Kalman filtering method for graph reasoning towards adaptive observation of industrial time series data

By incorporating the physical topology and temporal neighborhood information of the energy consumption metering system into the Kalman filtering method, a dynamic factor graph model is constructed, and the Kalman filter gain is adaptively adjusted. This solves the problems of anomaly identification and data quality management in complex industrial time-series data by traditional methods, and improves state estimation and data quality.

CN122293060APending Publication Date: 2026-06-26TAIJI COMPUTER CORPORATION LIMITED
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-02-12
Publication Date
2026-06-26

AI Technical Summary

Technical Problem

Traditional Kalman filtering methods are unable to effectively identify anomalies, quantify confidence, and integrate topological constraints when faced with complex industrial time-series data, resulting in inaccurate state estimation and poor quality of the input data.

Method used

By acquiring energy consumption measurements and predicted status, calculating deviations and evaluating single-point reliability, and combining time neighborhood information and the physical topology of the energy consumption metering system, a dynamic factor graph model is constructed for global inference, adaptively adjusting the Kalman filter gain, and performing data quality classification management.

Benefits of technology

It achieves adaptive anomaly identification, quantification of credibility, and fusion of topological constraints, thereby improving the accuracy of state estimation and the quality of the data entering the database.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122293060A_ABST
    Figure CN122293060A_ABST
Patent Text Reader

Abstract

This invention provides a Kalman filter method for graph inference based on adaptive observation of industrial time-series data. The method includes acquiring the current energy consumption measurement, the current predicted state, and the predicted observation; calculating the deviation between the energy consumption measurement and the predicted observation; fusing temporal neighborhood information to obtain a local comprehensive credibility factor; combining the physical topology of the energy consumption metering system with energy conservation constraints to obtain a topological consistency factor; constructing a dynamic factor graph model for global inference to obtain a global credibility factor; adaptively adjusting the Kalman filter observation update gain based on this global credibility factor to obtain the current state estimate; and performing data quality grading and differentiated data entry management based on the global credibility factor. This method achieves adaptive anomaly identification, credibility quantification, topological constraint fusion, and data quality management, improving the accuracy of state estimation and the quality of the data entered into the database.
Need to check novelty before this filing date? Find Prior Art