Design of a Renewable Energy Forecasting System for Anambra State Using Machine Learning

Effective planning and integration of solar and wind energy resources into Anambra State's electricity grid requires accurate renewable energy generation forecasting, a capability currently absent from state energy management frameworks. This study designs a machine learning-based renewable energy forecasting system for Anambra State, predicting daily solar irradiance and potential wind energy output using historical meteorological data from NASA POWER and the Nigerian Meteorological Agency. Three forecasting models, including Long Short-Term Memory (LSTM) recurrent neural networks, Support Vector Regression, and a gradient boosting ensemble, were trained and evaluated on a 10-year historical climate dataset for Awka and surrounding areas. Feature variables included temperature, humidity, cloud cover, wind speed, and solar zenith angle. Model performance was assessed using Root Mean Square Error and Mean Absolute Percentage Error. Results indicate that the LSTM model achieved the lowest RMSE of 0.41 kWh/m2 per day for solar irradiance prediction and a MAPE of 8.7 percent for wind output estimation, outperforming both baseline models. A web-based forecasting dashboard prototype was developed to visualize 7-day renewable output projections. The study concludes that machine learning forecasting systems can provide actionable renewable energy planning insights for Anambra State energy authorities. Recommendations include real-time sensor data integration from deployed solar installations and collaboration with the Anambra State Rural Electrification Agency.