Battery Cell Anomaly Detection Using Time-Series Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing battery management systems cannot trace the source of faults in battery packs due to differences among cells, leading to reduced power performance and inability to predict battery faults effectively.
Innovation Solution
A battery cell anomaly determination method using an ARIMA model to analyze time-series physical data, preprocess the data, and calculate difference features to identify anomalous cells based on a preset threshold, enabling early fault prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Power
If battery cells are connected in series to meet power demand, then the power performance of the battery pack is improved, but differences among battery cells arise due to uncontrollable factors during manufacture and use, leading to reduced power performance
Solution Approach 1:
The patent applies preliminary action by establishing a cell anomaly determination model before actual fault occurrence. The model is trained using historical data and preprocessing techniques (data cleaning, stationarity test, white noise test) to recognize anomaly patterns early. This allows the system to predict and identify potential faults in battery cells before they manifest as actual failures, enabling proactive maintenance and preventing power performance degradation.
2Reliability
If fault determination is based on collected information from BMS and infotainment system, then fault detection capability is improved, but the ability to trace to the source of the fault through big data is lost
Solution Approach 1:
The patent applies segmentation by dividing the battery pack into individual battery cells as independent analysis units. Each cell is monitored separately with its own time-series physical data (voltage, current, temperature). The anomaly determination model processes each cell's data independently, allowing precise identification of which specific cell is anomalous. This segmented approach preserves traceability to individual cells while maintaining overall system fault detection capability.
Solution Approach 2:
The patent introduces an intermediary - the cell anomaly determination model - that processes raw collected information from BMS and infotainment systems. This model acts as a mediator between data collection and fault determination, transforming raw data into meaningful anomaly predictions. The model uses preprocessing and ARIMA analysis to extract actionable insights, enabling both fault detection and source tracing through the big data cloud platform.
3Measurement precision
If real-time monitoring of all battery cells is implemented, then fault detection accuracy is improved, but system complexity and data processing requirements increase
Solution Approach 1:
The patent applies self-service by enabling the anomaly determination model to autonomously process and analyze cell data without requiring complex external intervention. The model automatically performs preprocessing (data cleaning, stationarity test, white noise test), selects appropriate ARIMA models, determines model parameters, and generates anomaly predictions. This self-service capability reduces the need for complex manual analysis systems while maintaining high fault detection accuracy through automated intelligent processing.
Data Source
Figure 1
AI summary
The present invention relates to a battery cell anomaly determination method and system, a processing device, and a storage medium. The method comprises: obtaining time-series of real physical data of each cell in a battery pack to be detected, performing preprocessing, inputting the preprocessed time-series of real physical data into a pre-constructed cell anomaly determination model, and calculating to obtain physical data predicted values of a plurality of frames of each cell in the battery pack to be detected; according to the physical data predicted values of the plurality of frames of each cell and physical data true values of corresponding frames, calculating to obtain difference features of the plurality of frames of each cell in the battery pack to be detected; integrating and summarizing the difference features of the plurality of frames of each cell to obtain an anomaly sorting result of each cell; and determining the serial number of an anomalous cell in the battery pack to be detected according to the anomaly sorting result of each cell and a preset threshold. According to the method, the fault of a battery cell can be predicted in advance, and the method can be widely used in the technical field of battery fault detection.