Machine learning-driven wind energy mapping enhanced by natural neighbor interpolation

In the present work, a prediction on the wind energy potential in Semarang City (Central Java Province, Indonesia) has been performed by leveraging a novel combination of machine learning and natural neighbor interpolation (NNI) methodology. This integrated approach uniquely combines the predictive power of machine learning to estimate wind speeds based on historical and spatial data, with the spatial mapping capabilities of NNI, which provides a more accurate and seamless visualization of wind speed distribution. This combination addresses challenges of data sparsity and variability, offering a more reliable and localized mapping approach than traditional methods. Additionally, air density is considered to calculate energy density, enabling a comprehensive evaluation of wind energy potential. The results show an average monthly wind speed of 5.23 m/s, ranging from 3.38 m/s to 7.39 m/s. Wind speeds between 7 m/s and 10 m/s are predicted to occur for up to 10 months annually, with an estimated energy density of 102.7 W/m². These findings underscore the feasibility of small-scale wind power generation in the study area and provide actionable insights for advancing renewable energy policies and implementations at the local level.