A soc estimation method based on driving segment data

Abstract

The application discloses a SOC estimation method based on driving segment data; RLS_OCV is obtained through parameter identification based on segment driving data, the problem of data saturation of the RLS algorithm is solved, and a forgetting factor does not need to be considered; segment data characteristics are easy to be counted based on the segment driving data, and the confidence of each estimation result can be conveniently judged by using the counted characteristic information, so that the SOC is reduced in error correction; the effectiveness of the RLS_OCV result is verified by using online DCR results RLS_DCR and KF_DCR, the false convergence of the RLS algorithm is avoided, and the RLS_OCV estimation result is ensured to be within a certain error range.

Description

Technical Field

[0001] This invention relates to the field of algorithms, and in particular to a SOC estimation method based on driving segment data. Background Technology

[0002] The SOC estimation method based on driving segment data includes modules such as driving segment data feature statistics, online DCR estimation, online OCV estimation, and SOC correction. It addresses the following problems of existing online SOC estimation algorithms based on parameter identification: poor reliability of the estimated SOC, i.e., existing technologies do not address the possible error of each calculation step and whether it is reliable; data saturation is easy during the estimation process, requiring the setting of a forgetting factor, but actual driving conditions have great uncertainty, so it is impossible to fix a forgetting factor; under constant current or certain uncertain operating conditions, it may converge to an incorrect result; online estimation based on rolling windows has large memory space requirements.

[0003] For example, a method for jointly estimating the state of charge, state of health, and state of power of a lithium-ion battery, disclosed in Chinese patent literature (publication number CN105301509B), includes online estimation of the battery's state of health: using a recursive least squares method with a forgetting factor to identify the open-circuit voltage and internal resistance online, and indirectly obtaining the state of charge based on a pre-established OCV-SOC correspondence, and then estimating the battery capacity based on the cumulative charge and discharge capacity between two SOC points; online estimation of the battery's state of charge: based on a second-order RC equivalent circuit model, using a Kalman filter algorithm to estimate the battery's state of charge, and updating the battery capacity parameters in the Kalman filter algorithm based on the battery capacity estimation results; and online estimation of the battery's state of power: based on the internal resistance obtained online, and based on the battery's own voltage and current limits, calculating the maximum chargeable and dischargeable current, and then further calculating the maximum chargeable and dischargeable power. This method uses the Kalman filter algorithm to filter and estimate the SOC. However, in practical applications, under certain specific working conditions, the RLS estimation of battery model parameters may be completely distorted. In addition, for SOC correction, only a portion of the RLS estimation results with high confidence are needed to correct and obtain an accurate SOC. Summary of the Invention

[0004] This invention primarily addresses the shortcomings of existing online SOC estimation methods based on forgetting factors, which cannot meet the accuracy requirements of SOC estimation under any working condition using a fixed forgetting factor. It provides a SOC estimation method based on vehicle segment data. By performing parameter identification on segment vehicle data to obtain RLS_OCV, the data saturation problem of the RLS algorithm is solved, and the forgetting factor is not considered. Using segment vehicle data facilitates the statistical analysis of segment data features, and the statistical feature information can be used to easily determine the confidence level of each estimation result, thereby reducing incorrect SOC corrections. The effectiveness of the RLS_OCV result is verified using online DCR results RLS_DCR and KF_DCR, avoiding false convergence of the RLS algorithm and ensuring that the RLS_OCV estimation result is within a certain error range.

[0005] The above-mentioned technical problems of the present invention are mainly solved by the following technical solutions: A SOC estimation method based on driving segment data, characterized in that the SOC estimation method includes the following steps: Step S1, update the SOC using the ampere-hour integral algorithm, and use the corresponding equation as the state equation; Where Cap is the battery capacity, measured in Ah, and I... k Let be the current at time k, in A or W. k Let Δt be the random noise input at time k, and Δt be the time interval between k and k+1, in seconds. Step S2: Estimate the open-circuit voltage value RLS_OCV based on the fragment RLS algorithm, and use the open-circuit voltage RLS_OCV as the observed value. The relationship between SOC and OCV is used as the observation equation SOC = f(OCV). Step S3: Calculate the confidence voltage error Calib_Thd based on the open-circuit voltage RLS_OCV estimated by the RLS algorithm based on the fragment data. The confidence voltage error Calib_Thd is applied to the SOC state estimation update equation of the Kalman filter algorithm. The update equation is determined by using the valid flag bit Valid_flg of RLS_OCV.

[0006] The RLS_OCV is obtained by parameter identification based on segmented driving data, which solves the data saturation problem of the RLS algorithm. The segmented driving data makes it easy to statistically analyze the features of the segmented data, and the statistical feature information is used to judge the confidence of each estimation result, reducing the false correction of SOC. The effectiveness of the RLS_OCV result is verified by online DCR results RLS_DCR and KF_DCR, avoiding false convergence of the RLS algorithm and ensuring that the RLS_OCV estimation result is within a certain error range.

[0007] Preferably, the confidence voltage error Calib_Thd is applied to the Kalman filter algorithm's SOC state estimation update equation as follows: Among them, C k Let K be the observation matrix. k Here, is the Kalman gain matrix, and Valid_flg is the valid flag for RLS_OCV. The valid flag is used to select the update equation, ensuring that the update equation can effectively utilize statistical feature information to determine the confidence level of each estimation result, thus reducing incorrect corrections to SOC.

[0008] As a preferred method, the calculation process for open-circuit voltage RLS_OCV, confidence voltage error Calib_Thd, and Valid_flg is as follows: Step S21: Set n = 1, Itot = 0, Vtot = 0; Step S22: Determine if n is 0. If n = 0, proceed to step S23; if n is not 0, proceed to step S24. Step S23: Iterate RLS once and output RLS_OCV and RLS_DCR; Calculate Calib_Thd based on the statistical features Itot and Vtot, and reset RLS, Itot, and Vtot to zero; Proceed to step S25. Step S24: RLS iterates once, Itot is updated to the original Itot plus Curr, and Vtot is updated to the original Vtot plus Volt value; at this time, the iteration number n is incremented by one; n is divided by the dt value and the remainder is taken, and the remainder value is assigned to n; Step S25: Determine whether the formula n=0, the formula RLS_DCR<(1+η)*KF_DCR, and the formula RLS_DCR>(1-η)*KF_DCR are all true. If all three are true, then Valid_flg=TURE; otherwise, then Valid_flg=FALSE. Where dt is the number of iterations required to output RLS_OCV and RLS_DCR (DCR calculated based on the RLS algorithm identification results) each time. Generally, the data segment length is 60s, and the algorithm scheduling cycle is 0.1s, so dt = 60 / 0.1 = 600; Curr is the current current, Volt is the current voltage, KF_DCR is the DCR calculated based on Delta current and voltage and Kalman filtering, and η is the threshold for the fluctuation range of DCR estimation results based on different methods (generally taken as 0.2); based on the DCR calculated from the RLS algorithm identification results, the open-circuit voltage RLS_OCV is obtained using the algorithm, the confidence voltage error Calib_Thd is obtained, the identifier number Valid_flg is calculated according to the algorithm flow, and the data is substituted into the formula to obtain the calculation result.

[0009] As a preferred method, the electrochemical polarization voltage Vp_Seg is calculated by using the first feature Itot, the second feature Vtot, and RLS_OCV, as follows: Vp_Seg=Vavg_Seg-Iavg_Seg*KF_DCR; Where Itot is the cumulative current Itot within the segment time; Iavg_Seg is the average current of the segment; and Vavg_Seg is the average voltage of the segment. Based on a large amount of test statistics, it was found that the larger the battery electrochemical polarization voltage, the larger the RLS_OCV error identified by the RLS algorithm. Therefore, the first feature Itot, the second feature Vtot, and RLS_OCV are used to inversely deduce the electrochemical polarization voltage Vp_Seg.

[0010] Preferably, the formula for calculating Calib_Thd is as follows: Calib_Thd=min{f Curr (Iavg_Seg), f Volt (yp_Seg), e k}; Among them, e k To identify voltage estimation errors in the RLS_OCV process, specific data information can be quickly obtained based on the formula.

[0011] As a preferred option, f Curr and f Volt The functional relationship between the eigenvalues ​​of Iavg_Seg and Vp_Seg obtained from statistical analysis and Calib_Thd is as follows: Where (LoIavg, LoIavgCalib) and (HiIavg, HiIavgCalib) represent the possible errors corresponding to different average segment currents calibrated under urban and high-speed operating conditions; (LoVp, LoVpCalib) and (HiVp, HiVpCalib) represent the possible errors corresponding to different average voltages of electrochemically polarized segments calibrated under urban and high-speed operating conditions. Selecting error ranges according to different cities can minimize the magnitude of errors and meet the SOC estimation accuracy requirements for any operating condition.

[0012] Preferably, the KF_DCR estimation process is as follows: Step S71: Assign the value of DCR_Ini to KF_DCR, set the value of Np to zero, set StartI = I(k), and set StartU = Volt(k); Step S72: Determine whether I(k)-I(k-1)>DeltaI is true; if true, proceed to step S73; if false, proceed to step S74. Step S73: Increment the value of Np by one; assign the value of KF_DCR(k-1) to KF_DCR(k); repeat step S72; Step S74: Determine if Np > DeltaIThd is true; if true, then... Set the value of Np to zero, set StartI = I(k), StartU = VolT, update KF_DCR(k) using the KF filtering algorithm, and jump to step S72; if not, assign the value of KF_DCR(k-1) to KF_DCR(k), set the value of Np to zero, set StartI = I(k), StartU = Volt(k), and jump to step S72.

[0013] The DCR is calculated in real time by detecting changes in incremental current and corresponding voltage. The DCR estimation result is smoothed using a filtering algorithm (Kalman filter) to obtain KF_DCR. The effectiveness of the RLS_OCV result is verified using online DCR results (RLS_DCR and KF_DCR), which can avoid false convergence of the RLS algorithm.

[0014] Preferably, DCR_Ini is the initial value of DCR, I(k) and Volt(k) are the current and voltage at time k, respectively, DeltaI is the Delta current threshold for determining whether the pulse ends at time k, Np is the number of times the Delta current threshold is continuously satisfied, and DeltaIThd is the Np condition threshold for determining the number of times the Delta current is continuously satisfied; the state equation in the KF filter update DCR algorithm is KF_DCR(k) = KF_DCR(k-1), and the observation equation is... Verifying the validity of the RLS_OCV results using online DCR results (RLS_DCR and KF_DCR) can avoid false convergence of the RLS algorithm and ensure that the RLS_OCV estimation results are within a certain error range.

[0015] The beneficial effects of this invention are: 1. To address the limitation that existing online identification SOC estimation methods based on forgetting factors cannot meet the SOC estimation accuracy requirements of any operating condition with a fixed forgetting factor; 2. To address the difficulty of existing algorithms meeting the reliability requirements of automotive applications, reliability is improved by increasing confidence and validity assessments to avoid erroneous corrections and convergence. 3. Requires only a small amount of memory, making it easy to implement on low-cost embedded hardware. Attached Figure Description

[0016] Figure 1 Here is a block diagram illustrating the algorithm's principle. Figure 2 The flowchart for calculating RLS_OCV, Calib_Thd, and Valid_flg; Figure 3 Flowchart for KF_DCR estimation; Figure 4 This is a diagram of a first-order RC model. Detailed Implementation

[0017] It should be understood that the embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined by the appended claims.

[0018] The technical solution of the present invention will be further described in detail below through embodiments.

[0019] A method for estimating SOC based on driving segment data, such as Figure 1 As shown, the basic principle of Kalman filtering for SOC correction in this invention is as follows: First, the update of SOC by the ampere-hour integral algorithm is used as the state equation; Where Cap is the battery capacity, measured in Ah, and I... k Let be the current at time k, in A or W. k Let Δt be the random noise input at time k, and let Δt be the time interval between k and k+1, in seconds.

[0020] Second, the open-circuit voltage value RLS_OCV estimated based on the fragment RLS algorithm is used as the observation value, and the relationship between SOC and OCV is used as the observation equation SOC=f(OCV). This observation equation needs to be obtained through offline testing.

[0021] Third, Calib_Thd is the confidence voltage error of the open-circuit voltage RLS_OCV estimated by the RLS algorithm based on fragment data. This confidence voltage error is applied to the SOC state estimation update equation of the Kalman filter algorithm: Where C k Let K be the observation matrix. k is the Kalman gain matrix, and Valid_flg is the flag indicating whether RLS_OCV is valid.

[0022] In the above steps, the SOC state estimation update differs from the standard Kalman filter algorithm, while the other steps are the same.

[0023] The calculation flowcharts for RLS_OCV, Calib_Thd, and Valid_flg are as follows: Figure 2 As shown: In the diagram, dt represents the number of iterations required to output RLS_OCV and RLS_DCR (the DCR calculated based on the RLS algorithm identification results). Typically, the data segment length is 60 seconds, and the algorithm scheduling cycle is 0.1 seconds, so dt = 60 / 0.1 = 600. Curr represents the current current, Volt represents the current voltage, KF_DCR is the DCR calculated based on Delta current and voltage and Kalman filtering, and η is the threshold for the fluctuation range of DCR estimation results based on different methods (typically taken as 0.2). Using a first-order RC model, such as Figure 4 The diagram illustrates the principle of the RLS algorithm for estimating RLS_OCV and RLS_DCR.

[0024] The mathematical formula for the first-order RC model is as follows: To meet the iteration requirements of the RLS algorithm, the mathematical formula of the first-order RC model needs to be discretized, as follows: make: The following RLS algorithm is used to identify the parameter θ. k The recursive formula is as follows: When the number of iterations equals dt, the output RLS_OCV = U ocv，k , RLS_DCR=R0+R1.

[0025] The principle of calculating Calib_Thd based on voltage and current statistical characteristics is as follows: Based on a large amount of test statistical data, it was found that the larger the battery electrochemical polarization voltage, the larger the RLS_OCV error identified by the RLS algorithm. In this invention, the cumulative current Itot within a segment time is used as the first feature, and Vtot is used as the second feature. The electrochemical polarization voltage Vp_Seg is inversely calculated using this feature and RLS_OCV. The calculation method is as follows: Vp_Seg=Vavg_Seg-Iavg_Seg*KF_DCR; Where Iavg_Seg is the average current of the segment, and Vavg_Seg is the average voltage of the segment.

[0026] The formula for calculating Calib_Thd is as follows: Calib_Thd=min{f Curr (Iavg_Seg), f Volt (Vp_Seg), e k}; Where e k To identify the voltage estimation error obtained during the RLS_OCV process.

[0027] f Curr and f Volt The functional relationship between the eigenvalues ​​of Iavg_Seg and Vp_Seg obtained from statistical analysis and Calib_Thd is as follows: Where (LoIavg, LoIavgCalib) and (HiIavg, HiIavgCalib) are the possible errors corresponding to the different average segment currents calibrated according to urban and high-speed operating conditions; (LoVp, LoVpCalib) and (HiVp, HiVpCalib) are the possible errors corresponding to the different average voltages of the electrochemically polarized segments calibrated according to urban and high-speed operating conditions.

[0028] The DCR is calculated in real time by detecting changes in incremental current and corresponding voltage. Then, the DCR estimation result is smoothed using a filtering algorithm (Kalman filtering) to obtain KF_DCR. The KF_DCR estimation flowchart is shown below. Figure 3 .

[0029] In the flowchart, DCR_Ini is the initial value of DCR (DCR at the last power-down), I(k) and Volt(k) are the current and voltage at time k, respectively, DeltaI is the Delta current threshold for whether the pulse at time k has ended, Np is the number of times the Delta current threshold is continuously met, and DeltaIThd is the Np condition threshold for judging the number of times the Delta current is continuously met. In the above KF filter update DCR algorithm, the state equation is KF_DCR(k)=KF_DCR(k-1), and the observation equation is DCR=(Volt(k)-StartU) / (I(k)-StartI). This filtering process is a standard Kalman filter.

Claims

1. A SOC estimation method based on driving segment data, characterized in that, The SOC estimation method includes the following steps: S1. Update the SOC using the ampere-hour integral algorithm, and use the corresponding equation as the state equation; ， Where Cap is the battery capacity, measured in Ah. Let be the current at time k, in amperes (A). Let k be the random noise input at time k. The time interval between k and k+1 is in seconds. S2. Estimating open-circuit voltage based on fragment RLS algorithm. and open circuit voltage As observed values, the relationship corresponding to SOC-OCV is used as the observation equation. ; Confidence voltage error and The calculation process is as follows: S21. Set n=1, Itot=0, Vtot=0; S22. Determine if n is 0. If yes, jump to S23; if n is not, jump to S24. S23, Iterate once and output and ; For based on DCR calculated using current, voltage, and Kalman filtering; based on statistical characteristics and calculate and reset to zero. , and Jump to S25; S24, One iteration, Updated to the original Apply current ,and Updated to the original Apply current voltage Value; at this point, the iteration count n is incremented by one; n is divided by The value is calculated and the remainder is taken, and the remainder value is assigned to n. For each output and The number of iterations required; S25. Determine if all formulas in the set are true. If so, then... ;if not, The formula group includes n=0, , ; S3. Open-circuit voltage estimated by RLS algorithm based on fragment data. Calculate the confidence voltage error Confidence voltage error Applying to the SOC state estimation update equation of the Kalman filter algorithm, using Valid identifier Determine the update equation.

2. The SOC estimation method based on driving segment data according to claim 1, characterized in that: The confidence voltage error The state estimation update equation for the Kalman filter algorithm is as follows: ， in, For the observation matrix, Here is the Kalman gain matrix. for Is it a valid identifier? 3. The SOC estimation method based on driving segment data according to claim 1, characterized in that: The threshold for the fluctuation range of DCR estimation results based on different methods.

4. The SOC estimation method based on driving segment data according to claim 1, characterized in that: The electrochemical polarization voltage is inversely calculated using the first feature Itot, the second feature Vtot, and RLS_OCV. The calculation formula is as follows: ； ； ； in, This represents the cumulative current over a given time interval. The average current of the segment; This represents the average voltage of the segment.

5. The SOC estimation method based on driving segment data according to claim 1, characterized in that: The The calculation formula is as follows: ， in, To identify voltage estimation errors in the RLS_OCV process.

6. The SOC estimation method based on driving segment data according to claim 5, characterized in that: and This is based on the analysis of statistical results. and Eigenvalues ​​and The functional relationship between them is as follows: ， ， in, and This represents the possible errors corresponding to different average segment currents calibrated for urban and high-speed operating conditions. and This represents the possible error corresponding to the average voltage of different electrochemical polarization segments calibrated for urban and high-speed operating conditions.

7. The SOC estimation method based on driving segment data according to claim 1, characterized in that: The The estimation process is as follows: Step S71, will The value assigned and will The value is set to zero, let And make ; Step S72, Judgment Is the condition met? If it is met, proceed to step S73; if it is not met, proceed to step S74. Step S73 Increment the value by one; The value assigned Repeat step S72; Step S74, Judgment Is it true? If it is true, then... ,make The value is set to zero. The value assigned ,Will The value assigned , Filtering algorithm update And jump to step S72; if it is not true, then The value assigned ,make The value is set to zero. The value assigned ,Will The value assigned And jump to step S72; the for The initial value, the and stated Let be the current and voltage at time k, respectively. Does the pulse end at time k? Current threshold, the To continuously satisfy the current The number of times the current threshold is reached, the To determine continued satisfaction Number of currents Conditional threshold; the state equation in the KF filter update DCR algorithm is The observation equation is .

Images