Battery SOC prediction method fusing online identification and adaptive algorithm

Abstract

The invention discloses a battery SOC prediction method fusing online identification and an adaptive algorithm, relates to the technical field of power battery monitoring, and aims to solve the problem of insufficient SOC prediction precision of an electric vehicle power battery under a dynamic working condition. The prediction method comprises the steps of obtaining lithium battery charging and discharging experiment data, establishing a dual-polarization equivalent circuit model to simulate a strong nonlinear behavior in the battery, and fitting an OCV-SOC curve based on the experiment data; performing online parameter identification on the equivalent circuit model by using a forgetting factor recursive least square method based on grid search-real-time voltage difference improvement; sOC prediction is carried out by combining online parameter identification with a self-adaptive expansion Kalman filtering algorithm; and the proposed prediction algorithm is verified under a mixed pulse power characteristic test working condition. According to the method, the accuracy and the real-time performance of SOC prediction are remarkably improved, and the method has a wide application prospect.

Description

Technical Field

[0001] The present invention relates to the technical field of power battery monitoring, and in particular to a battery SOC prediction method integrating online identification and adaptive algorithms. Background Art

[0002] Lithium batteries are widely used in new energy vehicles due to their high energy density, low self-discharge rate, long cycle life and environmental friendliness. To ensure the safe, reliable and efficient operation of lithium batteries, the battery management system (BMS) needs to monitor its working status in real time. The state of charge (SOC) of the power battery is one of the important states to be monitored. During the actual driving of electric vehicles, the SOC of the battery cannot be directly measured by sensors, and it needs to rely on various algorithms to predict it. The accuracy of the remaining battery power prediction directly determines the reference degree of the remaining driving range. Although electric vehicles have obvious advantages over traditional fuel vehicles in terms of power and environmental protection, drivers are uncertain about the accurate remaining driving range of electric vehicles, which leads to widespread range anxiety. Therefore, it is particularly urgent to solve the problem of the accuracy of electric vehicle remaining power prediction.

[0003] At present, the model-based method can obtain the highly dynamic behavior of the battery through the power supply, resistor and other components in the equivalent circuit. In previous studies, the parameters of the equivalent circuit model are usually obtained by offline identification. However, the parameters of offline identification cannot adjust the model parameters according to real-time data and reflect the changes in the internal state of the battery. Therefore, some scholars have proposed a method based on forgetting factor recursive least squares (FFRLS) to identify the parameters of the battery model online. However, the inventors found in the study that the value of the forgetting factor in the forgetting factor recursive least squares method needs to be adjusted manually to obtain a relatively stable identification result. The forgetting factor is too large or too small. The impact on the identification result is relatively large. Therefore, it is important to make the forgetting factor in the algorithm change dynamically with the system operation and determine the appropriate initial parameter range for the accuracy of parameter identification and SOC prediction results. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a power battery SOC prediction method that integrates online identification and adaptive algorithms to address the defects involved in the background technology. It can not only accurately track the dynamic changes of model parameters under cyclic conditions in real time, but also predict SOC under typical conditions with higher accuracy and robustness.

[0005] The present invention adopts the following technical solutions to solve the above technical problems:

[0006] A power battery SOC prediction method integrating online identification and adaptive algorithm is characterized in that a hybrid pulse power characteristic test condition (HPPC) is performed on the power battery using a battery charge and discharge test platform to obtain experimental data and fit the SOC-OCV curve;

[0007] Establish a lithium battery dual-polarization equivalent circuit model that can meet both the requirements of computational efficiency and real-time prediction accuracy, and obtain the electrical relationship of the dual-polarization equivalent circuit model from Kirchhoff's current-voltage law;

[0008] The electrical relationship of the dual-polarization equivalent circuit model is used to perform online parameter identification using a forgetting factor recursive least square method based on grid search-real-time voltage difference improvement to obtain the component parameters of the dual-polarization equivalent circuit model.

[0009] The adaptive extended Kalman filter algorithm is combined with the dual-polarization equivalent circuit model with known parameters to establish the SOC prediction model. By adding a noise adaptive module to the extended Kalman filter algorithm, the noise and covariance can be continuously updated according to data changes to obtain the SOC prediction value.

[0010] Furthermore, the power battery uses a cylindrical 4680 lithium battery with a diameter of 46mm and a length of 80mm produced by Panasonic. Its positive and negative poles are on the same side, the protruding circle in the middle is the positive pole, and the surrounding of the circle is the negative pole. The operating temperature range is -30℃-60℃, and the maximum discharge rate is 8C;

[0011] Furthermore, the battery charge and discharge test platform consists of a LANBTS (BT-2016E) battery tester, a host computer, and a SPX-150BE constant temperature incubator;

[0012] Furthermore, the specific method of establishing the dual-polarization equivalent circuit model is as follows: the strong nonlinear behavior inside the battery is replaced by two RC links and an ohmic internal resistance, and the electrical relationship of the dual-polarization equivalent circuit model is obtained according to Kirchhoff's current-voltage law;

[0013] Furthermore, the forgetting factor recursive least squares online parameter identification method based on grid search-real-time voltage difference improvement is specifically as follows: When the recursive least squares method integrating the forgetting factor is used for parameter identification, the forgetting factor needs to be manually adjusted many times to obtain a relatively stable identification result. In order to improve the calculation efficiency, the forgetting factor recursive least squares method is improved based on the error between the terminal voltage measurement value and the model prediction value so that the forgetting factor changes dynamically within a reasonable range. The strategy of improving the forgetting factor recursive least squares method with real-time voltage error is as follows:

[0014]

[0015] Among them, μ tis the forgetting factor at time t; μ min With μ max is the upper and lower limits of the forgetting factor; σ 2 is the variance of the voltage error; e is the voltage difference; a is the smoothing factor.

[0016] Since the improved forgetting factor recursive least squares method needs to determine the size of the smoothing factor and the upper and lower limits of the forgetting factor to ensure the accuracy and volatility of the parameter identification results, a flexible and simple grid search method is selected to traverse the size of the smoothing factor and the upper and lower limits of the forgetting factor within a certain range to find the optimal parameter combination that minimizes the error and balances the parameter volatility. The mathematical model for selecting the initial parameters of the improved forgetting factor least squares method by grid search is as follows:

[0017]

[0018] Among them, the present invention searches and selects a=0.95, μ min =0.95, μ max =1.

[0019] Furthermore, the electrical relationship of the established equivalent circuit model needs to be converted before the online parameter identification method, specifically as follows: the electrical relationship of the dual-polarization equivalent circuit model is Laplace transformed to obtain a transfer function; a discretized form of the transfer function can be obtained by using a bilinear transformation method, and then converted into a differential equation form;

[0020] Furthermore, HPPC test is used to obtain experimental data, and the obtained test data is used to solve the parameter values ​​of the equivalent circuit model based on the online parameter identification method;

[0021] Furthermore, the implementation of the adaptive algorithm is as follows: the extended Kalman filter is used to linearize the nonlinear model to retain the first-order Taylor expansion term of the nonlinear function, and a noise adaptive module is added on its basis, which can continuously update the noise and covariance according to the data changes. The recursive process of the adaptive extended Kalman filter algorithm is as follows:

[0022]

[0023] Among them, e t is the observation error at time, H t is the covariance matrix of the new data information based on the moving average filter with a window size of M, where M is the window size.

[0024] Compared with the prior art, the present invention adopts the above technical solution and has the following technical effects:

[0025] 1. The present invention develops an online parameter identification method based on grid search-real-time voltage difference improved forgetting factor recursive least squares, which can accurately track the dynamic changes of model parameters under complex working conditions in real time and verify the accuracy of the established dual-polarization equivalent circuit model.

[0026] 2. Based on the experimental data of 4680 power battery under HPPC conditions, the present invention carries out model parameter identification and SOC prediction of the joint algorithm, and verifies the accuracy and robustness of the joint algorithm for SOC prediction performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 The figure is a flow chart of the specific steps of the SOC prediction method of the present invention.

[0028] Figure 2 This is a dual-polarization equivalent circuit model diagram of the present invention.

[0029] Figure 3 It is a comparison diagram of the terminal voltage prediction value and the actual value of the equivalent circuit model of the present invention.

[0030] Figure 4 This is a terminal voltage prediction error diagram of the present invention.

[0031] Figure 5 This is a comparison chart of the SOC measurement value and the predicted value of the present invention.

[0032] Figure 6 This is the SOC prediction error diagram of the present invention.

[0033] Figure 7 It is the technical roadmap of the present invention. DETAILED DESCRIPTION

[0034] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings:

[0035] Example

[0036] like Figures 1 to 7As shown, a battery SOC prediction method integrating online identification and adaptive algorithm includes the following steps: using a battery charge and discharge test platform to perform a hybrid pulse power characteristic test condition (HPPC) on a power battery, obtain experimental data, and fit the SOC-OCV curve; establishing a lithium battery dual-polarization equivalent circuit model that meets both the requirements of computational efficiency and the requirements of real-time prediction accuracy, and obtaining the electrical relationship of the dual-polarization equivalent circuit model from Kirchhoff's current law and voltage law; using the obtained electrical relationship to perform online parameter identification using a forgetting factor recursive least squares method based on grid search-real-time voltage difference improvement to obtain the component parameters of the model; combining an adaptive extended Kalman filter algorithm with a dual-polarization equivalent circuit model with known parameters to establish a SOC prediction model, and by adding a noise adaptive module to the extended Kalman filter algorithm, the noise and covariance can be continuously updated according to data changes to obtain the SOC prediction value;

[0037] The specific experimental steps of the HPPC test are as follows: the experiment is mainly divided into two stages: long rest and short pulse. A long rest stage is performed every 10% SOC, so that the OCV corresponding to the current SOC can be obtained when the battery reaches stability, and the short pulse stage is mainly used to measure the internal resistance.

[0038] 1) Set the temperature of the thermostat to 25°C and place the battery in the thermostat for 1 hour;

[0039] 2) Fully charge the battery at constant current and constant voltage according to the standard, let it stand for 1 hour, and record the voltage at this time;

[0040] 3) Discharge at a constant current of 1C (23A) for 3 minutes, let stand for 1 hour, and record the voltage at this time;

[0041] 4) Discharge at 1C (23A) constant current for 10 seconds and let stand for 40 seconds;

[0042] 5) Charge at 0.75C (17.25A) constant current for 10 seconds and let stand for 40 seconds;

[0043] 6) Repeat steps 3 to 5 twice, and the SOC is now 90%;

[0044] 7) Discharge at a constant current of 1C (23A) for 6 minutes, let stand for 1 hour, and record the voltage at this time;

[0045] 8) Discharge at 1C (23A) constant current for 10 seconds and let stand for 40 seconds;

[0046] 9) Charge at 0.75C (17.25A) constant current for 10 seconds and let stand for 40 seconds;

[0047] 10) Repeat steps 7 to 9 for 7 times, and the SOC is now 20%;

[0048] 11) Discharge at a constant current of 1C (23A) for 3 minutes, let stand for 1 hour, and record the voltage at this time;

[0049] 12) Discharge at 1C (23A) constant current for 10 seconds and let stand for 40 seconds;

[0050] 13) Charge at 0.75C (17.25A) constant current for 10 seconds and let stand for 40 seconds;

[0051] 14) Repeat steps 11 to 13 twice, and the SOC is now 10%.

[0052] The specific method of establishing the dual-polarization equivalent circuit model is as follows: the strong nonlinear behavior inside the battery is replaced by two RC links and an ohmic internal resistance, and the following is established: Figure 2 The dual-polarization equivalent circuit model shown in the figure has the following electrical relationship according to Kirchhoff's current-voltage law:

[0053]

[0054] Where i is the current. In this paper, the discharge current is positive and the charging current is negative. U t is the terminal voltage; U is the open circuit voltage; R0 is the ohmic internal resistance; R a , R b are electrochemical polarization internal resistance and concentration polarization internal resistance respectively; C a , C b are the electrochemical polarization capacitance and concentration polarization capacitance respectively; U a , U b are the electrochemical polarization voltage and concentration polarization voltage respectively. In this paper, the coulomb efficiency η=1 and the sampling time T c =1S, maximum available capacity Q n =21.18Ah.

[0055] The grid search-real-time voltage difference-based improved forgetting factor recursive least squares online parameter identification method is specifically as follows: when the recursive least squares method integrating the forgetting factor is used for parameter identification, the forgetting factor needs to be manually adjusted many times to obtain a relatively stable identification result. In order to improve the calculation efficiency, the forgetting factor recursive least squares method is improved based on the error between the terminal voltage measurement value and the model prediction value so that the forgetting factor changes dynamically within a reasonable range. The strategy of improving the forgetting factor recursive least squares method with real-time voltage error is as follows:

[0056]

[0057] Among them, μ t is the forgetting factor at time t; μ min With μ max is the upper and lower limits of the forgetting factor; σ 2is the variance of the voltage error; e is the voltage difference; a is the smoothing factor.

[0058] Since the improved forgetting factor recursive least squares method needs to determine the size of the smoothing factor and the upper and lower limits of the forgetting factor to ensure the accuracy and volatility of the parameter identification results, a flexible and simple grid search method is selected to traverse the size of the smoothing factor and the upper and lower limits of the forgetting factor within a certain range to find the optimal parameter combination that minimizes the error and balances the parameter volatility. The mathematical model for selecting the initial parameters of the improved forgetting factor least squares method by grid search is as follows:

[0059]

[0060] Among them, the present invention searches and selects α=0.95, μ min =0.95, μ max =1.

[0061] Before the online parameter identification method, the electrical relationship of the established equivalent circuit model needs to be converted as follows:

[0062] From the electrical relationship of the above dual-polarization equivalent circuit model and Laplace transform, we can get:

[0063]

[0064] Among them, τ a =R a C a , τ b =R b C b , τ a , τ b is the time constant.

[0065] Depend on Let Y(s) = U(s) - U t (s) The transfer function can be obtained:

[0066]

[0067] The bilinear transformation method is used to map equation (5) from the s plane to the z plane, The discretized form of the above transfer function can be obtained:

[0068]

[0069] Among them, k1, k2, k3, k4, k5 are corresponding coefficients, T c is the sampling time.

[0070] Formula (6) can be expressed in the form of a difference equation:

[0071] Y t =U t -U t,t =k1Y t-1 +k2Y t-2 +k3I t +k4I t-1 +k5I t-2 (7)

[0072] Among them, Y t is the output at time t, U t is the open circuit voltage value. According to the corresponding relationship between the open circuit voltage and the state of charge, I t，t with U t，t is the input value at time t, obtained from experiments.

[0073] It can also be roll out:

[0074]

[0075] Substituting formula (8) into formula (6), we can obtain:

[0076]

[0077] Combining equation (5) and equation (9), the relationship between the battery model parameters and the corresponding coefficients can be deduced:

[0078]

[0079] The component parameters of the dual-polarization equivalent circuit can be further solved by equation (7):

[0080]

[0081] Will I t As the circuit system input, Y t As the circuit system output, let:

[0082]

[0083] θ t =[k1,k2,k3,k4,k5] T (13) in, θ t are the data matrix and parameter matrix respectively. The parameters in the above estimation matrix can be identified only through the current and voltage values ​​collected in real time in the experiment. According to equations (10) and (11), the model parameter R in the dual-polarization equivalent circuit model can be obtained. o , R a , R b , C a and C bThe numerical value of .

[0084] In this embodiment, in order to illustrate the accuracy of the parameter identification result, the terminal voltage prediction result after the identification result is used to show that the terminal voltage prediction result under the HHHPPC condition is as follows: Figure 3 As shown, the corresponding terminal voltage prediction error is Figure 4 As shown, the maximum absolute error MAX of the predicted terminal voltage by the online identification method is 0.08V, and the average absolute error MAE is 0.01mV, which has stable convergence and good dynamic tracking ability.

[0085] The state transfer equation and observation equation of the dual-polarization equivalent circuit model system are as follows:

[0086]

[0087] The nonlinear state equation and the observation equation are combined at the current estimated value Performing a first-order Taylor expansion at gives:

[0088]

[0089] The state transfer matrix and observation matrix are The nonlinear observation equation is calculated at the current estimate The linearized observation equation is obtained by performing a first-order Taylor expansion at:

[0090]

[0091] The implementation method of the adaptive algorithm is as follows: the extended Kalman filter is used to linearize the nonlinear model to retain the first-order Taylor expansion term of the nonlinear function, and a noise adaptive module is added on its basis to continuously update the noise and covariance according to the data changes. The recursive process of the adaptive extended Kalman filter algorithm is as follows:

[0092]

[0093] Among them, e t is the observation error at time t, H t is the covariance matrix of the new data information based on the moving average filter with a window size of M, where M is the window size.

[0094] In order to verify the accuracy and robustness of the algorithm proposed in this invention, the battery was tested through a mixed power pulse experiment. The effectiveness of the proposed joint algorithm was verified under HPPC conditions. Figure 5 The SOC prediction results shown in Figure 6 The SOC prediction errors shown have a maximum absolute error of 0.3% and an average absolute error of 0.08%, both of which are within 1%.

[0095] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as those generally understood by those skilled in the art in the art to which the present invention belongs. It should also be understood that terms such as those defined in common dictionaries should be understood to have a meaning consistent with the context of the prior art, and will not be interpreted with an idealized or overly formal meaning unless defined as herein.

[0096] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A battery SOC prediction method integrating online identification and adaptive algorithm, characterized in that: The method comprises the following steps: Use the battery charge and discharge test platform to perform hybrid pulse power characteristic test (HPPC) on the power battery, obtain experimental data, and fit the SOC-OCV curve; Establish a lithium battery dual-polarization equivalent circuit model that can meet both the requirements of computational efficiency and real-time prediction accuracy, and obtain the electrical relationship of the dual-polarization equivalent circuit model from Kirchhoff's current-voltage law; The electrical relationship of the dual-polarization equivalent circuit model is used to perform online parameter identification using a forgetting factor recursive least square method based on grid search-real-time voltage difference improvement to obtain the component parameters of the dual-polarization equivalent circuit model. The adaptive extended Kalman filter algorithm is combined with a dual-polarization equivalent circuit model with known parameters to establish a SOC prediction model. By adding a noise adaptive module to the extended Kalman filter algorithm, the noise and covariance can be continuously updated according to data changes to obtain the SOC prediction value.

2. The power battery according to claim 1, characterized in that: A cylindrical 4680 lithium battery with a diameter of 46mm and a length of 80mm produced by Panasonic is selected. The positive and negative poles are on the same side, the protruding circle in the middle is the positive pole, and the surrounding area of ​​the circle is the negative pole. The operating temperature range is -30℃-60℃, and the maximum discharge rate is 8C.

3. The battery charge and discharge test platform according to claim 1, characterized in that: It consists of a LANBTS (BT-2016E) battery tester, a host computer and a SPX-150BE constant temperature incubator.

4. The dual-polarization equivalent circuit model according to claim 1, characterized in that: The specific method of establishing the dual-polarization equivalent circuit model is as follows: the strong nonlinear behavior inside the battery is replaced by two RC links and an ohmic internal resistance, and the electrical relationship of the dual-polarization equivalent circuit model is obtained according to Kirchhoff's current and voltage law.

5. The online parameter identification method according to claim 1, characterized in that: The specific method of online parameter identification based on grid search-improved forgetting factor recursive least squares of real-time voltage difference is as follows: When the recursive least squares method with forgotten factor is used for parameter identification, the forgetting factor needs to be manually adjusted many times to obtain a relatively stable identification result. In order to improve the calculation efficiency, the forgetting factor recursive least squares method is improved based on the error between the terminal voltage measurement value and the model prediction value, so that the forgetting factor changes dynamically within a reasonable range. The strategy of improving the forgetting factor recursive least squares method with real-time voltage error is as follows: Among them, μ t is the forgetting factor at time t; μ min With μ max is the upper and lower limits of the forgetting factor; σ 2 is the variance of the voltage error; e is the voltage difference; α is the smoothing factor. Since the improved forgetting factor recursive least squares method needs to determine the size of the smoothing factor and the upper and lower limits of the forgetting factor to ensure the accuracy and volatility of the parameter identification results, a flexible and simple grid search method is selected to traverse the size of the smoothing factor and the upper and lower limits of the forgetting factor within a certain range to find the optimal parameter combination that minimizes the error and balances the parameter volatility. The mathematical model for selecting the initial parameters of the improved forgetting factor least squares method by grid search is as follows: Among them, the present invention searches and selects ɑ=0.95, μ min =0.95, μ max =1. Before the online parameter identification method is used, the electrical relationship of the established equivalent circuit model needs to be converted, specifically as follows: the electrical relationship of the dual-polarization equivalent circuit model is Laplace transformed to obtain the transfer function; the discretized form of the transfer function can be obtained by using the bilinear transformation method, and then converted into a differential equation form.

6. The mixed pulse operating condition test according to claim 1, characterized in that: The experimental data are obtained by HPPC testing, and the obtained test data are solved based on the online parameter identification method to obtain the parameter values ​​of the equivalent circuit model.

7. The adaptive algorithm according to claim 1, characterized in that: The specific implementation method of the adaptive algorithm is as follows: the extended Kalman filter algorithm is used to linearize the nonlinear model to retain the first-order Taylor expansion term of the nonlinear function, and a noise adaptive module is added on its basis to continuously update the noise and covariance according to data changes. The recursive process of the adaptive extended Kalman filter algorithm is as follows: Among them, e t is the observation error at time, H t is the covariance matrix of the new data information based on the moving average filter with a window size of M, where M is the window size.

Images