Multi-kernel correntropy based extended Kalman filtering for state-of-charge estimation

Area of Science:

• * Electrical Engineering
• * Control Systems
• * Signal Processing

Background:

• * Real-time battery management relies on accurate state of charge (SOC) estimation.
• * The standard extended Kalman filter (EKF) can produce biased SOC estimates due to sensor faults, bias, and noise.
• * Complex non-Gaussian disturbances challenge traditional EKF performance in battery systems.

Purpose of the Study:

• * To propose a robust extended Kalman filter (EKF) for accurate SOC estimation under abnormal operating conditions.
• * To enhance the EKF's resilience against non-Gaussian noise and sensor-related disturbances.
• * To develop a computationally efficient method for practical battery pack state estimation.

Main Methods:

• * Formulation of a batch-mode regression integrating process and measurement uncertainties.
• * Application of the maximum multi-kernel correntropy (MMKC) criterion to mitigate abnormal condition influences.
• * Utilization of an optimization method for MMKC parameter determination and fixed-point iteration for state estimation.
• * Update of the posterior error covariance matrix using the total influence function for improved robustness.
• * Implementation of a novel filtering scheme to reduce computational complexity.

Main Results:

• * The proposed MMKC-EKF demonstrates superior accuracy and robustness in SOC estimation compared to standard methods.
• * Effective suppression of influences from complex non-Gaussian disturbances and sensor anomalies.
• * Validation through extensive simulations under various noise conditions (Gaussian and non-Gaussian).

Conclusions:

• * The MMKC-EKF provides a robust and accurate solution for online SOC estimation in battery management systems.
• * The method effectively handles sensor faults, bias, noise, and non-Gaussian disturbances.
• * The reduced computational complexity makes the MMKC-EKF suitable for practical battery pack state estimation applications.