Learning operando impedance function for battery health with aging-aware equivalent circuit model

Zihao Zhou,Antti Aitio,David Howey

2024-07-14

Abstract:The wide usage of Lithium-ion batteries (LIBs) requires a deep understanding about battery health. Estimation of battery state-of-health (SOH) is a crucial but yet still challenging task. Pure model-based methods may suffer from inaccuracy and instability of parameter estimations in lifelong aging. While pure data-driven methods rely heavily on quality and quantity of training set, causing lack of generality when extrapolating into unseen cases. In this paper, we propose an aging-aware equivalent circuit model (ECM), which combines model-based and data-driven techniques for SOH estimation. Gaussian process (GP) regression is incorporated in ECM to modelling parameters dependency on operating condition and aging time. The state space formulation of GP is used to enable a computationally efficient co-estimation framework of both parameters and states. Samples from two Open datasets are used to validate model performance, which gives accurate estimation for capacity and impedance. The learnt impedance function can be lined to the shape change of open circuit voltage (OCV) versus SOC curve and thus providing a way to further estimate changes of differential voltage (dV/dQ) curves.

Systems and Control

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the challenges in the state - of - health (SOH) estimation of lithium - ion batteries (LIBs). Specifically:

1. **Limitations of pure model - based methods**: Pure model - driven methods have deficiencies in the accuracy and stability of parameter estimation during the battery aging process.
2. **Limitations of pure data - driven methods**: Pure data - driven methods are highly dependent on the quality and quantity of the training set, resulting in a lack of generalization ability when dealing with unseen situations.
3. **Joint estimation of capacity and impedance**: Existing methods usually only focus on capacity decay and ignore the impact of impedance increase on battery health. The impedance value is relatively easy to estimate from different types of cycle data, so it can provide more available labels.

To solve these problems, the paper proposes an aging - aware equivalent circuit model (ECM) that combines model - driven and data - driven techniques and uses Gaussian process regression (GP) to model the dependence of parameters on operating conditions and aging time. Through the GP in state - space representation, this model can achieve efficient joint estimation of parameters and states.

### Main contributions:
- **Combination of model - and data - driven methods**: By combining model - driven and data - driven methods, the accuracy and generalization ability of battery SOH estimation are improved.
- **Application of Gaussian process regression**: Gaussian process regression is used to model the dependence of parameters on operating conditions and aging time, thereby more accurately capturing the changes during the battery aging process.
- **Joint estimation framework**: Through the Extended Kalman Filter (EKF), the joint estimation of parameters and states is achieved, improving the computational efficiency of the model.
- **Verification and analysis**: The performance of the model is verified using two public data sets. The results show that the model can accurately estimate the capacity and impedance of the battery and can further analyze the battery degradation situation.

### Key technical points:
- **State - space representation of the Gaussian process**: The Gaussian process is transformed into a state - space form, so that the computational complexity of the model grows linearly, which is suitable for large - scale data processing.
- **Improvement of the equivalent circuit model**: By introducing the Gaussian process to model the capacity and impedance of the battery, it can better reflect the changes of the battery under different operating conditions and aging times.
- **Joint estimation framework**: Through the EKF, the joint estimation of the battery state and Gaussian process parameters is achieved, ensuring the accuracy and stability of the model.

### Conclusion:
The model proposed in this paper can not only accurately estimate the capacity and impedance of the battery but also provide rich battery aging information, providing a new way for further analysis of battery degradation.