SOH Estimation of BESS Using Operational Data Based on Adaptive UKF and AWTLS

Abstract:
Battery Energy Storage Systems (BESS) help in the maintenance of grid flexibility. The State of Health(SOH) of BESS is an important parameter to optimize their performance and lifetime. Direct capacity testing of these systems can result in revenue loss due to operational downtime. Data-driven techniques like neural network, fuzzy logic, support vector machines require large amounts of data to train. Mathematical-model-based algorithm considers fixed internal characteristics, process noise, and measurement noise of battery to model a Li-ion battery. This study is continuation of earlier work in which Unscented Kalman Filter (UKF) was used on historical operational data to estimate the State of Charge (SOC) and SOH was then estimated by using the Approximate Weighted Total Least Square (AWTLS). The objective of this study is to estimate the SOH of the battery with adaptive noise and battery internal characteristics to model battery aging process. To this end, UKF is used to estimate SOC based on adaptive process and measurement noise on operational historical data comprising of current and voltage data over long period of time. Noise values are updated based on error offsets. SOH is then estimated by using the AWTLS with a forgetting factor on differentiated SOC (dSOC). Adaptive UKF caters to the aging phenomena of the battery. AUKF estimates are improved over non-adaptive UKF estimates. The proposed method can be integrated into industrial BESS monitoring systems for real-time, data-efficient SOH tracking under diverse operational conditions.