Extension on Photovoltaic Power Generation for Short-term Solar Forecasting

Abstract

This study aims to explore the application of deep learning architectures to improve the accuracy and optimize the resource use of solar power generation forecasting for practical application. Addressing the challenge of variability in solar power output, the study evaluates three different Convolutional Neural Network (CNN) models using an extensive dataset of sky images and photovoltaic (PV) power generation collected over three years. This research fills a gap in previous methodologies by leveraging advanced image processing and deep learning for real-time prediction. The models were tested for efficiency and accuracy over four metrics. The findings suggest that precise solar power predictions can enhance grid stability, optimize energy management, and reduce operational costs for energy providers. This study contributes to the broader goal of sustainable energy development, offering a pathway to integrating renewable energy more effectively into existing power grids.