This paper takes the creation of user profiles, prediction of electricity demand, construction of an electricity service optimization model, and satisfaction of electricity user needs as its research approach. Using electricity big data technology, it obtains residential electricity consumption behavior data from aspects such as basic electricity consumption, equipment electricity consumption, advanced electricity consumption, and abnormal electricity consumption. Through quantitative analysis of the obtained user electricity consumption behavior data, it generates user behavior feature tags from aspects such as basic and behavioral characteristics. By combining the generated user behavior tags with the characteristics of changes in electricity consumption behavior data, the core content of user electricity consumption behavior profiles is derived, thereby achieving precise user profiling for residential users. Additionally, based on existing research, short-term and medium-to-long-term influencing factors are screened out, and the Attention-Bi-LSTM model is used for electricity demand forecasting. Y State Grid Power Marketing Unit was selected as the experimental subject, and the power user behavior characteristics were calculated and analyzed. The proposed model was used to predict power user demand. The proposed prediction model not only fits the original data curve well but also maintains the prediction error within the range of [-5000, 6000], demonstrating high-precision prediction performance.
