A Long Term Wind Speed Forecasting Model Using Enhanced Hybrid Feature Selection and Deep Learning Approaches

Abstract

Wind energy has become a pivotal renewable energy source as the world newlineconfronts climate change and the depletion of fossil fuel reserves. As renewable energy newlinesystems play a growing role in reducing global carbon emissions, accurate wind speed newlineforecasting (WSF) has emerged as a critical component for the efficient design, newlineoperation, and integration of wind power systems, particularly those connected to the newlineelectrical grid. Improved wind forecasting is essential to better managing intermittent newlineenergy generation, thereby reducing the need for fossil fuel-based backup power, newlinedirectly contributing to climate change mitigation and supporting global sustainability newlinegoals. However, the nonlinear, noisy, and highly stochastic nature of wind data presents newlinesignificant challenges, complicating efforts to achieve precise long-term wind speed newlinepredictions. This research demonstrates a 22.8% improvement in prediction accuracy newlineover traditional models, addressing these challenges head-on. The models were trained newlineand tested on extensive datasets collected from nine weather stations in the Tirunelveli newlinedistrict of Tamil Nadu, India, covering data from 2010 to 2022, ensuring robust and newlinegeographically diverse validation newline