@inproceedings{scholars20567,
         journal = {7th International Conference on Energy, Power and Environment, ICEPE 2025},
            year = {2025},
             doi = {10.1109/ICEPE65965.2025.11139390},
       publisher = {Institute of Electrical and Electronics Engineers Inc.},
           title = {AI-Based Hybrid Artificial Neural Network and Particle Swarm Optimization Model for Energy Demand Forecasting},
             url = {https://www.scopus.com/inward/record.uri?eid=2-s2.0-105016859871&doi=10.1109\%2fICEPE65965.2025.11139390&partnerID=40&md5=7fb59d00fa060848bc1ad9db731b79a2},
        keywords = {Demand side management; Electric load forecasting; Electric power plant loads; Energy management; Errors; Mean square error; Neural networks; Artificial neural network; Demand-side; Energy consumption forecasting; Energy-consumption; Load profile analyze; Load profiles; Neural-networks; Particle swarm; Particle swarm optimization; Profile analysis; Swarm optimization; Particle swarm optimization (PSO)},
          author = {Felix, Franciecya and Jamahori, Hanis Farhah and Isa, Siti Salwa Mat},
            isbn = {979-833159706-1},
        abstract = {Energy consumption forecasting plays a crucial role in power system planning, load management, and energy optimization. Traditional forecasting models, such as Artificial Neural Networks (ANN), often suffer from convergence to local minima and non-optimal parameter selection, leading to reduced prediction accuracy. To address these limitations, this study proposes the hybridization of Particle Swarm Optimization (PSO) with ANN to enhance the model's forecasting performance by developing an accurate demand prediction model. Three years of energy consumption collected from January 2022 to December 2024 is used to predict future energy demand. ANN-PSO is proposed to improve forecasting accuracy by utilizing PSO's optimization capability to fine-tune the ANN model. The methodology includes data preprocessing, ANN configuration, and PSO-based parameter optimization. After training, the optimized ANN-PSO model forecasts future energy consumption, which is further analyzed through daily 24-hour load profiles. Three indices are used as fitness functions to measure the ANN's performance matrices: Mean Square Error (MSE), Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE). The simulation results demonstrate the superiority of ANN-PSO over conventional ANN. The ANN-PSO model achieves an MSE of 0.0209, RMSE of 0.1447, and MAPE of 17.09, significantly outperforming the standalone ANN, with an MSE of 0.1728, RMSE of 0.4157, and MAPE of 48.19. {\copyright} 2025 IEEE.}
}