Energy management depends heavily on the accuracy of load forecasting, both of which are critical to the stability of the power grid and the operational efficiency of the system. In this work, we propose a hybrid deep learning model that combines Temporal Convolutional Networks (TCNs), Bidirectional Gated Recurrent Units (Bi-GRU), Long Short-Term Memory (LSTM), and Transformer attention to address the complexities of load time series data. Through advanced feature engineering, lag feature generation, and hyperparameter tuning the model provides accurate forecasts for daily, weekly, and monthly periods. The empirical results showed an improvement in the accuracy (with the RMSE/ MAE drops), demonstrating that the model can capture both long-term and short-term dependencies.