Hybrid Grey Wolf Optimization and Gorilla Troop Optimizer Algorithms in ANN for Short-Term Load Forecasting

Abstract

Accurate short-term electricity load forecasting is critical for efficient power system operation and management. This study presents a hybrid metaheuristic model, the Grey Wolf Optimizer–Gorilla Troops Optimizer (GWO-GTO), for next-day load prediction. The model’s performance was evaluated using actual load data for April 24, 2021, and compared against conventional forecasting techniques. The GWO-GTO model achieved a Mean Absolute Percentage Error (MAPE) of 0.2741%, a Mean Absolute Error (MAE) of 12.8864 MW, and a Root Mean Square Error (RMSE) of 43.2202 MW, demonstrating superior forecasting accuracy. The model also attained a coefficient of determination (R²) of 0.99999667 and a Pearson Correlation Coefficient (PCC) of 0.99966458, indicating near-perfect alignment between actual and predicted loads. A comparative analysis over a seven-day period (April 24–30, 2021) confirmed the robustness of GWO-GTO, with consistently low MAPE values, peaking at 1.5518% on April 28. In a 168-hour comparative study, GWO-GTO outperformed other models, achieving the lowest MAPE of 2.5072%, MAE of 108.4440 MW, and RMSE of 154.1433 MW, confirming its effectiveness in capturing load variations. Compared to traditional models such as Artificial Neural Networks (ANN) and Genetic Algorithm (GA), GWO-GTO showed a 22–30% improvement in accuracy. These results establish GWO-GTO as a computationally efficient and highly accurate model for short-term electricity load forecasting.