A Machine Learning–Based Adaptive Framework for Wind Energy Potential Assessment across Nigeria’s Climatic Zones

The current focus on transitioning of energy to cleaner and more sustainable energy systems has led to global increase in the search for effective methods to assess renewable energy potential. In this study, a machine learning-based adaptive framework is proposed for the assessment of wind energy potential across he different climatic zones in Nigeria. The framework proposed the modelling of a complex temporal and spatial dependencies in wind patterns by training a Long Short-Term Memory (LSTM) algorithm, enhanced with the attention-based mechanism on the Nigerian Meteorological Agency (NiMet) long-term data with attributes such as elevation, wind speed, wind direction, maximum temperature (Tmax), and minimum temperature (Tmin).  The proposed framework incorporates the comparism of the enhanced deep learning model with the baseline statistical approach (Weibull distribution), to evaluate the robustness of the proposed model. Additionally, the proposed framework is designed to integrate the output of the deep learning model into the Geographic Information System (GIS) for spatial analysis and wind farm site suitability mapping. The proposed framework in this study addresses some limitations in previous studies, including the utilization of conventional statistical models, and the focus on single locations. The expected result from the proposed framework is a higher accuracy and precision, with improved spatial adaptability when compared with the Weibull-based models, enabling more reliable wind energy assessment across Nigeria’s climatic zones. This study provides a data-driven approach that can be utilized for forecasting wind energy with higher accuracy and precision, aiding decision-makers in renewable energy planning and national grid optimization.