Seasonal Short-Term Load Forecasting (STLF) using combined Social Spider Optimisation (SSO) and African Vulture Optimisation Algorithm (AVOA) in Artificial Neural Networks (ANN)

Abstract

Accurate short-term load forecasting (STLF) is critical for efficient energy management, especially in regions like Nigeria, where electricity demand fluctuates due to climatic and socio-economic factors. This study proposes a hybrid model combining Social Spider Optimisation (SSO) and African Vulture Optimisation Algorithm (AVOA) to optimise Artificial Neural Networks (ANN) for improved STLF accuracy. The model was trained and validated using actual load data from the Nigerian grid for February, March, May, and June 2021. Quantitative evaluation using Mean Absolute Percentage Error (MAPE), Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Pearson Correlation Coefficient, and Coefficient of Determination (R²) showed superior performance of the SSO-AVOA model. The most stable results were recorded in May 2021, with MAPE of 0.202%, MAE of 8.47 MW, RMSE of 28.83 MW, and R² of 0.999, indicating nearly perfect forecasting. February and June periods showed relatively higher errors (e.g., MAPE up to 1.043% in February), reflecting the difficulty of forecasting during seasonal transitions. Findings confirm the robustness and adaptability of the hybrid model, which consistently maintains high correlation between actual and forecasted loads. However, error patterns during volatile periods suggest potential for improvement. Future work should integrate weather and socio-economic indicators, apply dynamic seasonal adaptations, and validate the model across Nigeria’s geopolitical zones. This study demonstrates that hybrid bio-inspired algorithms like SSO-AVOA are practical, high-performing tools for real-world load forecasting in dynamic and complex environment