In this work, a novel Digital Twin model using attention mechanism integrated with LSTM to forecast the future power load of a specific user is developed. The power load prediction research is done in detail by taking into account important factors such as temperature, humidity, and the price of electricity. Therefore, LSTM networks are adopted for deep learning of the historical power load data, while the attention mechanism is used to assign weights to the significance of various factors that affect the power load and make better predictions of the future power load. The results of the presented experiment show the improved prediction accuracy and stability of the model in comparison with the existing power load prediction models. The present study also introduces a new and effective method for the power load forecasting.