Battery Cell Remaining-Life Prediction Through Partial Cycling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting the remaining useful life of battery cells are time-consuming, costly, and require extensive cycling, leading to low throughput and inefficient production processes.
Innovation Solution
A machine learning-based system that uses a small stimulus current and partial cycling to gather data for predicting battery degradation, employing a computational model trained with ground truth data from a large number of cells, allowing for fast and accurate predictions of remaining useful life.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional cycling methods are used to test battery cells, then the remaining useful life can be predicted, but the testing process becomes time-consuming and low throughput
Solution Approach 1:
The patent applies partial action by using a reduced cycling protocol that performs only a subset of the full cycling tests required by traditional methods. Instead of completing the entire cycling life test, the system uses a limited number of cycles (e.g., 10-50 cycles) combined with machine learning models to predict remaining useful life, thereby reducing testing time while maintaining acceptable prediction accuracy.
Solution Approach 2:
The patent introduces machine learning models as an intermediary between the cycling test data and the remaining useful life prediction. The ML models process the limited cycling data and produce predictions without requiring complete cycling tests, acting as a mediator that enables fast predictions from partial test data.
2Measurement precision
If extensive cycling is performed to determine cell health, then prediction accuracy improves, but testing time and hardware complexity increase
Solution Approach 1:
The patent applies preliminary action by pre-training machine learning models on extensive cycling data from many cells before actual production testing. This pre-training allows the models to make accurate predictions during production without requiring time-consuming tests, as the heavy computational work is done beforehand during model training rather than during cell testing.
Solution Approach 2:
The patent uses virtual copies of the physical cells in the form of digital twins represented by machine learning models. Instead of physically testing each cell for its entire life, the system creates a virtual model that can predict cell behavior and remaining useful life based on limited physical test data, effectively copying the testing process in silico.
3Ease of manufacture
If fixed cycling protocols with constant current are used, then the testing process is simple, but the method is limited to short lab experiments and cannot scale to long cycling studies
Solution Approach 1:
The patent applies dynamics by replacing fixed cycling protocols with adaptive cycling protocols that can adjust parameters based on cell responses and test objectives. The system dynamically modifies cycling conditions during testing to optimize data collection for machine learning models, enabling scalable long-term studies while maintaining protocol flexibility rather than rigidity.
Data Source
AI summary
A system and method for predicting a remaining useful life of a cell of a battery is described. The system include a circuit adapted to provide a stimulus signal to the cell; a processor; a tangible, non-transitory computer readable medium that stores instructions, which when executed by the processor, cause the processor to: cycle a plurality of target cells for a test set of cycles; determine a set of features from data from reference cells; determine parameters of a computational model of the remaining useful life of the cell; apply a computational model using the determined set of feature values; and provide a prediction of the remaining useful life of target cells.