基于贝叶斯记忆模块增强的用电数据异常检测方法及装置
By introducing a Bayesian memory module and a feature extraction framework constrained by a joint reconstruction-regression task in electricity consumption data anomaly detection, the problem of insufficient robustness and generalization ability of existing methods in complex scenarios is solved. This enables the capture of long-term dependencies and the suppression of contaminated data, thereby improving the accuracy and adaptability of anomaly detection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING UNIV OF POSTS & TELECOMM
- Filing Date
- 2025-11-27
- Publication Date
- 2026-07-17
AI Technical Summary
Existing methods for detecting anomalies in electricity consumption data struggle to capture the contextual dependencies and complex nonlinear relationships of time series data when faced with complex and ever-changing real-world scenarios. Furthermore, they are unable to effectively learn long-term dependencies that extend beyond the time window size, resulting in insufficient robustness and generalization ability. In particular, the unexpected generalization problem of the model to anomalies is severe when contaminated data is present.
We employ a Bayesian memory module-based enhancement method to construct a feature extraction framework constrained by the reconstruction-regression joint task by inserting positional information within and between windows. By combining the reconstruction autoencoder and the Bayesian memory module, we suppress the influence of contaminated data, extract temporal features, and perform multi-objective optimization.
It improves the robustness and accuracy of electricity consumption data anomaly detection, effectively captures long-term dependencies in complex scenarios, enhances the model's adaptability to fluctuations in input features, suppresses interference from contaminated data, and achieves more efficient anomaly detection.
Smart Images
Figure CN121659145B_ABST