Enhancing Renewable Energy Storage Conversion Efficiency using ERFE with FFNN

Pari-Condori E.Y.; Rao G.R.K.; Abdulkader R.; Kumar V.K.; Jeyaraj J.P.G.; Quispe-Ramos E. · Journal of Machine and Computing (2024) · DOI: 10.53759/7669/jmc202404005

Resumen

The 21st century witnesses a pivotal global shift towards Renewable Energy Sources (RES) to combat climate change. Nations are adopting wind, solar, hydro, and other sustainable energy forms. However, a primary concern is the inconsistent nature of these sources. Daily fluctuations, seasonal changes, and weather conditions sometimes make renewables like the sun and wind unreliable. The key to managing this unpredictability is efficient Energy Storage Systems (ESS), ensuring energy is saved during peak periods and used during low production times. However, existing ESSs are not flawless. Energy conversion and storage inefficiencies emerge due to temperature changes, inconsistent charge rates, and voltage fluctuations. These challenges diminish the quality of stored energy, resulting in potential waste. There is a unique chance to address these inefficiencies using the vast data from renewable systems. This research explores Machine Learning (ML), particularly Neural Networks (NN), to improve REES efficiencies. Analyzing data from Palm Springs wind farms, the study employs an Entropy-Based Recursive Feature Elimination (ERFE) coupled with Feed-Forward Neural Networks (FFNN). ERFE utilizes entropy to prioritize essential features, reducing redundant data and computational demands. The tailored FFNN then predicts energy conversion rates, aiming to enhance energy storage conversion and maximize the usability of generated Renewable Energy (RE).

Instituciones: Universidad Nacional del Altiplano; Universidad Andina del Cusco