TL;DR
This paper introduces a novel filter-based method for accurately estimating the State of Health (SOH) of lithium-ion batteries in real-time, even under varying temperatures and dynamic operating conditions. By dynamically adjusting battery model parameters, including capacity, their approach significantly reduces estimation errors, offering a more reliable solution for electric vehicle battery management.
What Problem Does This Paper Address?
Lithium-ion batteries are critical for electric vehicles (EVs), but their performance degrades over time (aging). Accurately monitoring their "State of Health" (SOH) is crucial for safety, performance optimization, and predicting remaining lifespan. However, estimating SOH is challenging because battery aging is influenced by complex internal electrochemical processes and external factors like temperature, which cause battery parameters to change dynamically. Traditional methods often struggle to adapt to these variations, leading to inaccurate SOH estimations. This paper aims to bridge this gap by developing a robust online SOH estimation method that explicitly accounts for these dynamic parameter changes, particularly focusing on how temperature affects battery capacity.
Methods
The core of this work lies in a filter-based co-estimation method that integrates parameter identification and state estimation.
- Reconstructed Equivalent Circuit Model: They start by reconstructing a second-order hybrid equivalent circuit model (ECM) of the lithium-ion battery. This model is designed to represent both the steady-state and transient behaviors of the battery. A key innovation here is reformulating the terminal voltage derivative to express it as a product of measurable variables, non-measurable variables, and an unknown parameter matrix.
- Adaptive Parameter Update Rate: To handle the time-varying nature of battery parameters (including capacity), they design an adaptive update rate for these parameters. This is achieved by incorporating signal filters into the parameter estimation process, allowing for online updates.
- Nonlinear Observer for SOH Estimation: Based on the adaptive model, a nonlinear observer is developed. This observer integrates filters to estimate the battery's SOH. A significant aspect is that the capacity is dynamically adjusted within the parameter update rate, directly influencing the SOH estimation.
- Comparison with Existing Methods: The proposed method is benchmarked against established techniques like the Dual Extended Kalman Filter (DEKF) and Dual Unscented Kalman Filter (DUKF) to demonstrate its superiority.
Datasets
The paper utilizes experimental data collected from a lithium-ion battery (INR18650 MH1) under controlled laboratory conditions.
- Experimental Setup: The experiments were conducted using a NEWARE CT-8008 battery testing system and a RIUKAI programmable constant temperature and humidity test chamber.
- Testing Conditions: The batteries were subjected to:
- Dynamic stress testing (DST) and Federal City Driving Schedule (FUDS) at 25 °C.
- Basic performance tests (capacity and rapid test procedures) at various temperatures (0 °C, 10 °C, 25 °C, and 45 °C).
- The DST and FUDS tests were repeated at 0 °C, 10 °C, and 45 °C to evaluate the algorithm's robustness across different temperature ranges.
Key Insights and Ideas
- Capacity as a Dynamic Parameter: Recognizing that battery capacity isn't static but changes with aging and operating conditions is crucial. The paper's approach of dynamically adjusting capacity within the parameter estimation framework is a key differentiator.
- Temperature's Impact on Model Accuracy: The study highlights how temperature variations significantly affect battery model parameters. Their method's ability to adapt to these changes, even without explicit temperature compensation in the core observer, leads to improved accuracy.
- Co-estimation Advantage: By jointly estimating parameters and states, the method avoids the complexity of separate estimation steps and potentially reduces error propagation.
- Filter-Based Approach for Robustness: The use of filters in both parameter adaptation and state estimation contributes to the algorithm's robustness against noise and dynamic changes.
Key Results
The proposed filter-based method demonstrated excellent performance across various conditions:
- SOH Estimation Accuracy: The mean absolute error (MAE) for SOH estimation was consistently below 0.5%, and the root mean square error (RMSE) was below 0.2% across different temperatures (0 °C, 10 °C, 25 °C, and 45 °C).
- SOC Estimation Accuracy: For SOC estimation, MAE was kept below 2% and RMSE below 1%.
- Comparison: The filter-based observer showed better estimation accuracy and faster dynamic response compared to DEKF and DUKF, especially under varying temperature conditions.
- Temperature Adaptability: While temperature variations did cause some increase in SOH error at extreme low and high temperatures, the proposed method's error increase was significantly less than that of DEKF and DUKF, indicating good environmental adaptability.
Limitations and Future Work
- Temperature Impact Nuances: The paper acknowledges that while their method accounts for temperature's effect on model parameters, the temperature itself doesn't change linearly during operation. The segmented fitting curves for SOC and Voc might not fully capture these dynamic temperature effects.
- Future Optimization: The authors suggest that future work could involve exploring more advanced algorithms, such as machine learning and deep learning techniques, to further enhance estimation accuracy and robustness. This could lead to even more sophisticated battery management systems.
Summary updated