An electric carbon data anomaly detection method, device, medium and equipment

CN122241546APending Publication Date: 2026-06-19STATE GRID ZHEJIANG ELECTRIC POWER CO LTD

Patent Information

Authority / Receiving Office: CN · China
Patent Type: Applications(China)
Current Assignee / Owner: STATE GRID ZHEJIANG ELECTRIC POWER CO LTD
Filing Date: 2026-05-25
Publication Date: 2026-06-19

AI Technical Summary

Technical Problem

Existing technologies for detecting anomalies in carbon dioxide data suffer from low detection accuracy, high false alarm and false negative rates, inability to accurately identify multiple types of anomalies, and poor adaptability to complex data scenarios.

Method used

A multi-scale constrained anomaly detection model is adopted, combined with a deep support vector description network. Through multi-timescale embedding offset modeling and anomaly distance constraint, an adaptive anomaly score is generated to identify short-term, medium-term, and long-term anomalies in the carbon data. An auxiliary classifier is used to identify the anomaly type.

Benefits of technology

It enables accurate and efficient identification of carbon dioxide data, improves the accuracy and interpretability of detection, adapts to power data anomalies in different time dimensions, and provides reliable real-time monitoring and anomaly early warning support.

✦ Generated by Eureka AI based on patent content.

Smart Images

Figure CN122241546A_ABST

Patent Text Reader

Abstract

This invention discloses a method, apparatus, medium, and device for anomaly detection in electrocarbon data. Belonging to the field of anomaly detection, this application first acquires multidimensional time-series electrocarbon data and then inputs it into a preset multi-scale constrained anomaly detection model. The model's temporal modeling and reconstruction submodule is responsible for extracting temporal embedding features and generating a data reconstruction sequence, then calculating the residual between the multidimensional time-series electrocarbon data and the reconstruction sequence to obtain the data reconstruction error. The multi-scale embedding calculation submodule calculates the multi-scale embedding offset score of the temporal embedding based on multiple preset time scales, and concatenates these scores with the data reconstruction error to form a joint feature. A preset deep support vector description network is used to perform high-dimensional mapping on the joint feature, and the distance from the high-dimensional mapping result to the center of a preset minimum closed sphere is calculated, finally generating an adaptive anomaly score. This application effectively solves the problem that existing technologies cannot accurately and efficiently detect anomalies in electrocarbon data.

Need to check novelty before this filing date? Find Prior Art