Lithium Battery Voltage Anomaly Detection With Time-Series Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing battery management systems fail to accurately identify anomalies in voltage acquisition of adjacent cells, leading to false alarms and missed faults, which can compromise safety and efficiency.
Innovation Solution
A lithium battery data acquisition anomaly identification method using time series analysis and anomaly detection, involving data preprocessing, extraction of voltage time series features, calculation of mean and variance, and N-sigma method to identify outlying points, followed by time series analysis to pinpoint anomalous voltage acquisition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If voltage data is monitored in real-time using conventional methods, then the battery management system can detect faults, but false alarms and missed faults occur due to inability to identify adjacent cell acquisition anomalies
Solution Approach 1:
The patent segments the voltage time series data into different components (trend component, seasonal component, residual component) using decomposition methods. This segmentation allows the system to analyze different aspects of the data separately, improving the ability to distinguish between normal variations and actual anomalies in adjacent cell voltage acquisitions
Solution Approach 2:
The patent introduces a new dimension of analysis by calculating the difference between adjacent cell voltages and analyzing this difference series separately. This dimensional transformation enables the system to detect acquisition anomalies that are not apparent in the original voltage data, thereby improving anomaly identification accuracy without compromising fault detection reliability
2Reliability
If the BMS monitors all cell voltages continuously, then safety can be ensured, but storage limitations prevent effective anomaly identification
Solution Approach 1:
The patent extracts only the essential features from the raw voltage data, specifically the difference series between adjacent cells and the decomposed components. By storing and analyzing only these extracted features rather than the complete raw data, the system maintains safety monitoring capability while significantly reducing storage requirements
Solution Approach 2:
The patent transforms the raw voltage data into different parameter representations (difference values, decomposed components) that capture the essential information needed for anomaly detection. This parameter transformation allows the system to monitor all cell voltages continuously for safety while using compact representations for anomaly identification, thus managing storage limitations effectively
3Ease of operation
If single rule anomaly detection is used, then the system is simple to implement, but it cannot identify acquisition anomalies in various cases
Solution Approach 1:
The patent develops a multi-functional anomaly detection framework that combines multiple detection methods (decomposition-based detection, difference series detection, statistical threshold detection) into a unified system. This universal approach can handle various types of anomalies (acquisition failures, voltage jumps, sensor faults) while maintaining relatively simple implementation through standardized processing steps
Solution Approach 2:
The patent creates a composite anomaly detection strategy that combines multiple detection approaches and analysis methods. By integrating different detection techniques (time series decomposition, difference analysis, statistical methods) into a composite framework, the system achieves broad anomaly identification coverage while keeping each individual component relatively simple and easy to implement
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention relates to a lithium battery data acquisition anomaly identification method and system, a storage medium and a device. The method includes: performing data preprocessing on voltage data of a battery management system to form derived voltage time series features; extracting voltage time series features of adjacent cells in a battery pack according to the derived voltage time series features, and calculating the mean and variance of voltages of all cells, to form new voltage time series features; performing anomaly identification on the new voltage time series features to find out an outlying time series point of a cell voltage; and performing time series analysis on voltages of adjacent cells according to the outlying time series point of a cell voltage, and identifying a position where voltage acquisition of adjacent cells is anomalous. According to the present invention, the accuracy of anomaly identification of voltage acquisition of adjacent cells can be improved, and the method can be applied to the fields of lithium-ion battery state acquisition, monitoring and early warning.