Research On Key Technologies For Lithium-ion Battery Second Use

Lithium- ion batteries retired from electric vehicles still have high potential to be reused in other applications if their functional components are effective and their appearance is in good condition. Therefore, these retired lithium- ion batteries without damage may be considered expanding their useful life for second use, which is to reduce their cost and is conductive to energy conservation and emissions reduction. The retired batteries have been separated from battery management system, in other words they are offline, and their residual capacity and performance are different from each other even in the same battery pack, so it is necessary to revaluate their performance. In this paper, an effective battery performance test was designed, which is to estimate state of health(SOH) of the retired batteries and form a novel method for further performance evaluation, even when they are under complicated test conditions and hostile environment. The main research contents are as follows:1. Based on the analysis of working characteristics of lithium- ion batteries, a battery test bench was established and the battery basic performance tests have been carried out yet. A comparative study on topology of the commonly used equivalent circuit models was performed and the RC equivalent circuit model is considered as the most suitable one to describe the lithium- ion battery dynamic working characteristics. The offline modeling method was then introduced for setup of the RC model and the identified parameters were extracted from the data of HPPC test, and the effectiveness of the established RC model was verified finally. However, the test data is always accompanied by random measurement error that would result in uncertainty error of the state estimation of lithium- ion battery. So the quantitative analysis of the uncertainty error effects was performed on the identified internal resistance of lithium-ion battery.2. For the lithium- ion batteries dynamic working states affected by external environment factors and load changes, the online modeling method was proposed. IV However, there are some issues in the dynamic modeling proc ess, such as data saturation, model adaptability of dynamic working characteristics, colored noise and so on. Therefore, the variable forgetting factor least square identification method and bias compensation least square identification method were put forward to solve these issues, and the effectiveness and reliability of the proposed adaptive identification methods were verified by the experimental results. The existing open circuit voltage(OCV) tests for lithium- ion battery commonly includes three kinds of test methods and the relevant test profiles are summarized as constant current charge and/or discharge with long-time rest test(CCLRT), constant current charge and discharge with short-time rest test(CCSRT) and low current charge and discharge test( LCT). Eight typical OCV test profiles were designed. Comparing the modeling performance based on RC model with different OCV curves, CCSRT profile was verified the most suitable test method in practice due to its equivalent modeling performance and less test duration time.3. Because the Kalman filters(KF) estimation method has more complex matrix operations and complicated algorithm, a discrete-time sliding mode observer(DSMO) estimation method was proposed and the stability of the first-order DSMO and the second-order DSMO has been verified in detail. Additionally, the observability of RC models has been analyzed. The experimental results show that EKF, the first-order DSMO and second-order DSMO are practically effective estimation methods, and second-order DSMO can reduce the chattering phenomena caused by abrupt changes of model parameters with better robustness, higher estimation accuracy. Through setting up the battery pack test experimental bench, the battery module measurement unit circuit was designed and the DSMO algorithm was carried out in microprocessor by half execution time of EKF algorithm.4. Due to the analysis of commonly used battery health indicators(HIs) test and health feature extraction method, the battery health test schedule was planned to select internal resistance as the objective HI. The three HIs, average internal resistance, minimum internal resistance and resistance-SOC curve, and relevant health features were obtained from health test data. Based on the three HIs, the battery health models were established and their effectiveness has been validated by experimental data finally.5. According to the difference of test circumstance in real conditions, the working characteristics and related problems of retired batteries for second use were specifically analyzed. Considering the limited test time, absence of important data and intermittent discharging process, a suitable test profile for performance evaluation of retired batteries was designed and the health feature extraction method was proposed based on the identifiable I-ARX model. However, the proposed health models are still linear and there is individual difference of these retired batteries. Therefore, a decision method using multi- model fusion technology was proposed to solve the issue by BP neural network, which is a commonly used data fusion algorithm, and its effectiveness was verified by experimental data. Finally, a SOC estimation method suitable for retired batteries was also proposed based on the SOC health model and the novel method for second use battery performance evaluation was given.6. The experimental platform for lithium- ion battery pack was built, and its hardware circuits and software were designed and debugged together. Based on the designed platform, the proposed SOC estimation method, the performance evaluation test profile and decision method of retired batteries were verified effectively.