Battery Cell Capacitance Estimation Using SVM
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing battery management systems face challenges in accurately estimating battery cell capacity due to errors introduced by linearizing state and observation models, which also increase computational load.
Innovation Solution
The use of a Support Vector Machine (SVM) to learn a capacity function and adapt the state model, allowing for a linear state model and reducing the need for frequent recalculations by executing the supervised learning algorithm only when necessary, thereby improving estimation accuracy and reducing computational burden.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a nonlinear observation model is used for capacity estimation, then measurement precision is improved, but device complexity and computational load increase due to the need for linearization
Solution Approach 1:
The patent replaces the traditional mechanical/mathematical approach of linearizing nonlinear models with a neural network-based system. The neural network learns the nonlinear relationships directly from data, substituting the need for analytical linearization and its associated computational burden while maintaining estimation accuracy.
Solution Approach 2:
The patent transforms the problem from estimating capacity directly using nonlinear models to estimating state-of-charge using a linearized model with neural network-predicted parameters. The neural network outputs parameters (like OCV-SOC curve characteristics) that enable accurate estimation through simpler linear calculations.
2Adaptability or versatility
If the supervised learning algorithm is executed frequently to update the state model, then adaptability is improved, but productivity decreases due to increased computational burden
Solution Approach 1:
The patent implements periodic execution of the supervised learning algorithm rather than continuous execution. The algorithm is triggered at specific intervals or under specific conditions (e.g., when significant cell changes are detected), balancing model adaptability with computational efficiency by avoiding unnecessary frequent recalculations.
3Device complexity
If linearization of the state model is performed to simplify calculations, then device complexity is reduced, but measurement precision deteriorates due to linearization errors
Solution Approach 1:
The patent substitutes the linearization approximation with a neural network that captures nonlinear effects. The neural network compensates for the simplifications made in the linear state model, maintaining accuracy while enabling the use of computationally simpler linear calculations for real-time estimation.
Data Source
AI summary
Estimating cell capacitance includes learning a function that returns capacitance as a function of an earlier capacitance and a physical characteristic that is correlated with changes in that capacitance and using a supervised learning algorithm on a learning database that has values of the capacitance, values of that physical characteristic measured concurrently with the capacitance and a preceding value of the capacitance.