Real‐time identification of partnership for a new generation of vehicles battery model parameters based on the model reference adaptive system

Abstract

Partnership for a new generation of vehicles (PNGV) model is a conventional battery equivalent circuit model (ECM). However, identifying the best parameters for this model is a challenge. In this study, the PNGV model is transformed into a directly identifiable difference equation to identify its parameters. Subsequently, the model reference adaptive system (MRAS) is used to realize the real-time identification of the model parameters. The identification accuracy of the MRAS is found to be superior to that of the recursive extended least square algorithm. For a single hybrid pulse power characterization (HPPC), the PNGV model identified by the MRAS can achieve a high-precision terminal voltage estimation. For lithium iron phosphate, lithium titanate, and nickel-metal hydride batteries, the root mean square errors are 0.024, 0.048, and 0.020 V, respectively. Besides, the real-time state of charge (SOC) estimation can be realized by the identified open-circuit voltage (OCV). The average errors of the three batteries are only −0.02, −0.01, and −0.01, respectively. Since the PNGV model has a capacity of describing the change of OCV with the current accumulation effect, the model is only suitable for simulating single HPPC or positive-negative pulses with equal amplitude and not for other current pulses. This is a major drawback of the PNGV model. The real-time PNGV model parameters identification method proposed in this study can provide a solid foundation for various state estimation of a battery. Novelty Statement• Transformation of the PNGV model into a difference equation that can be directly identified.• Real-time identification of the PNGV model parameters via the MRAS.• The identified OCV realized the real-time SOC estimation and discussed the deficiencies of the PNGV model.

Improved fractional‐order hysteresis‐equivalent circuit modeling for the online adaptive high‐precision state of charge prediction of urban‐electric‐bus lithium‐ion batteries

“…Pulse tests are performed at different SOC points, with the minimum RMSE of the terminal voltage as the fitness function, as shown in Equation (13).…”

“…12 There are many algorithms for estimating SOC based on ECM. 13 Fu et al 14 used the weighted multi-innovation cubature Kalman filter (KF) method to estimate SOC based on the Thevenin model, which improved the accuracy significantly. 15 Considering the effects of temperature and C-rates based on the Thevenin model and the SOC of lithium batteries was estimated using the unscented Kalman filter (UKF).…”

“…The partnership for a new generation of vehicles (PNGV) model has higher estimation accuracy, but parameter identification is more complex 12 . There are many algorithms for estimating SOC based on ECM 13 . Fu et al 14 used the weighted multi‐innovation cubature Kalman filter (KF) method to estimate SOC based on the Thevenin model, which improved the accuracy significantly 15 .…”

On‐road estimation of state of charge of lithium‐ion battery by extended and dual extended Kalman filter considering sensor bias

“…The average function of charging and discharging must be determined to obtain the accurate function, 35 as shown in Figure 5. The steps for low current OCV test to obtain the SOC‐OCV relationship are as follows: Step 1 : Bring the battery voltage to its lower cutoff value (ie, 0.0 SOC) and rest for 2 hours. Step 2 : Charge the battery at a 0.05°C rate from full discharge state to full charge state so that SOC goes from 0.0 to 1.0. Step 3 : Rest the battery for 2 hours. Step 4 : Discharge the battery at a 0.05°C rate from full charge state to full discharge state so that SOC goes from 1.0 to 0.0. The OCV is represented as a ninth‐order polynomial function of SOC 36 and the coefficients are obtained by the Matlab curve fitting toolbox in (). …”

“…In some circumstances, this relationship is formulated using the CC charging component, whereas, in others, it is formulated using the CC discharging part. The average function of charging and discharging must be determined to obtain the accurate function, 35 The OCV is represented as a ninth-order polynomial function of SOC 36 and the coefficients are obtained by the Matlab curve fitting toolbox in (25).…”