The battery in electric vehicles is a key focus of battery health maintenance. The Battery Management System (BMS) maintains the optimal state of the battery by evaluating its State of Health (SOH). Accurately identifying SOH can determine battery replacement time, avoid battery failure, and extend its service life. This article aims to improve the performance of BMS by identifying SOH parameters. Based on the Thevenin battery model, key parameters such as R0, Rp, and Cp are obtained. Two adaptive algorithms, Coulomb counting and open circuit voltage, are used to complete parameter identification. The results of the two algorithms in terms of error, mean absolute error (MAE), root mean square error (RMSE), and final SOH value are compared. The research focuses on obtaining estimation error data and reliable BMS performance information. The results show that the Coulomb counting method has a smaller error in SOH estimation than the open circuit voltage method, with an error of 1.770%. The final SOH value is 17.33%, and the Thevenin battery model has a modeling error of 0.0451% for the battery.
1. Introduction
Electric Vehicle Battery and Battery Management System (BMS): In electric vehicles, the battery is the primary energy source, providing power to the engine and other systems. Unlike traditional cars, electric vehicle batteries have relatively small capacity and voltage, and are usually packaged in battery modules. The battery system consists of multiple batteries, which are managed by BMS. Its functions include optimizing the battery working system, involving two key parameters: state of charge (SOC) and state of health (SOH). SOC is the ratio of remaining capacity to total capacity, while SOH is the comparison value between current performance and new battery performance, which cannot be directly measured and needs to be estimated.
Research background and related methods: State of Health (SOH) can quantify battery performance and lifespan. Quality degradation, changes in internal resistance and capacity parameters may occur during battery use. Identifying SOH parameters helps determine the actual condition of the battery, recommend replacement times, and extend battery life. There are currently multiple methods for estimating State of Health (SOH) or State of Charge (SOC), but there are few methods that simultaneously identify both and generate appropriate parameters to reduce the computational burden on BMS. The algorithm for monitoring battery parameters needs to adapt to parameter changes and estimate battery condition. The methods can be divided into three categories, including spectral impedance method, circuit model equation method, and electrochemical impedance model method.
Review of related work: Multiple methods have been commonly used in previous research to identify battery parameters. Coulombic counting (CC) and open circuit voltage (OCV) methods are widely used in electric vehicle BMS, each with its own advantages and disadvantages. The CC method estimates SOH by monitoring the in and out capacity of the battery, taking into account power loss during the charging cycle, and can also provide relevant information through voltage recovery; The OCV method can be considered as a balanced voltage after the battery has fully rested, and the State of Health (SOH) is estimated by considering the BMS battery parameter conditions.
The focus of this study is to identify accurate SOH parameters to extend battery life. A battery model-based method is used to evaluate and identify SOH parameters. The Thevenin battery model is used to identify R0, Rp, and Cp parameters through an adaptive algorithm (Recursive Least Squares, RLS). Based on the evaluation results, accurate SOH estimates are obtained to reduce computational burden.
Research contribution: The results of testing battery parameters provide reasonable estimates and small error rates for evaluating BMS system performance. The Coulomb counting method is convenient for calculating battery capacity, and the maximum power of the battery decreases with the increase of charge and discharge cycles. The relative error of Thevenin battery model is less than 2%. In terms of SOH estimation accuracy, CC method is superior to RLS, and CC method can estimate battery terminal voltage and SOC, while OCV method can only estimate battery parameters.
2. Battery management system
Battery components (functions and composition of BMS): BMS regulates the battery system composed of hundreds or thousands of batteries in electric vehicles, and has important functions such as monitoring, estimating parameters, protecting, providing reports, and balancing batteries. Its main functions include protecting the battery from damage, operating the battery within appropriate voltage and temperature ranges, and maintaining the battery to operate at parameters that meet system requirements such as SOC, SOH, and SOF. BMS consists of sensors, actuators, and controllers, with inputs including sensor signals such as current, voltage, temperature, and pedals, and outputs including modules for thermal management, balance, safety management, charging indication, fault alarm, and communication. The BMS software includes multiple functional modules such as battery parameter detection, estimation, and fault diagnosis. Battery voltage measurement, parameter estimation, balancing, and fault diagnosis are the main issues of BMS, among which battery voltage measurement faces difficulties such as voltage differences caused by battery series connection and high-precision requirements.
Battery modeling: This article determines the State of Health (SOH) parameters through battery modeling, and converts the input battery voltage, current, and temperature parameters into SOH to obtain accurate estimates. Using the Thevenin battery model, the voltage transient response of the battery polarization process is described by selecting the internal resistance and capacitance parameters of the battery. The mathematical equations of the battery model and the calculation methods for related parameters (Voc, R0, Rp, and Cp) are provided, which are obtained through the RLS algorithm and applied to the Thevenin battery model.
3. Determine health condition parameters
The importance and methods of identifying health status parameters: Accurate SOH parameters are crucial for BMS performance. This study uses Coulomb counting as an adaptive algorithm to identify these parameters to obtain SOH initialization values and evaluate BMS performance. The Thevenin battery model is used to determine the battery model parameters and OCV-SOC function. The specific process involves inputting current to the battery model, analyzing terminal voltage data, converting from time domain to SOC domain, and curve fitting to obtain the OCV-SOC function. The parameter identification process is repeated until the SOH estimation is reasonable and the error rate is small.
OCV-SOC function: Based on the Thevenin battery model, OCV (SOC) is a source voltage parameter obtained by testing the voltage of the battery without a connected load and the voltage before connecting the battery pack. The SOC OCV curve is estimated using constant load test data and fitted with a twelfth order polynomial. The tenth order polynomial has the highest accuracy in estimating Voc and the smallest root mean square error (RMSE), which has a significant impact on the accuracy of SOC and OCV functions.
R0, Rp, and Cp parameters: The Thevenin battery model requires OCV at SOC as the source voltage, which is obtained through pulse testing. R0 is an internal resistance with a value greater than other resistances. Due to the sampling period issue, it is difficult to capture small data changes. The relationship between R0 and SOC was obtained through second-order polynomial curve fitting, with an average R0 value of 0.027735 Ω. R0, Rp, and Cp provide input data for voltage and current pulse testing and obtain output parameter values.
Experimental result
By analyzing the State of Health (SOH) parameters monitored by the battery, BMS performance is achieved, and physical parameter data such as terminal voltage and inlet/outlet current of the battery are obtained. Based on battery modeling, parameter data is identified and used for battery status monitoring and protection systems. The SOH estimation method includes measuring the changes in battery resistance and capacity, respectively using Ohm's law and Coulomb counting method, and substituting the OCV value into the OCV-SOC relationship equation to obtain the SOC and SOH values.
The static discharge test was conducted, and the results showed that the CC algorithm obtained the SOH change by multiplying the current value by time, while the OCV algorithm obtained the SOH value by using the terminal voltage or OCV value of the battery model. The SOH change curves of the two algorithms were similar. The test also obtained battery parameter identification results, and the battery relaxation characteristics can be used for parameter identification. The faster the test cycle, the more accurate the SOH estimation. The CC algorithm is superior to the OCV algorithm in SOH initialization, which can better understand the internal resistance of the battery and simultaneously estimate the terminal voltage Vt, SOC, and SOH of the battery, with an estimation error of less than 2%.
From the error data of identifying SOH parameters, the mean square error (MSE) of CC algorithm is 0.0111, the final SOH value is 17.33%, the error percentage is 1.770%, and the root mean square error (RMSE) is 0.0132
Research result discussion: The impact of internal battery resistance on CC and OCV algorithms is similar, and CC algorithm can better understand internal resistance with smaller errors. The CC algorithm can successfully estimate the terminal voltage Vt, SOC, and SOH of the battery simultaneously, with an estimation error of less than 2%. In the discharge test, the CC algorithm is more accurate than the OCV algorithm in SOH initialization, with an estimated mean square error (MSE) of 1.770% for the CC algorithm and 3.256% for the OCV algorithm. These results provide reference for parameter identification in BMS evaluation.
4. Summary
The performance evaluation results of BMS based on SOH parameter identification show that the Coulomb counting algorithm has better estimation results, with a SOH estimation error of 1.770% and a final SOH value of 17.33%. The modeling error of Thevenin battery model for batteries is 0.0451%. In terms of the accuracy of SOH estimation using two methods (Coulomb counting and open circuit voltage), Coulomb counting has higher accuracy. In addition, adaptive algorithms based on battery modeling can estimate the terminal voltage and SOH of the battery.