Forecasting Power Generation in Photovoltaic Plants using Artificial Neural Networks and Transfer Learning

Abstract

The global energy demand is constantly rising, driven by population growth, changes in consumption patterns, and technological advancements. However, concerns about climate change and the limited availability of fossil fuels are driving the search for renewable sources, with solar energy standing out. Solar production, while abundant, is sensitive to weather conditions, making its prediction challenging. New photovoltaic plants present an additional challenge due to their lack of historical data. This work proposes to address the challenge of energy generation forecast in photovoltaic plants using Transfer Learning and Artificial Neural Networks (ANNs), aiming to improve predictions through previous knowledge. Transfer Learning strategies, the benefits of using pre-trained models, and their effectiveness are discussed and compared to traditional approaches. The results indicate that Transfer Learning in ANNs shows promise in improving energy generation forecasting from photovoltaic plants.