An optimized system for predicting energy usage in smart grids using temporal fusion transformer and Aquila optimizer.

This research presents an optimized system for predicting energy usage in smart grids by integrating the Temporal Fusion Transformer (TFT) with the Aquila Optimizer (AO). The study addresses the growing need for accurate energy consumption forecasts in smart grids, driven by the increasing adoption of renewable energy and real-time data collection through smart meters. The TFT model leverages self-attention mechanisms to handle complex time-series data, improving forecasting accuracy across various time horizons. To enhance predictive performance, the Aquila Optimizer, a nature-inspired algorithm, is employed to fine-tune critical hyperparameters, ensuring optimal model convergence and performance. The proposed AO-TFT model is evaluated against traditional models like LSTM and CNN-BiLSTM, demonstrating superior accuracy, lower RMSE, and faster computation times. The research also analyses the impact of various factors, including building types, weather conditions, and load variations on energy prediction. The proposed AO-TFT model achieved a significantly lower RMSE of 0.48 and MAE of 0.31, demonstrating superior accuracy compared to traditional models. Future work is suggested to explore hybrid optimization techniques and real-time adaptive models for dynamic grid management.