A two-stage electricity fee prediction method and device based on user portrait and multi-dimensional data feature fusion
By employing a two-stage electricity bill forecasting method based on user profiles and multi-dimensional data feature fusion, and by filtering weather, holiday, and comprehensive information feature values, combined with anomaly event correction, the problem of large computational load and accuracy fluctuation in electricity bill forecasting for ordinary residential users is solved, thereby improving the forecasting accuracy and practicality.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-04-02
- Publication Date
- 2026-07-03
AI Technical Summary
In predicting electricity costs for ordinary residential users, existing methods suffer from problems such as large amounts of computational data, scattered processing results, and fluctuating prediction accuracy for individual users, which affect the stability and practicality of the prediction results.
A two-stage electricity bill prediction method based on user profiles and multi-dimensional data feature fusion is adopted. By obtaining user profiles of target users, filtering feature values of weather information, holiday information and other comprehensive information, and inputting them into the prediction model to obtain the predicted daily electricity consumption, and calling the fitted correction relationship to correct when there are abnormal events, the electricity bill is finally calculated.
It reduces the amount of computational data required for predicting electricity bills for ordinary residential users, improves the accuracy and practicality of predictions, reduces the bias of prediction results caused by abnormal events, and enhances the credibility of prediction results.
Smart Images
Figure CN122335381A_ABST