Battery State Estimation Calibration With Minimal Input Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional battery state estimation methods, particularly for lithium-ion batteries, face challenges in handling cell-to-cell variation and nonlinear electrochemical impedance changes due to aging, requiring impractical amounts of calibration data and risking overfitting, especially when using machine learning models.
Innovation Solution
A model-input-based calibration method that uses minimal calibration measurements as input features to enhance battery state estimation, employing multivariable polynomial regression and neural networks, with techniques like linear least squares and iterative training to adjust model parameters, and deploying a linearized model for wider calibration condition tolerance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional calibration methods using large amounts of measured data are employed, then estimation accuracy can be improved, but the practicality and computational workload increase significantly
Solution Approach 1:
The patent changes the approach from using large amounts of calibration data to using minimal calibration data combined with physics-based models. The model incorporates electrochemical impedance parameters and temperature relationships to estimate battery state without requiring extensive calibration measurements, thereby reducing computational workload while maintaining accuracy.
Solution Approach 2:
The patent introduces an intermediary physics-based model that mediates between the minimal calibration data and the battery state estimation. This model uses electrochemical impedance spectroscopy parameters and temperature relationships to bridge the gap, allowing accurate estimation without direct extensive measurement of all battery states.
2Measurement precision
If machine learning models are used for battery state estimation, then estimation accuracy can be improved, but the risk of overfitting increases when using limited calibration data
Solution Approach 1:
The patent replaces pure machine learning approaches with a physics-based model that incorporates electrochemical impedance relationships. This substitution reduces overfitting risk by grounding the model in physical principles rather than purely statistical patterns, allowing accurate estimation even with minimal calibration data.
Solution Approach 2:
The patent changes the model's reliance on data-driven parameters to physics-based parameters such as electrochemical impedance spectroscopy measurements and temperature relationships. This parameter transformation reduces the model's sensitivity to calibration data quantity, thereby reducing overfitting risk while maintaining estimation accuracy.
3Measurement precision
If extensive calibration data is collected to handle cell-to-cell variation, then estimation accuracy across different cells can be improved, but the time and resources required increase
Solution Approach 1:
The patent changes the calibration approach from extensive multi-cell calibration data collection to minimal calibration data combined with physics-based models. The model uses electrochemical impedance parameters that can be measured with minimal data, reducing calibration time while maintaining accuracy across different cells through physical principle generalization.
Solution Approach 2:
The patent segments the calibration process into minimal essential measurements rather than requiring comprehensive calibration across all cells. By identifying and measuring only the critical electrochemical impedance parameters, the system achieves cell-to-cell variation handling without extensive calibration time.
Data Source
AI summary
Techniques for estimating a battery state of a battery include receiving a present battery measurement associated with an unknown battery state and a calibration measurement associated with a known calibration condition. An input set is generated based on the present battery measurement and the calibration measurement. The input set is inputted into a battery state model to receive an output of the battery state model. The output includes a present battery state estimation for the unknown battery state.