Battery Aging Prediction Using Sigmoidal Rate Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies struggle to accurately predict the aging of rechargeable batteries, leading to inefficiencies in managing battery performance and safety due to the lack of effective methods for determining capacity fade and power loss over the service life.
Innovation Solution
The use of sigmoidal rate expressions (SRE) and machine-learning models, combined with Monte Carlo simulations, to analyze historical and measured battery data, allowing for the prediction of future battery aging states based on a small subset of early test data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional battery aging prediction methods are used, then comprehensive battery performance data can be obtained, but testing costs and time are excessively high
Solution Approach 1:
The patent applies preliminary action by using early-cycle battery data (first subset of cycles) to train machine learning models and establish degradation patterns before the battery reaches end-of-life. This allows the system to predict future aging trajectories without requiring complete lifecycle testing, thereby reducing testing time while maintaining prediction accuracy.
Solution Approach 2:
The patent uses copying by creating virtual representations of battery aging through machine learning models that replicate the degradation behavior observed in early test data. These models generate synthetic aging predictions that mirror real battery performance, eliminating the need for prolonged physical testing while preserving measurement precision.
2Measurement precision
If comprehensive battery testing is conducted to ensure accurate aging predictions, then prediction accuracy improves, but testing costs increase
Solution Approach 1:
The patent applies taking out by extracting only the essential early-cycle data features needed for training machine learning models, rather than conducting comprehensive full-lifecycle testing. This selective extraction of critical information maintains prediction accuracy while significantly reducing the energy and resource costs associated with extensive battery testing.
Solution Approach 2:
The patent uses parameter changes by transforming early-cycle battery parameters (capacity, voltage, temperature) into predictive aging parameters through machine learning model training. This parameter transformation allows accurate aging predictions to be derived from limited early-test data, reducing the need for costly comprehensive testing while maintaining measurement precision.
3Loss of energy
If early test data is used for predictions, then testing costs are reduced, but prediction accuracy may deteriorate
Solution Approach 1:
The patent applies feedback by continuously refining machine learning models using early-cycle test data, where model predictions are compared against actual battery performance measurements. This iterative feedback process improves prediction accuracy over time, allowing reliable aging predictions to be achieved using only early test data without requiring comprehensive lifecycle testing.
Solution Approach 2:
The patent uses preliminary action by performing extensive model training and validation during the early testing phase, establishing robust prediction capabilities before full-scale deployment. This preliminary model development ensures that accurate predictions can be made from early data alone, eliminating the need for costly additional testing while maintaining high measurement precision.
Data Source
AI summary
Embodiments disclosed herein include methods, systems and/or devices configured to generate historical parameters of a sigmoidal rate expression based on the sigmoidal rate expression and measured data of a battery. The embodiments may further be configured to predict future parameters of the sigmoidal rate expression based on the sigmoidal rate expression and the historical parameters. The embodiments may further be configured to predict an aging state of the battery based on the sigmoidal rate expression and the future parameters. Additional embodiments are directed to methods, systems, and/or devices configured to synthesize training data based on measured battery data, a sigmoidal rate expression, and ranges for parameters of the sigmoidal rate expression. The additional embodiments may further be configured to train a machine-learning model using the synthesized training data.