Autoencoder Battery Data Analysis for Odd-Cell Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for diagnosing and predicting the state or lifetime of energy storage devices are inaccurate due to variations in material properties and manufacturing differences among individual devices, especially when new or differently charged devices are integrated into a group, leading to increased errors in large-scale systems.
Innovation Solution
A data processing device and method that uses an autoencoder to determine the 'oddity' of energy storage devices by training a determination model on measured data, which can be retrained based on environmental conditions and usage time, and applies smoothing processing to prevent erroneous determinations, allowing for accurate diagnosis and prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If measured data from all energy storage devices in a group is used for diagnosis and prediction, then the system can handle large-scale energy storage systems, but the accuracy of diagnosis and prediction deteriorates due to oddity among devices
Solution Approach 1:
The patent extracts and identifies odd energy storage devices from the group using an autoencoder-based determination model. By detecting devices with abnormal patterns (oddity) and separating them from the normal group, the system maintains high diagnostic accuracy while managing large-scale systems. The odd devices are identified through reconstruction error analysis and can be handled separately or excluded from group-based predictions.
Solution Approach 2:
The patent segments the group of energy storage devices into normal devices and odd devices based on their operational patterns and characteristics. This segmentation allows the system to apply different processing approaches to different subsets, maintaining overall system scalability while ensuring accurate diagnosis for each segment.
2Measurement precision
If a determination model is trained on measured data to identify odd energy storage devices, then the accuracy of identifying odd devices improves, but the complexity of the system increases
Solution Approach 1:
The patent employs an autoencoder-based determination model that performs self-learning to automatically identify odd energy storage devices. The model trains itself on measured data without requiring manual feature engineering or complex external algorithms, achieving high determination accuracy while keeping the system relatively simple through self-service learning.
3Adaptability or versatility
If new energy storage devices with different charge-discharge history are added to a group, then the system capacity and versatility increase, but the oddity among devices increases leading to reduced prediction accuracy
Solution Approach 1:
The patent implements a dynamic determination model that can be retrained periodically with new measured data from the group. As new energy storage devices are added with different characteristics, the model adapts by incorporating recent data, maintaining its ability to accurately identify odd devices despite changes in system composition and device diversity.
4Reliability
If smoothing processing is applied to measured data to prevent erroneous determination, then the reliability of oddity determination improves, but the time required for processing increases
Solution Approach 1:
The patent applies smoothing processing selectively and适度 (moderately) to the measured data. Rather than excessive smoothing that would oversimplify the data, the patent uses just enough smoothing to reduce noise and prevent erroneous determination while preserving the essential characteristics needed for accurate oddity detection, thus balancing reliability and processing efficiency.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Provided are a data processing device, a data processing method and a computer program for improving the accuracy of diagnoses, estimations and predictions regarding power storage elements on the basis of measurement data associated with the power storage elements. The data processing device processes measurement data for power storage elements. The data processing device comprises: a storage unit storing a determination model using an autoencoder that has been trained such that when measurement data for each of the power storage elements or for each power storage element group obtained by grouping a plurality of the power storage elements is input, the measurement data is reproduced; and a specification unit that specifies the measurement data for power storage elements that are different in the measurement data for each of the power storage elements or each of the power storage element groups on the basis of the error between the reproduced data that is output when the measurement data for each of the power storage elements or each of the power storage element groups is input to the determination model and the measurement data itself.