Online Identification with Reliability Criterion and State of Charge Estimation Based on a Fuzzy Adaptive Extended Kalman Filter for Lithium-Ion Batteries

Zhongwei Deng, Lin Yang, Yishan Cai, Hao DengEnergies 2016 · DOI

TL;DR

This paper tackles the challenge of accurately estimating the State of Charge (SOC) in lithium-ion batteries by proposing a novel Fuzzy Adaptive Extended Kalman Filter (FAEKF). The FAEKF combines online parameter identification with a reliability criterion and a fuzzy logic-based approach to compensate for inherent model errors, leading to more robust and accurate SOC estimations across various operating conditions.

What Problem Does This Paper Address?

Accurate State of Charge (SOC) estimation is crucial for the safe and efficient operation of battery management systems (BMS) in electric vehicles and other applications. While the Extended Kalman Filter (EKF) is a popular method, its performance is highly sensitive to the accuracy of the battery model and the noise covariance parameters. Existing methods often struggle with:

  • Model Inaccuracies: Simplified battery models (like the Thevenin equivalent circuit model) inherently have "intrinsic model errors," especially under dynamic conditions with high or rapidly changing currents. These errors can significantly degrade SOC estimation accuracy.
  • Parameter Drift: Battery model parameters can change over time due to aging and varying operating conditions (temperature, current load). Relying on static, offline-identified parameters leads to inaccuracies.
  • Reliability of Online Identification: While online parameter identification methods can adapt to changing conditions, the identified parameters themselves might not always be reliable, especially during complex driving cycles.

This paper aims to address these limitations by developing a more robust and adaptive SOC estimation framework.

Methods

The proposed approach integrates several key components:

  • Online Parameter Identification: The authors employ a Recursive Least Squares (RLS) algorithm, specifically the Recursive Extended Least Squares (RELS) algorithm, to identify the parameters of the Thevenin equivalent circuit model in real-time. This allows the model to adapt to changing battery conditions.
  • Parameter Reliability Criterion: To ensure the identified parameters are trustworthy, a criterion based on the difference between the offline and online Open Circuit Voltage (OCV) is introduced. If this difference exceeds a threshold (ΔmV), the online parameters are deemed unreliable and not used for SOC estimation. This leverages the fact that OCV-SOC curves are relatively invariant with battery aging.
  • Fuzzy Model for Model Error Compensation: A fuzzy logic-based model is developed to compensate for the intrinsic model error. This model uses battery current, current change, and SOC as input variables to dynamically adjust the measurement noise covariance (R). By increasing R when model errors are expected to be high (e.g., large currents), the algorithm gives more weight to the Ampere-Hour integral method, and vice-versa.
  • Fuzzy Adaptive Extended Kalman Filter (FAEKF): The core algorithm is an Adaptive Extended Kalman Filter (AEKF) that incorporates the fuzzy model for updating R. The process noise covariance (Q) is estimated from the innovation sequence. This combination aims to provide a more accurate and resilient SOC estimation.

Datasets

The experiments were conducted using LiFePO4 batteries with a capacity of 10 Ah. The data was collected using a Digatron UBT 300-060 multifunctional battery tester. The paper utilizes:

  • Constant current discharge/charge profiles: To evaluate performance under steady-state conditions.
  • Pulse current profiles: To assess behavior under dynamic, short-term load changes.
  • Urban Dynamometer Driving Schedule (UDDS) cycle: A standard driving cycle used to simulate real-world vehicle operation and evaluate the algorithm's performance under complex, varying current conditions.

The paper does not introduce a new benchmark dataset but rather uses established testing methodologies.

Key Insights and Ideas

  • OCV Error as a Reliability Indicator: Using the OCV difference as a reliability criterion for online parameters is an elegant way to filter out potentially erroneous model updates, especially during transient or unusual operating conditions.
  • Fuzzy Logic for Model Error Compensation: The insight that intrinsic model errors are correlated with current magnitude and rate of change, and that these can be compensated by dynamically adjusting the measurement noise covariance (R), is a powerful concept. Fuzzy logic provides a flexible way to implement this compensation without requiring a precise analytical model of the error.
  • Synergy of Online Identification and Fuzzy Compensation: The paper highlights the benefit of combining a "positive strategy" (online identification to adapt parameters) with a "negative strategy" (fuzzy model to compensate for inherent model limitations). This dual approach leads to a more robust system than either strategy alone.
  • OCV Plateau Handling: The fuzzy model's membership functions are designed to account for the OCV plateau region in LiFePO4 batteries, which is a known challenge for SOC estimation methods relying solely on OCV.

Key Results

The proposed FAEKF algorithm demonstrated improved SOC estimation accuracy and convergence compared to a standard AEKF without the reliability criterion (AEKF(NO-R)) and an AEKF with online identification but without fuzzy compensation (AEKF(Online)).

  • Pulse Test: FAEKF achieved a mean error (ME) of 0.012, significantly lower than AEKF(Online)'s ME of 0.022.
  • Constant Current Test: Under 1C discharge, FAEKF's ME was 0.028, compared to 0.057 for AEKF(Online). The FAEKF also showed better performance under higher currents (2C).
  • UDDS Cycle Test: The FAEKF consistently converged to the real SOC, while AEKF(Online) maintained a noticeable error. For initial SOC=0.7, FAEKF's ME was 0.018, versus 0.029 for AEKF(Online).
  • Reliability Criterion: The reliability criterion was shown to prevent parameter divergence and improve SOC estimation accuracy, particularly under challenging conditions like pulse currents.

Limitations and Future Work

  • Parameter Calibration: The parameter 'A' in the reliability criterion and the upper bound of R in the fuzzy model require calibration for different battery types. The paper suggests this can be done through experiments.
  • Generalizability: While the authors state the method can be applied to other battery types with modifications, further validation on different chemistries would be beneficial.
  • Computational Complexity: Although the paper aims for integration into BMS, the computational load of the FAEKF, especially with online identification and fuzzy logic, might still be a consideration for resource-constrained embedded systems.

Future work could involve exploring more advanced fuzzy logic structures, investigating other online identification algorithms, and validating the approach on a wider range of battery chemistries and operating conditions.

Summary updated

Abstract

In the field of state of charge (SOC) estimation, the Kalman filter has been widely used for many years, although its performance strongly depends on the accuracy of the battery model as well as the noise covariance. The Kalman gain determines the confidence coefficient of the battery model by adjusting the weight of open circuit voltage (OCV) correction, and has a strong correlation with the measurement noise covariance (R). In this paper, the online identification method is applied to acquire the real model parameters under different operation conditions. A criterion based on the OCV error is proposed to evaluate the reliability of online parameters. Besides, the equivalent circuit model produces an intrinsic model error which is dependent on the load current, and the property that a high battery current or a large current change induces a large model error can be observed. Based on the above prior knowledge, a fuzzy model is established to compensate the model error through updating R. Combining the positive strategy (i.e., online identification) and negative strategy (i.e., fuzzy model), a more reliable and robust SOC estimation algorithm is proposed. The experiment results verify the proposed reliability criterion and SOC estimation method under various conditions for LiFePO4 batteries.
BibTeX
@inproceedings{deng2016online,
  author = {Zhongwei Deng and Lin Yang and Yishan Cai and Hao Deng},
  title = {{Online Identification with Reliability Criterion and State of Charge Estimation Based on a Fuzzy Adaptive Extended Kalman Filter for Lithium-Ion Batteries}},
  booktitle = {Energies},
  year = {2016},
  doi = {10.3390/en9060472},
}
Figure