Electric Load Forecasting and Management in Smart Grids Using Optimized Long Short-Term Memory Network: A Real-World Evaluation

OpenAlex İngilizce

To enhance the efficiency and stability of modern smart grids, accurate short-term electricity demand forecasting is essential.The objective of this study is to present the use of Long Short-Term Memory (LSTM) networks to predict electrical load based on high-resolution, real-world data from the Belgian Elia grid, sampled at 15-minute intervals.The methodology includes data preprocessing, temporal feature extraction, sequence generation, and model optimization.Exploratory data analysis highlights important consumption patterns and seasonal variations.The LSTM model effectively captures both short-term fluctuations and long-term dependencies, achieving an RMSE of 119.41 MW, a MAPE of 1.30%, and an R² score of 0.992 on the test set.Compared to alternative forecasting approaches, including more complex hybrid architectures, the LSTM model demonstrates superior accuracy and generalization capability.For instance, compared with ARIMA-LSTM models that reported a MAPE of 2.83% and CNN-LSTM models with 2.72%, the proposed model achieves markedly better performance.These findings support the integration of LSTM-based forecasting systems into smart grid operations for real-time energy management.