Machine Learning-Based Wind Speed Estimation for Renewable Energy Optimization in Urban Environments: A Case Study in Kano State, Nigeria

Abstract

Climate change always had a massive effect on worldwide cities. which can only be decreased through considering renewable energy sources (wind energy, solar energy). However, the need to focus on wind energy prediction will be the best solution to the world electricity petition. Wind power (WP) estimating techniques have been used for diverse literature studies for many decades. The hardest way to improve WP is its nature of differences that make it a tough undertaking to forecast. In line with the outdated ways of predicting wind speed (WS), employing machine learning methods (ML) has become an essential tool for studying such a problem. The methodology used for this study focuses on sanitizing efficient models to precisely predict WP regimens. Two ML models were employed “Gaussian Process Regression (GPR), and Feed Forward Neural Network (FFNN)” for WS estimation. The experimental methods were used to focus the WS prediction. The prophecy models were trained using a 24-hour’ time-series data driven from Kano state Region, one of the biggest cities in Nigeria. Thus, investigating the (ML) forecast performance was done in terms of coefficient of determination (R²), linear correlation coefficient (R), Mean Square Error (MSE), and Root Mean square error (RMSE). Were. The predicted result shows that the FFNN produces superior outcomes compared to GPR. With R²= 1, R = 1, MSE = 6.62E-20, and RMSE = 2.57E-10