Sectional type on-line SOC dynamic calibration method with forgetting factor, computer equipment and electric vehicle

Abstract

The invention relates to the technical field of battery management, and discloses a sectional type on-line SOC dynamic calibration method with a forgetting factor, computer equipment and an electric vehicle. According to the online SOC dynamic calibration method, a battery parameter table is constructed, online parameter identification is performed by adopting a least square method with a forgetting factor, the forgetting factor is dynamically adjusted through a segmented error threshold, numerical stability is ensured by combining overrun judgment, SOC is estimated and calibrated in real time by utilizing an extended Kalman filtering algorithm, accumulative errors are effectively inhibited, and the accuracy of SOC calibration is improved. The method significantly improves the estimation precision and system adaptability, and is more suitable for the actual operation condition of the automobile.

Description

Technical Field

[0001] This invention relates to the field of battery management technology, and more specifically to a segmented online SOC dynamic calibration method with a forgetting factor, a computer device, and an electric vehicle. Background Technology

[0002] In recent years, electric vehicles and related technologies have developed rapidly. Estimating the state of charge (SOC) of a battery is a crucial component of the battery management system. Accurate SOC estimation helps to fully utilize the power performance of the battery system, prevent overcharging and over-discharging, and ensure the battery's lifespan and safety during use.

[0003] Currently, commonly used SOC estimation methods include the ampere-hour integration method, the open-circuit voltage method, the neural network method, and the Kalman filter method. Among these, the ampere-hour integration method is low-cost and simple to implement, but it requires other methods to determine the initial SOC value, and due to current measurement errors, it can accumulate significant errors over long periods. The open-circuit voltage method uses the terminal voltage of the electric vehicle after a long period of inactivity as the open-circuit voltage, and determines the calibration value through the correspondence between the open-circuit voltage and SOC, making dynamic estimation difficult in actual operation. As a complex nonlinear system, the battery uses a neural network method to estimate SOC, achieving high accuracy, but its computational complexity and large data storage requirements make it difficult to implement on a microcontroller. The Kalman filter method estimates SOC based on the battery's equivalent circuit model, calculating the estimated SOC value through the system state equation, and then correcting the estimate based on the current voltage measurement value, achieving a minimum variance estimate of the system state. While suitable for nonlinear systems, it suffers from parameter identification errors and numerical instability. Summary of the Invention

[0004] To overcome the aforementioned technical problems, this invention provides a segmented online SOC dynamic calibration method with a forgetting factor. This online SOC dynamic calibration method constructs a battery parameter table and uses the least squares method with a forgetting factor for online parameter identification. It dynamically adjusts the forgetting factor through segmented error thresholds, combines over-limit judgment to ensure numerical stability, and uses the extended Kalman filter algorithm to estimate and calibrate SOC in real time, effectively suppressing accumulated errors, significantly improving estimation accuracy and system adaptability, and is more suitable for the actual operating conditions of automobiles.

[0005] The first aspect of this invention provides a segmented online SOC dynamic calibration method with a forgetting factor, the dynamic calibration method comprising: Construct a battery model parameter table; Battery model parameters are identified online using the least squares method with a forgetting factor. The EKF method was used to estimate and calibrate the SOC online.

[0006] Preferably, a battery model parameter table is constructed, including: HPPC experimental method was used to obtain current, voltage and time series data of battery under different operating conditions, and equivalent circuit model was established according to formula group (1). (1) in, for Battery terminal voltage at time, for Open-circuit voltage at any given time for The first polarization voltage at time 1, for The second polarization voltage at time , for The first polarization voltage at time 1, for The second polarization voltage at time , For ohmic internal resistance, For the first polarization resistor, This is the second polarization resistor. This is the first polarization capacitor. This is the second polarization capacitor. for Current at any moment for Current at any moment It is a natural constant. For time step; Determine the SOC breakpoint based on the changing trends of current and time; Based on the breakpoints, the data is divided into multiple data subsets, and multiple SOC intervals are determined. For each subset of data, the battery response curve is fitted, and the ohmic internal resistance parameter and polarization parameter are calculated and obtained. An nth-order polynomial fit is performed on the open-circuit voltage and the state of charge (SOC) to generate a battery model parameter table.

[0007] Preferably, battery response curve fitting is performed on each data subset to calculate and obtain ohmic internal resistance parameters and polarization parameters, including: Identify and extract key curve segments from the battery response curve; The ohmic resistance parameters are obtained by fitting the curve segment of the instantaneous voltage jump point in the impulse response using formula (2). (2) in, , , , This refers to the voltage point in the HPPC experiment. It is a pulse current; The relaxation curve segment is fitted using formula (3) to obtain the polarization parameters. (3) in, This refers to the battery terminal voltage. Open circuit voltage, For current, It is a natural constant. For time step, For the first polarization resistor, This is the second polarization resistor. This is the first polarization capacitor. This is the second polarization capacitor.

[0008] Preferably, the battery model parameters are identified online using the least squares method with a forgetting factor, including: Based on the equivalent circuit model, a continuous S-domain transfer function is constructed using formula (4). (4) in, Let be the transfer function in the S-domain. The first time constant, The second time constant, For Laplace variables; Based on the bilinear transformation, the discretized transfer function is constructed using formula (5). (5) in, It is a constant. Let Z be the variable for transformation; The vector of parameters to be estimated and the amount of input data are defined using formulas (6) and (7), respectively. (6) (7) - (8) in, for The vector of parameters to be estimated at time t. , , , , , It is a constant. for The amount of input data at any given time. for Output variables at time 10:00 for Output variables at time 10:00 for Output variables at time 10:00 for Input variables at time, for Input variables at time, for Input variables at time, for The output variable at time t is The difference between the open-circuit voltage and the terminal voltage at a given time; Linearization is performed using the least squares method, and the output variables are obtained according to formula (9). (9) in, for Output variables at time 10:00 for The transpose of the input data vector at time t. for The vector of parameters to be estimated at time t. for The prediction error at any given time; The estimated parameter vector is updated using a recursive algorithm based on formula (10). (10) in, for Prediction error at time, for The transpose of the input data vector at time t. for The input data vector at time t, for The estimated parameter vector at time step, for The estimated parameter vector at time step, for The Kalman gain matrix at time t. for The covariance matrix at time t, for The covariance matrix at time t, Forgetting factor parameters, It is the identity matrix; Different allowable errors are set according to different SOC intervals, and the forgetting factor parameters are dynamically adjusted in a segmented manner using formula (11). (11) in, for The forgetting factor parameter at time, The parameter is the minimum forgetting factor. The parameter for the maximum forgetting factor. The coefficient of variation, for Error scaling factor at time step The allowable error threshold; Based on the dynamic forgetting factor parameters, the constants in the discretized transfer function are identified, and the intermediate variables are obtained by formula group (12). The parameter over-limit processing method is used to process the over-limit values ​​in the intermediate variables. (12) in, , , , , Here, T is an intermediate variable, and T is the sampling time. Based on the intermediate variables, the parameters are transformed using formula (13) to obtain the battery model parameters. (13) in, The first time constant, This is the second time constant.

[0009] Preferably, a parameter over-limit handling method is used to handle over-limit values ​​within intermediate variables, including: Determine whether the value of the denominator of the intermediate variable is less than a preset first threshold; If the value of the denominator of the intermediate variable is less than a preset first threshold, obtain the value of the denominator of the intermediate variable at the previous time step. If the value of the denominator of the intermediate variable is greater than or equal to a preset first threshold, obtain the value of the denominator of the intermediate variable at the current time.

[0010] Preferably, the SOC is estimated and calibrated online using the EKF method, including: Determine whether the EKF calibration SOC condition has been triggered; When the EKF calibration SOC condition is triggered, the SOC is estimated and calibrated online. Determine whether the EKF termination calibration SOC condition has been triggered; If the EKF termination calibration SOC condition is not triggered, the steps of iteratively re-estimating and calibrating the online SOC are performed. If the EKF termination calibration SOC condition is triggered, output the current calibrated SOC.

[0011] Preferably, when the EKF calibration SOC condition is triggered, the SOC is estimated and calibrated online, including: Set the initial state and noise characteristics of the EKF algorithm, and initialize the state vector according to formula (14). (14) in, For state vectors, The initial state of charge, This is the first polarization voltage. This is the second polarization voltage. The state estimates the covariance matrix. The process noise covariance matrix is... To measure the noise covariance matrix; Formulas (15)-(16) are used to predict the state vector and error covariance matrix of the battery model at the current moment. (15) (16) in, for The prior estimate of the state vector at time t. for The state transition matrix at time t, for The optimal estimate of the updated state vector at time step [time]. for The input matrix at time step, for Measuring current at any given time for The predicted value of the error covariance matrix at time t. for The transpose of the state transition matrix at time t. The process noise covariance matrix; The predicted state vector is corrected using formula (17); (17) in, for The predicted value of the terminal voltage at time [time]. for The output matrix at time 10:00. It is a feedforward matrix; The innovation vector and gain matrix are obtained using formula (18). (18) in, For innovation vectors, for The input unit voltage value at any given time. for Gain matrix at time step for The transpose of the output matrix at time t. To measure the noise covariance matrix; Based on the innovation vector and gain matrix, the state vector and error covariance matrix are updated and calibrated using formulas (19)-(20) to obtain the optimal estimate at the current time. (19) (20) in, for The optimal estimate of the updated state vector at time step [time]. For the updated The error covariance matrix at time t.

[0012] A second aspect of the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the dynamic calibration method as described in any one of the preceding claims.

[0013] A third aspect of the present invention provides an electric vehicle, the electric vehicle including a battery system for performing the dynamic calibration method as described in any one of the preceding claims.

[0014] Through the above technical solution, this online SOC dynamic calibration method first constructs battery model parameter tables for different SOC ranges through experiments, establishing an accurate second-order RC equivalent circuit model. Then, it employs the least squares method with a forgetting factor for online parameter identification. The forgetting factor is dynamically adjusted through a piecewise error threshold to achieve adaptive updating of model parameters. An over-limit judgment mechanism ensures numerical stability. Based on the extended Kalman filter algorithm, the identified parameters are used for real-time estimation and calibration of SOC. Through state prediction, measurement update, and posterior estimation recursive calculation, the cumulative error of the ampere-hour integration method is effectively suppressed. The piecewise design enables the algorithm to adapt to the nonlinear characteristics of different SOC ranges of the battery, significantly improving parameter identification accuracy and operating condition adaptability. The online identification mechanism shortens the offline calibration time. The fusion of EKF and FFRLS effectively controls the SOC estimation error, enabling it to converge to within 3% even under complex dynamic operating conditions, significantly improving the estimation accuracy and reliability of the battery management system. Attached Figure Description

[0015] Figure 1This is a connection block diagram of a segmented online SOC dynamic calibration method with forgetting factor according to an embodiment of the present invention; Figure 2 This is a connection block diagram generated from the battery model parameter table of a segmented online SOC dynamic calibration method with forgetting factor according to an embodiment of the present invention. Figure 3 This is a connection block diagram for identifying battery model parameters in a segmented online SOC dynamic calibration method with forgetting factor according to an embodiment of the present invention. Figure 4 This is a connection block diagram of parameter over-limit handling in a segmented online SOC dynamic calibration method with forgetting factor according to an embodiment of the present invention. Figure 5 This is a connection block diagram illustrating the triggering and termination of online SOC calibration in a segmented online SOC dynamic calibration method with a forgetting factor according to an embodiment of the present invention. Figure 6 This is a connection block diagram of online SOC estimation and calibration in a segmented online SOC dynamic calibration method with forgetting factor according to an embodiment of the present invention. Detailed Implementation

[0016] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0017] like Figure 1 The diagram shown is a connection block diagram of a segmented online SOC dynamic calibration method with a forgetting factor according to an embodiment of the present invention; Figure 1 The online SOC dynamic calibration method includes: In step S10, a battery model parameter table is constructed; In step S11, the battery model parameters are identified online using the least squares method with a forgetting factor; In step S12, the SOC is estimated and calibrated online using the EKF method.

[0018] Step S10: Obtain battery dynamic response data under different SOC ranges, temperatures and rates through HPPC experiments. Based on the second-order RC equivalent circuit model, obtain accurate parameters such as ohmic internal resistance, polarization resistance, polarization capacitance and open circuit voltage through offline fitting, and establish a parameter lookup table. Step S11 dynamically adjusts the forgetting factor by introducing a piecewise error threshold. When the prediction error exceeds the allowable threshold of the current SOC interval, the forgetting factor is automatically reduced to enhance the response to new data; conversely, the forgetting factor is increased to maintain stability. Simultaneously, an over-limit judgment module is added to prevent computational divergence, achieving high-precision, adaptive online identification of model parameters.

[0019] Step S12 inputs the battery parameters obtained from online identification into an extended Kalman filter, and combines them with the real-time measured voltage and current values ​​to make an optimal estimate of the SOC state vector. Through the recursive process of state prediction, measurement update, and posterior estimation, the cumulative error of the ampere-hour integration method is effectively suppressed, and the optimal SOC estimate is output.

[0020] Through the above technical solution, the online SOC dynamic calibration method constructs a battery parameter table and uses the least squares method with a forgetting factor for online parameter identification. It dynamically adjusts the forgetting factor through a segmented error threshold, combines over-limit judgment to ensure numerical stability, and uses the extended Kalman filter algorithm to estimate and calibrate SOC in real time. This effectively suppresses accumulated errors, significantly improves estimation accuracy and system adaptability, and is more suitable for the actual operating conditions of automobiles.

[0021] like Figure 2 The diagram shown is a connection block diagram for generating a battery model parameter table according to a segmented online SOC dynamic calibration method with a forgetting factor, based on an embodiment of the present invention. Figure 2 In order to construct an equivalent circuit model and generate a battery model parameter table, in one embodiment of the present invention, constructing the battery model parameter table may include the following steps: In step S20, the HPPC experimental method is used to obtain the current, voltage, and time series data of the battery under different operating conditions, and an equivalent circuit model is established according to formula (1). (1) in, for Battery terminal voltage at time, for Open-circuit voltage at any given time for The first polarization voltage at time 1, for The second polarization voltage at time , for The first polarization voltage at time 1, for The second polarization voltage at time , For ohmic internal resistance, For the first polarization resistor, This is the second polarization resistor. This is the first polarization capacitor. This is the second polarization capacitor. for Current at any moment for Current at any moment It is a natural constant. For time step; In step S21, the SOC breakpoint is determined based on the changing trends of current and time. In step S22, the data is divided into multiple data subsets based on the breakpoints to determine multiple SOC intervals; In step S23, battery response curve fitting is performed on each data subset to calculate and obtain ohmic internal resistance parameters and polarization parameters; In step S24, an nth-order polynomial fit is performed on the open-circuit voltage and the state of charge (SOC) to generate a battery model parameter table.

[0022] Battery dynamic data was acquired through HPPC testing, and a second-order equivalent circuit model was established. First, SOC breakpoints were identified based on current change trends, and multiple SOC intervals were defined. Then, dynamic response curves were fitted to subsets of data from each interval to accurately calculate ohmic internal resistance and polarization parameters. Subsequently, a functional relationship between open-circuit voltage and SOC was established through polynomial fitting, ultimately generating a battery model parameter table containing parameters under multiple operating conditions, providing an accurate benchmark for online calibration.

[0023] To obtain the ohmic internal resistance parameters and polarization parameters, in one embodiment of the present invention, battery response curve fitting is performed on each subset of data, and the calculation of the ohmic internal resistance parameters and polarization parameters may include the following steps: In step S30, the key curve segments of the battery response curve are identified and extracted; In step S31, the curve segment of the instantaneous voltage jump point in the impulse response is fitted using formula (2) to obtain the ohmic resistance parameter. (2) in, , , , This refers to the voltage point in the HPPC experiment. It is a pulse current; In step S32, the relaxation curve segment is fitted using formula (3) to obtain the polarization parameters. (3) in, This refers to the battery terminal voltage. Open circuit voltage, For current, It is a natural constant. For time step, For the first polarization resistor, This is the second polarization resistor. This is the first polarization capacitor. This is the second polarization capacitor. During the parameter extraction stage, by identifying key segments of the battery response curve (such as pulse jump points and relaxation segments), the ohmic internal resistance is obtained by fitting the instantaneous voltage jump segment, and the polarization resistance and capacitance parameters are obtained by fitting the relaxation segment, thereby completing the high-precision calibration of the battery model parameters.

[0024] like Figure 3 The diagram shown is a connection block diagram for identifying battery model parameters in a segmented online SOC dynamic calibration method with a forgetting factor according to an embodiment of the present invention; Figure 3 In this invention, considering the segmented dynamic adjustment of the forgetting factor to achieve online identification of battery model parameters, in one embodiment, online identification of battery model parameters based on the least squares method with a forgetting factor may include the following steps: In step S40, a continuous S-domain transfer function is constructed based on the equivalent circuit model using formula (4). (4) in, Let be the transfer function in the S-domain. The first time constant, The second time constant, For Laplace variables; In step S41, the discretized transfer function is constructed using formula (5) based on the bilinear transformation. (5) in, It is a constant. Let Z be the variable for transformation; In step S42, the vector of parameters to be estimated and the amount of input data are defined using formulas (6)-(7), respectively. (6) (7) - (8) in, for The vector of parameters to be estimated at time t. , , , , , It is a constant. for The amount of input data at any given time. for Output variables at time 10:00 for Output variables at time 10:00 for Output variables at time 10:00 for Input variables at time, for Input variables at time, for Input variables at time, for The output variable at time t is The difference between the open-circuit voltage and the terminal voltage at a given time; In step S43, the process is linearized using the least squares method, and the output variable is obtained according to formula (9). (9) in, for Output variables at time 10:00 for The transpose of the input data vector at time t. for The vector of parameters to be estimated at time t. for The prediction error at any given time; In step S44, the estimated parameter vector is updated using a recursive algorithm based on formula group (10). (10) in, for Prediction error at time, for The transpose of the input data vector at time t. for The input data vector at time t, for The estimated parameter vector at time step, for The estimated parameter vector at time step, for The Kalman gain matrix at time t. for The covariance matrix at time t, for The covariance matrix at time t, Forgetting factor parameters, It is the identity matrix; In step S45, different allowable errors are set according to different SOC intervals, and the forgetting factor parameters are dynamically adjusted in a segmented manner using formula (11). (11) in, for The forgetting factor parameter at time, The parameter is the minimum forgetting factor. The parameter for the maximum forgetting factor. The coefficient of variation, for Error scaling factor at time step The allowable error threshold; In step S46, the constants in the discretized transfer function are identified based on the dynamic forgetting factor parameters, intermediate variables are obtained using formula group (12), and the out-of-limit values ​​in the intermediate variables are processed using the parameter out-of-limit processing method. (12) in, , , , , Here, T is an intermediate variable, and T is the sampling time. In step S47, the battery model parameters are obtained by performing parameter transformation using formula (13) based on the intermediate variables. (13) in, The first time constant, This is the second time constant.

[0025] First, the continuous battery model is transformed into a discrete transfer function form, and the vector of parameters to be estimated and the vector of historical data are defined. The parameter estimates are updated in real time by predicting errors, and the correction process is optimized using the Kalman gain matrix. Differential error thresholds are set according to different SOC ranges of the battery. When the prediction error increases, the forgetting factor is automatically reduced to enhance the response to new data, and vice versa, the algorithm stability is maintained, thereby significantly improving the adaptability of parameter identification to dynamic operating conditions. Through intermediate variable calculation and over-limit judgment processing, the identified discrete coefficients are converted into battery physical parameters (such as ohmic internal resistance, polarization resistance / capacitance), providing an accurate model basis for EKF real-time calibration and effectively ensuring the accuracy and numerical stability of online identification.

[0026] like Figure 4 The diagram shown is a connection block diagram for parameter over-limit handling in a segmented online SOC dynamic calibration method with a forgetting factor according to an embodiment of the present invention; Figure 4In order to reduce the computational load, facilitate microcontroller calculations, and avoid the denominator term being zero due to its precision, thus affecting parameter identification, in one embodiment of the present invention, the method of handling out-of-limit values ​​in intermediate variables using a parameter out-of-limit processing method may include the following steps: In step S50, it is determined whether the value of the denominator of the intermediate variable is less than a preset first threshold. In step S51, if the value of the denominator of the intermediate variable is less than a preset first threshold, the value of the denominator of the intermediate variable at the previous time step is obtained. In step S52, if the value of the denominator of the intermediate variable is greater than or equal to a preset first threshold, the value of the denominator of the intermediate variable at the current time is obtained.

[0027] The algorithm optimizes computational efficiency and prevents errors such as zeroing the denominator by handling parameter over-limits. It judges in real time whether the denominator term is less than the preset threshold. If it is less, the value of the previous time step is used to avoid computational divergence. Otherwise, the current value is maintained. This significantly improves the stability and adaptability of the algorithm in a microcontroller environment.

[0028] like Figure 5 The diagram shown is a connection block diagram for triggering and terminating online SOC calibration according to an embodiment of the present invention, which describes a segmented online SOC dynamic calibration method with a forgetting factor. Figure 5 In order to accurately estimate and calibrate the State of Charge (SOC) online and iteratively correct the SOC, in one embodiment of the present invention, the online estimation and calibration of the SOC using the EKF method may include the following steps: In step S60, it is determined whether the EKF calibration SOC condition has been triggered; In step S61, when the EKF calibration SOC condition is triggered, the SOC is estimated and calibrated online; In step S62, it is determined whether the EKF termination calibration SOC condition has been triggered; In step S63, if the EKF termination calibration SOC condition is not triggered, the steps of online estimation and calibration of SOC are iteratively repeated. In step S64, if the EKF termination calibration SOC condition is triggered, the current calibrated SOC is output.

[0029] The EKF calibration process is controlled by an intelligent triggering and termination mechanism. When a sudden current change or a sustained voltage deviation from the predicted value is detected (dynamic operating condition), EKF calibration is automatically triggered. During calibration, the system continuously iterates and updates the SOC estimate until any termination condition is met: state estimation convergence, current returning to steady state, or the maximum number of iterations is reached. At this point, the final calibrated SOC value is output. This approach ensures rapid correction of accumulated ampere-hour integral errors under dynamic operating conditions (keeping the estimation error within 3%) while avoiding unnecessary computational resource consumption under steady-state conditions.

[0030] like Figure 6 The diagram shown is a connection block diagram of online SOC estimation and calibration in a segmented online SOC dynamic calibration method with a forgetting factor according to an embodiment of the present invention. Figure 6 In order to achieve SOC calibration using the EKF algorithm and output the optimal SOC value, in one embodiment of the present invention, when the EKF calibration SOC condition is triggered, online estimation and calibration of SOC may include the following steps: In step S70, the initial state and noise characteristics of the EKF algorithm are set, and the state vector is initialized according to formula (14). (14) in, For state vectors, The initial state of charge, This is the first polarization voltage. This is the second polarization voltage. The state estimates the covariance matrix. The process noise covariance matrix is... To measure the noise covariance matrix; In step S71, formulas (15)-(16) are used to predict the state vector and error covariance matrix of the battery model at the current moment. (15) (16) in, for The prior estimate of the state vector at time t. for The state transition matrix at time t, for The optimal estimate of the updated state vector at time step [time]. for The input matrix at time step, for Measuring current at any given time for The predicted value of the error covariance matrix at time t. for The transpose of the state transition matrix at time t. The process noise covariance matrix; In step S72, the predicted state vector is corrected using formula (17); (17) in, for The predicted value of the terminal voltage at time [time]. for The output matrix at time 10:00. It is a feedforward matrix; In step S73, the innovation vector and gain matrix are obtained using formula group (18). (18) in, For innovation vectors, for The input unit voltage value at any given time. for Gain matrix at time step for The transpose of the output matrix at time t. To measure the noise covariance matrix; In step S74, the state vector and error covariance matrix are updated and calibrated using formulas (19)-(20) based on the innovation vector and gain matrix to obtain the optimal estimate at the current time. (19) (20) in, for The optimal estimate of the updated state vector at time step [time]. For the updated The error covariance matrix at each time step is calculated. By initializing the state vector (SOC, polarization voltage) and noise parameters, the battery state and error covariance are predicted using the state equation. An innovative vector between the measured and predicted terminal voltage values ​​is used to calculate the optimal weights of the Kalman gain. Finally, feedback correction is applied to the state estimate, outputting the optimal SOC value. This process is implemented through a recursive "prediction-measurement-update" structure. The EKF algorithm integrates battery model predictions with real-time voltage measurements, effectively suppressing the accumulated error of the ampere-hour integration method. The synergistic mechanism of the innovative vector and Kalman gain enables the SOC estimate to quickly track the actual state under dynamic operating conditions, converging the error to within 3%. Its recursive calculation structure significantly improves adaptive capability, ensuring the safe management of electric vehicle battery systems.

[0031] A second aspect of the present invention provides a computer device including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the dynamic calibration method as described in any of the preceding claims.

[0032] A third aspect of the present invention provides an electric vehicle including a battery system for performing the dynamic calibration method as described in any of the preceding claims.

[0033] Through the above technical solution, this online SOC dynamic calibration method first constructs battery model parameter tables for different SOC ranges through experiments, establishing an accurate second-order RC equivalent circuit model. Then, it employs the least squares method with a forgetting factor for online parameter identification. The forgetting factor is dynamically adjusted through a piecewise error threshold to achieve adaptive updating of model parameters. An over-limit judgment mechanism ensures numerical stability. Based on the extended Kalman filter algorithm, the identified parameters are used for real-time estimation and calibration of SOC. Through state prediction, measurement update, and posterior estimation recursive calculation, the cumulative error of the ampere-hour integration method is effectively suppressed. The piecewise design enables the algorithm to adapt to the nonlinear characteristics of different SOC ranges of the battery, significantly improving parameter identification accuracy and operating condition adaptability. The online identification mechanism shortens the offline calibration time. The fusion of EKF and FFRLS effectively controls the SOC estimation error, enabling it to converge to within 3% even under complex dynamic operating conditions, significantly improving the estimation accuracy and reliability of the battery management system.

[0034] The preferred embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the specific details of the above embodiments. Within the scope of the technical concept of the present invention, various simple modifications can be made to the technical solutions of the present invention, and these simple modifications all fall within the protection scope of the present invention. Furthermore, it should be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the present invention will not describe the various possible combinations separately.

Claims

1. A segmented online SOC dynamic calibration method with a forgetting factor, characterized in that, The dynamic calibration method includes: Construct a battery model parameter table; Battery model parameters are identified online using the least squares method with a forgetting factor. The EKF method was used to estimate and calibrate the SOC online.

2. The dynamic calibration method according to claim 1, characterized in that, Construct a battery model parameter table, including: HPPC experimental method was used to obtain current, voltage and time series data of battery under different operating conditions, and equivalent circuit model was established according to formula group (1). ，（1） in, for Battery terminal voltage at time, for Open-circuit voltage at any given time for The first polarization voltage at time 1, for The second polarization voltage at time , for The first polarization voltage at time 1, for The second polarization voltage at time , For ohmic internal resistance, For the first polarization resistor, This is the second polarization resistor. This is the first polarization capacitor. This is the second polarization capacitor. for Current at any moment for Current at any moment It is a natural constant. For time step; Determine the SOC breakpoint based on the changing trends of current and time; Based on the breakpoints, the data is divided into multiple data subsets, and multiple SOC intervals are determined. For each subset of data, the battery response curve is fitted, and the ohmic internal resistance parameter and polarization parameter are calculated and obtained. An nth-order polynomial fit is performed on the open-circuit voltage and the state of charge (SOC) to generate a battery model parameter table.

3. The dynamic calibration method according to claim 2, characterized in that, For each subset of data, battery response curves were fitted, and ohmic internal resistance and polarization parameters were calculated, including: Identify and extract key curve segments from the battery response curve; The ohmic resistance parameters are obtained by fitting the curve segment of the instantaneous voltage jump point in the impulse response using formula (2). ，（2） in, , , , This refers to the voltage point in the HPPC experiment. It is a pulse current; The relaxation curve segment is fitted using formula (3) to obtain the polarization parameters. ，（3） in, This refers to the battery terminal voltage. Open circuit voltage, For current, It is a natural constant. For time step, For the first polarization resistor, This is the second polarization resistor. This is the first polarization capacitor. This is the second polarization capacitor.

4. The dynamic calibration method according to claim 1, characterized in that, Battery model parameters are identified online using the least squares method with a forgetting factor, including: Based on the equivalent circuit model, a continuous S-domain transfer function is constructed using formula (4). ，（4） in, Let be the transfer function in the S-domain. The first time constant, The second time constant, For Laplace variables; Based on the bilinear transformation, the discretized transfer function is constructed using formula (5). ，（5） in, It is a constant. Let Z be the variable for transformation; The vector of parameters to be estimated and the amount of input data are defined using formulas (6) and (7), respectively. ，（6） ，（7） - ，（8） in, for The vector of parameters to be estimated at time t. , , , , , It is a constant. for The amount of input data at any given time. for Output variables at time 10:00 for Output variables at time 10:00 for Output variables at time 10:00 for Input variables at time, for Input variables at time, for Input variables at time, for The output variable at time t is The difference between the open-circuit voltage and the terminal voltage at a given time; Linearization is performed using the least squares method, and the output variables are obtained according to formula (9). ，（9） in, for Output variables at time 10:00 for The transpose of the input data vector at time t. for The vector of parameters to be estimated at time t. for The prediction error at any given time; The estimated parameter vector is updated using a recursive algorithm based on formula (10). ，（10） in, for Prediction error at time, for The transpose of the input data vector at time t. for The input data vector at time t, for The estimated parameter vector at time step, for The estimated parameter vector at time step, for The Kalman gain matrix at time t. for The covariance matrix at time t, for The covariance matrix at time t, Forgetting factor parameters, It is the identity matrix; Different allowable errors are set according to different SOC intervals, and the forgetting factor parameters are dynamically adjusted in a segmented manner using formula (11). ，（11） in, for The forgetting factor parameter at time, The parameter is the minimum forgetting factor. The parameter for the maximum forgetting factor. The coefficient of variation, for Error scaling factor at time step The allowable error threshold; Based on the dynamic forgetting factor parameters, the constants in the discretized transfer function are identified, and the intermediate variables are obtained by formula group (12). The parameter over-limit processing method is used to process the over-limit values ​​in the intermediate variables. ，（12） in, , , , , Here, T is an intermediate variable, and T is the sampling time. Based on the intermediate variables, the parameters are transformed using formula (13) to obtain the battery model parameters. ，（13） in, The first time constant, This is the second time constant.

5. The dynamic calibration method according to claim 4, characterized in that, Methods for handling out-of-limit values ​​in intermediate variables are employed, including: Determine whether the value of the denominator of the intermediate variable is less than a preset first threshold; If the value of the denominator of the intermediate variable is less than a preset first threshold, obtain the value of the denominator of the intermediate variable at the previous time step. If the value of the denominator of the intermediate variable is greater than or equal to a preset first threshold, obtain the value of the denominator of the intermediate variable at the current time.

6. The dynamic calibration method according to claim 1, characterized in that, The EKF method is used for online estimation and calibration of SOC, including: Determine whether the EKF calibration SOC condition has been triggered; When the EKF calibration SOC condition is triggered, the SOC is estimated and calibrated online. Determine whether the EKF termination calibration SOC condition has been triggered; If the EKF termination calibration SOC condition is not triggered, the steps of iteratively re-estimating and calibrating the online SOC are performed. If the EKF termination calibration SOC condition is triggered, output the current calibrated SOC.

7. The dynamic calibration method according to claim 6, characterized in that, When the EKF calibration SOC condition is triggered, estimate and calibrate the SOC online, including: Set the initial state and noise characteristics of the EKF algorithm, and initialize the state vector according to formula (14). ，（14） in, For state vectors, The initial state of charge, This is the first polarization voltage. This is the second polarization voltage. The covariance matrix is ​​estimated for the state. The process noise covariance matrix is... To measure the noise covariance matrix; Formulas (15)-(16) are used to predict the state vector and error covariance matrix of the battery model at the current moment. ，（15） ，（16） in, for The prior estimate of the state vector at time t. for The state transition matrix at time t, for The optimal estimate of the updated state vector at time step [time]. for The input matrix at time step, for Measuring current at any given time for The predicted value of the error covariance matrix at time t. for The transpose of the state transition matrix at time t. The process noise covariance matrix; The predicted state vector is corrected using formula (17); ，（17） in, for The predicted value of the terminal voltage at time [time]. for The output matrix at time 10:

00. It is a feedforward matrix; The innovation vector and gain matrix are obtained using formula (18). ，（18） in, For innovation vectors, for The input unit voltage value at any given time. for Gain matrix at time step for The transpose of the output matrix at time t. To measure the noise covariance matrix; Based on the innovation vector and gain matrix, the state vector and error covariance matrix are updated and calibrated using formulas (19)-(20) to obtain the optimal estimate at the current time. ，（19） ，（20） in, for The optimal estimate of the updated state vector at time step [time]. For the updated The error covariance matrix at time t.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the dynamic calibration method according to any one of claims 1 to 7.

9. An electric vehicle, characterized in that, The electric vehicle includes a battery system for performing the dynamic calibration method as described in any one of claims 1 to 7.

Images