Battery State Classification Using Cycle-Level Data Labelling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing battery management systems face challenges in securing a large amount of learning data for classification models due to the time-consuming process of charging and discharging cycles, which limits the performance of the learned model, especially with a limited number of batteries.
Innovation Solution
A battery management apparatus and method that extracts feature values at every charging and discharging cycle, judges the state of the battery, sets labels based on these states, and learns a classification model to quickly accumulate learning data, even from a limited number of batteries, by considering the degradation and potential temporary restoration of battery states.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If charging and discharging cycles are performed to secure training data, then learning data can be obtained, but the process takes a lot of time
Solution Approach 1:
The patent creates virtual copies of battery data through data generation unit 104, which synthesizes additional training data by modifying existing battery measurement data. This copying approach allows the system to expand the training dataset without performing additional physical charging and discharging cycles, thereby resolving the contradiction between obtaining more learning data and reducing time consumption.
Solution Approach 2:
The patent performs preliminary data processing and augmentation before the actual model training phase. By pre-generating and storing virtual training data samples through various transformations (noise addition, parameter modification, etc.), the system prepares a comprehensive dataset in advance, eliminating the need for time-consuming real-time data collection during training.
2Duration of action of stationary object
If battery life expectancy is increased, then battery performance is improved, but the time required for charging and discharging cycle increases proportionally
Solution Approach 1:
The patent generates virtual training data that simulates long-term battery behavior patterns without requiring actual long-duration charging and discharging cycles. By copying and transforming existing battery data to represent various stages of battery life, the system can train models on extended operational scenarios while maintaining the same physical battery lifespan.
Solution Approach 2:
The patent performs preliminary data synthesis to create training samples representing various battery life stages and degradation patterns. This preliminary action allows the system to prepare comprehensive training data for long-term battery performance analysis without actually subjecting batteries to extended charging and discharging cycles, thus resolving the time-life expectancy contradiction.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A battery management apparatus according to an embodiment of the present disclosure includes: a feature value extracting unit configured to extract a feature value of a learning battery at every charging and discharging cycle; a first state judging unit configured to judge a state of the learning battery at every charging and discharging cycle, based on the feature value extracted by the feature value extracting unit and a criterion value preset to correspond to the feature value; a labelling unit configured to set a label for the feature value of the learning battery at every charging and discharging cycle, based on the state of the learning battery judged by the first state judging unit; a model learning unit configured to learn a classification model for judging a state of an analysis battery based on the feature value for which the label is set by the labelling unit; and a second state judging unit configured to judge the state of the analysis battery based on the classification model learned by the model learning unit.