Computer vision-based regression techniques for renewable energy : predicting energy output and performance

Chaudhuri, Arindam and Jiang, Richard (2024) Computer vision-based regression techniques for renewable energy : predicting energy output and performance. In: Computer Vision and Machine Intelligence for Renewable Energy Systems :. Elsevier, pp. 41-66. ISBN 9780443289484

Abstract

Renewable energy sources have provided us with several alternatives in comparison to regular and climate-hostile fuels. However, their usage comes with several issues. Using renewable options, users consume as well as produce and supply energy. It results in the bidirectional flow of energy. In spite of increased flexibility from renewable sources, management of supply and demand has become more challenging. In view of this considerable attention has been concentrated toward smart grid stability. In this research, we address this issue with computer vision-based regression techniques which are basically ensemble methods. The hybrid ensemble comprises an improved Cox proportional hazard method, improved embedded semisupervised learning, and improved long short-term memory as its integral components. Here, sensor and historical data are collected and prepared using benchmarking methods. The stability modeling and prediction on multisource data is successfully achieved. The results obtained are compared with stated and state-of-the-art methods and several benchmarks. This framework offers great potential for achieving more profitable, efficient, and sustainable smart grid solutions. This helps better energy data implementation toward desired business solutions. The method is superior in comparison to other methods as evident from results.