Efficient Energy Consumption and Demand Response Using Deep Learning-Based Load Forecasting for Green Grid

This paper presents a deep learning-based load forecasting framework for efficient energy consumption and demand response in green grid computing. We employ Long Short-Term Memory (LSTM) networks with Recurrent Neural Network (RNN) cells to predict energy consumption patterns by analysing historical load data and external factors, including weather conditions, user activity, and temporal patterns. The proposed model achieves high prediction accuracy with Mean Absolute Error (MAE) of 5.2 kW, Root Mean Square Error (RMSE) of 7.1 kW, Mean Absolute Percentage Error (MAPE) of 1.3%, and coefficient of determination (R²) of 0.95, outperforming baseline methods including ARIMA (46% improvement in MAE) and traditional neural networks (29% improvement). Statistical validation using paired t-tests (p < 0.001) and the Diebold-Mariano test confirms significance. These results demonstrate the model’s effectiveness in enabling real-time decision-making for data centres and grid operators, with potential applications in day-ahead energy procurement, demand response optimisation, and renewable energy integration for medium-scale data centre facilities.