacity worldwide. However, unlike traditional sources of electricity, solar generation is subject to the diurnal cycle of the sun as well as transitory changes in weather. As a result, there is a significant need to predict solar generating capacity on both an hour-ahead and day-ahead schedule so utilities can dispatch appropriate resources to meet demand. This paper explores the hour-ahead prediction of solar radiation and makes two contributions to the field. First, we develop a suite of machine learning based models for predicting hour-ahead ground level solar global irradiance in Seattle, WA, an area known for dense cloud cover and long periods of rain. The best of these models is implemented using XGBoost with a Root Mean Squared Error (RMSE) of 49.843 W/m2 and a Normalized Root Mean Squared Error (NRMSE) of 0.334. We demonstrate that this model improves upon alternative architectures in terms of the trade-off between test performance, training efficiency, and prediction efficiency. Additionally, we identify several properties of past work that make it difficult to fully evaluate the current landscape. These include critical differences in how training and testing sets are produced and how error is measured. Based on this analysis, we provide a set of recommendations for future work that we believe will help to improve understanding of the field at large.
Metrics
1 File views/ downloads
15 Record Views
Details
- Title
- Machine Learning for Hour-Ahead Solar Radiation Prediction in Seattle, WA
- Creators
- Aiden Dickson
- Contributors
- Scott Wallace (Chair)Grant Williams (Committee Member)Xinghui Zhao (Committee Member)
- Awarding Institution
- Washington State University
- Academic Unit
- School of Engineering and Computer Science (VANC)
- Theses and Dissertations
- Master of Science (MS), Washington State University
- Publisher
- Washington State University
- Number of pages
- 47
- Identifiers
- 99901221251501842
- Language
- English
- Resource Type
- Thesis