Accurate residential load forecasting is crucial for the stable operation of the energy internet, which plays a significant role in advancing sustainable development. As the construction of the energy internet progresses, the proportion of residential electricity consumption in end-use energy consumption is increasing, the peak load on the grid is growing year on year, and seasonal and regional peak power supply tensions, mainly for household electricity consumption, grow into common problems across countries. Residential load forecasting can assist utility companies in determining effective electricity pricing structures and demand response operations, thereby improving renewable energy utilization efficiency and reducing the share of thermal power generation. However, due to the randomness and uncertainty of user load data, forecasting residential load remains challenging. According to prior research, the accuracy of residential load forecasting using machine learning and deep learning methods still has room for improvement. This paper proposes an improved load-forecasting model based on a time-localized attention (TLA) mechanism integrated with LSTM, named TLA-LSTM. The model is composed of a full-text regression network, a date-attention network, and a time-point attention network. The full-text regression network consists of a traditional LSTM, while the date-attention and time-point attention networks are based on a local attention model constructed with CNN and LSTM. Experimental results on real-world datasets show that compared to standard LSTM models, the proposed method improves R² by 14.2%, reduces MSE by 15.2%, and decreases RMSE by 8.5%. These enhancements demonstrate the robustness and efficiency of the TLA-LSTM model in load forecasting tasks, effectively addressing the limitations of traditional LSTM models in focusing on specific dates and time-points in user load data.

With the proliferation of distributed energy resources, advanced metering infrastructure, and advanced communication technologies, the grid is transforming into a flexible, intelligent, and collaborative system. Short-term electric load forecasting for individual residential customers is playing an increasingly important role in the operation and planning of the future grid. Predicting the electrical load of individual households is more challenging with higher uncertainty and volatility at the household level compared to the total electrical load at the feeder and regional levels. The previous research results show that the accuracy of forecasting using machine learning and a single deep learning model is far from adequate and there is still room for improvement.