JZUSA. Lithium-ion battery state-of-charge estimation based on deconstructed equivalent circuit at different. open-circuit voltage relaxation times

Transcription

1 Journal of Zhejiang University-SCIENCE A Cite this as: Xi-ming CHENG, Li-guang YAO, Michael PECHT, Lithium-ion battery state-of-charge estimation based on deconstructed equivalent circuit at different opencircuit voltage relaxation times. Journal of Zhejiang University-SCIENCE A (Applied Physics & Engineering), 18(4): Lithium-ion battery state-of-charge estimation based on deconstructed equivalent circuit at different Key words: open-circuit voltage relaxation times Lithium-ion batteries, Open-circuit voltage, State-of-charge, Recursive least squares, Extended Kalman filters

2 Background The circuit-based model for lithium-ion battery is more popular than the electrochemical model for SOC estimation in real time. However, the OCV measurements are quite timeconsuming and their values have effects on the circuit model-based SOC estimation. Until now, there has been little literature discussing an appropriate relaxation period explored to ensure accuracy of the circuit model-based SOC estimation for a short SOC-OCV test time.

3 Method Influence of OCV measurement values at different relaxation time and temperature on two methods for battery SOC estimation including the OCV and EKF are investigated. A first-order RC circuit and one RC loop deconstructed from the second-order RC circuit model are used to estimate positive model parameters by the two-stage RLS for EKF-based SOC estimation. The two-stage RLS and EKF are combined to estimate circuit-model parameters and battery SOC for cylindrical lithium-ion cells.

4 Circuit Deconstruction for Two-Stage RLS (a) R 1 u 1 C 1 R 0 i r1 i c1 u rrc i u ocv u t R 1 the first-rc circuit model u 1 u 2 C 1 R 0 i r1 i c1 R 2 C 2 u ocv i r2 i c2 i u t the second-rc circuit model (b) R 2 i r2 i c2 i C 2 u 2 the RC loop

5 Combination Algorithm of RLS and EKF Initialization i k ˆθ 1,0 Θ 1,0 Stage 1: RLS Plant y 1,k T ˆθ 2,0 h1, ˆθ 1 ˆθ k Θ 2 2,0 Stage 2: RLS ˆX 0 R k Q k P 0 ˆθ 1 Parameter input Model & EKF-based SOC estimation D G k-1 Φ k-1 H k M k H k T Φ k P k-1 Φ k T y 2,k X k ˆX k-1 M k P k-1 z -1 z -1 ˆX k P k Z = k u t,k u ocv M k H kt (H k M k H kt R k ) -1 f(soc,t) Multi-stage RLS Parameter Identification F k K k K k (I-K k H k )M k SOC State output Block diagram of two-stage RLS - EKF SOC estimation algorithm Z k

6 Experiments Three CBAK 2.0 Ah power Li-ion cells in series were placed in a temperature chamber and discharged/charged by a Digatron battery tester BNT ME. Three types of battery charged/discharged tests: capacity, OCV, and drive cycle tests. The battery charging or discharging capacities were limited up to the nominal value in the OCV and drive cycle tests. During the charging OCV test, cells were rested for 2 h after every 10 % nominal capacity pulse charge at 0.5C above 10 C or 0.2C below 10 C. After they were fully charged and rested for 2 h, they were discharged at 20 C.

7 JZ U SA Results for OCV Measurements re = 100% ( yt y2h ) y2h Where re denotes the relative error voltage of OCV measurements. yt denotes OCV measurements at different relaxation periods. y2h denotes the measurement value at the 2-h relaxation period.

8 JZ U SA Results for SOC Estimation Estimated SOC and errors impacted by different temperature OCVs for battery FTP profiles at 0 /20 /40 C: (a) real and estimated SOC curves; (b) SOC estimation errors.

9 Conclusions Using the OCV-based SOC estimation method, different relaxation times were required to get SOC estimation errors less than 5% at 0 C, 5 min for the discharging process, and 10 min for the charging process. Less errors, higher temperature. The proposed adaptive algorithm shows that the SOC estimation accuracy improves as the OCV relaxation periods increase and that the SOC estimation errors can be reduced compared to the OCV-based SOC estimation. Under the FTP profiles, the model-based SOC estimation errors can show less than 3% even if the OCV test time is reduced down to a 5-min relaxation period for the SOC- OCV relationships at the experimental temperatures.