ENHANCING GHANA'S ENERGY SECTOR SUSTAINABILITY THROUGH MACHINE LEARNING-BASED SOLAR FORECASTING AND DATA-DRIVEN POLICY INTEGRATION

Ghana’s energy sector remains heavily reliant on thermal and hydroelectric sources, wherein
both are susceptible to climate variability and high operational costs. This study enhances the
sustainability of Ghana’s energy mix through machine learning-based solar forecasting and data-
driven policy integration. The study focused on utilizing machine learning models to learn from
historical solar irradiance data from 1984 to 2022 for Ghana, with projections extended to 2030.
Solar irradiance refers to the amount of solar radiation received per unit area and it is typically
measured in watts per square meter (W/m2). Solar irradiance is critical as it determines the
potential power output of solar panels and serves as an input for predicting solar energy
availability over different time horizons.

 The research trained and evaluated three models:
Random Forest, XGBoost and Deep Neural Networks for national solar irradiance prediction.
XGBoost achieved the highest accuracy with a coefficient of determination of 0.9575 and a Root
Mean Square Error of 0.1086 kWh/m2/day which outperformed the other models. The findings
identified the Upper East Region, Upper West Region and Northern Region as high potential
solar zones with dry-season months of November to March showing peak irradiance levels
exceeding 5.5kWh/m2/day. The study recommends integrating these forecasts into Ghana’s
renewable energy policies to guide infrastructure investments while supporting Sustainable
Development Goal 7. Despite data resolution and computational limitations, the findings provide
an evidence-based framework for advancing clean and resilient energy planning in Ghana.