Enhancing Medium-Term Electric Load Forecasting Accuracy Leveraging Swarm Intelligence and Neural Networks Optimization

For power systems to be designed, planned, and managed effectively, electric load forecasting is essential. This study employs the Artificial Bee Colony (ABC) and Particle Swarm Optimization (PSO) methods to optimize Medium-Term Load Forecasting (MTLF) for the Abuja Municipal Area Council (AMAC). The study uses the Abuja Electricity Distribution Company (AEDC) load consumption data from January 2012 to December 2017 and trains the dataset using the Levenberg-Marquardt and Bayesian Regularization techniques. The models used for load forecasting are Multiple Linear Regression (MLR) and Non-linear Autoregressive Neural Network (NARX-NN). The metrics used for evaluation are Mean Absolute Percentage Error (MAPE) and Root Mean Square Error (RMSE). The results show that NARX-NN, which was trained using Bayesian regularization, performs better than MLR; it is especially noteworthy for its ability to accurately anticipate the uneven load curve of 2018. Forecasting accuracy is much increased when NARX-NN and PSO are combined; these findings are further supported by ABC optimization outcomes. The goal of obtaining a MAPE value of less than 0.1% is effectively attained by this method, confirming the effectiveness of the NARX-NN and PSO integration in improving load forecasting accuracy.