JP2015224919A - Device for estimating parameters of equivalent circuit for vehicle secondary battery - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a device for estimating parameters of an equivalent circuit model for a vehicle secondary battery, capable of highly accurately estimating parameters of the equivalent circuit model during actual running of a vehicle.SOLUTION: An arithmetic unit divides measured voltage values v(t) acquired after change timing at which a measured current value i(t) transiently changes into those in a high frequency region RN1 and those in a low frequency region RN2, performs thinning-out processing on the measured voltage values v(t) in the region RN2, and estimates parameters of an equivalent circuit model 9 by a predetermined fitting method.

Description

  The present invention relates to a parameter estimation device for an equivalent circuit of a vehicular secondary battery.

  A vehicle such as a hybrid vehicle (HV) or an electric vehicle (EV) is equipped with a secondary battery such as a lithium ion battery. The state of charge (SOC) of the secondary battery for a vehicle is an index used to calculate the travelable distance and the like, and the state of charge SOC of the secondary battery for the vehicle is estimated sequentially. This is important because it greatly contributes to the reduction of the capacity.

  In order to estimate the state of charge SOC of a vehicular secondary battery, a method using an equivalent circuit model has been studied. In this estimation method, the battery is expressed by an equivalent circuit model of an electric circuit, and the waveform is fitted according to the time response (charge / discharge characteristics) of the voltage when the current is changed transiently (for example, stepped). This is a method for estimating (identifying) parameters of an equivalent circuit.

  The response waveform is equivalent to the discharge waveform of the RC parallel circuit and becomes an exponential response. For this reason, as this fitting method, for example, there is a method of estimating (identifying) a parameter using a nonlinear least square method such as an LM (Levenberg Marquardt) method. Here, in the nonlinear least square method, the actual measured waveform and the pseudo response waveform based on the estimated parameter are compared in each sampling data, and the parameter is set so as to minimize the sum Σe ^ 2 of the square of the error e. It is a method to adjust. As a technique related to the present application, there is provided a technique for estimating a state of charge (charging rate) based on a convergence value of an open circuit voltage (see, for example, Patent Document 1).

Japanese Patent Laying-Open No. 2005-43339 (Japanese Patent No. 4015128)

FIG. 10 shows an example in which the detection voltage is sampled at equal intervals according to a general control method. FIG. 10 shows the transient response characteristics of the voltage when the current decreases in a stepped manner. As shown in FIG. 10, the sampling data D has an overwhelmingly larger number of data in the low frequency region RN2 than the number of data in the high frequency region RN1. Therefore, when the sum Σe ² of the squares of the errors e is minimized and the fitting function F0 shown in FIG. 10 is obtained, the parameters can be accurately estimated (identified) in the low frequency region RN2, but the parameter estimation accuracy is relative in the high frequency region RN1. Will be reduced. Further, since the sampling period needs to be as long as the time constant of the desired RC circuit to be estimated, a long time is particularly required in the low frequency region RN2. However, when estimating (identifying) parameters during actual driving, there is a possibility that signals cannot be acquired for a long time.

  An object of the present invention is to provide a parameter estimation device for an equivalent circuit model of a secondary battery for a vehicle that can estimate the parameters of the equivalent circuit model with high accuracy during actual traveling of the vehicle.

  According to the first aspect of the present invention, the estimation unit determines effective data by partially thinning out measurement voltage data acquired after a change timing at which the measurement current data measured by the current value measurement unit changes transiently. The parameters of the equivalent circuit model are estimated by using a predetermined fitting method using the effective data. As in the first aspect of the invention, after the effective data is determined by partially thinning the measured voltage data, the fitting function as accurate as possible can be obtained by estimating the parameters of the equivalent circuit model by a predetermined fitting method. In addition, the parameters of the equivalent circuit model can be estimated with high accuracy during actual driving of the vehicle.

  According to the eighth aspect of the present invention, the estimation unit sets an error weighting factor for measurement voltage data acquired after a change timing at which the current measurement data measured by the current value measurement unit changes transiently, and the weighting factor is set. The parameter of the equivalent circuit model is estimated by a predetermined fitting method using the error weighted according to the above. If an error weighting coefficient is set for the measured voltage data as in the eighth aspect of the invention, and the error weighted according to the weighting coefficient is used, the parameters of the equivalent circuit model are estimated by a predetermined fitting technique. Thus, the fitting function as accurate as possible can be obtained, and the parameters of the equivalent circuit model can be estimated with high accuracy during actual running of the vehicle.

1 is an electrical configuration diagram schematically illustrating an example of a system block configuration of an apparatus for estimating an equivalent circuit parameter of a secondary battery for a vehicle in one embodiment. A diagram schematically showing an example of an equivalent circuit model of a secondary battery A flowchart schematically showing a method for estimating the state of charge Timing chart that represents the current and voltage changes of the equivalent circuit model in principle Image example for classifying into high frequency region / low frequency region Explanatory drawing which shows an example of an estimation method roughly (the 1) Explanatory drawing which shows an example of the estimation method roughly (the 2) Explanatory diagram schematically showing an example when the number of data acquisition is small Characteristic diagram schematically showing open circuit voltage vs. state of charge Explanatory drawing which shows the comparative example of an estimation method roughly

Hereinafter, an embodiment of the present invention will be described. FIG. 1 is a schematic block diagram showing an electrical configuration of an apparatus 1 for estimating an equivalent circuit parameter of a vehicle secondary battery according to the present embodiment.
A load 2 by a main motor is mounted in the vehicle. The load 2 is supplied with power from a cell group of the secondary battery 3. The secondary battery 3 is a lithium ion battery or the like. The voltage value measuring unit 4 is provided for measuring the voltage value of the cell of the secondary battery 3, and the temperature measuring unit 5 is provided for measuring the temperature of the secondary battery 3. A current value measuring unit 7 is provided for measuring the current flowing through the load 2.

  The calculation unit 8 is configured using, for example, a microcomputer. The calculation unit 8 includes, for example, an A / D conversion unit 8 a and a memory 8 b in a microcomputer. The voltage information of the voltage value v (t) measured by the voltage value measurement unit 4 and the temperature measured by the temperature measurement unit 5. The temperature information of T (t) and the current information of the current value i (t) measured by the current value measurement unit 7 are acquired through the A / D conversion unit 8a, and the measurement voltage data, the measurement temperature data, The measured current data is stored in the memory 8b. The computing unit 8 computes parameter values of the equivalent circuit model 9 of the secondary battery 3 using at least some or all of these values.

  A method for estimating the state of charge SOC of the secondary battery 3 in the above configuration will be described with reference to FIGS. The flowchart shown in FIG. 3 schematically shows the processing contents executed by the arithmetic unit 8 in response to the timer interrupt for each cycle T0, and the processing performed in the environment where the estimation device 1 is mounted in the vehicle. Indicates. The period T0 indicates the sampling interval of the current value i (t) by the current value measuring unit, the voltage value v (t) by the voltage value measuring unit, and the temperature T (t) by the temperature measuring unit. FIG. 2 schematically shows an equivalent circuit model 9.

In the equivalent circuit model 9 of the secondary battery 3 shown in FIG. 2, V _OCV (OCV: Open Circuit Voltage) indicates an open circuit voltage. The open circuit voltage V _OCV indicates the potential difference between the electrodes in an electrochemical equilibrium state. An equivalent circuit model 9 of the secondary battery 3 shown in FIG. 2 includes a pure voltage element 10 of the secondary battery 3 and an internal impedance element 11 between terminals. The internal impedance element 11 includes a resistor R0 and an RC parallel circuit group 12 that is connected in series to the resistor R0, and capacitors C1... Cn and resistors R1. The resistor R0 is a zero-order component resistor. When the secondary battery 3 is represented by the equivalent circuit model 9 shown in FIG. 2, a reaction in which the time constants τ1, τ2,..., N of the parallel circuits 12a of the RC parallel circuit group 12 are mixed is shown. Strictly speaking, there are inductor components and the like, but they are not shown because they are negligible, for example. Here, the current value measuring unit 7 can measure the current flowing to the secondary battery 3 by measuring the current value I flowing from the secondary battery 3 to the load 2, and the voltage value measuring unit 4 can measure the current of the secondary battery 3. The voltage value between both terminals can be measured.

  The calculation unit 8 performs the process shown in FIG. 3 at every constant sampling period T0 in order to estimate the state of charge SOC of the secondary battery 3. The memory 8b has a storage area for storing a “target data storage flag”. As will be described in detail later, the calculation unit 8 determines whether or not the target data is in a transiently changed timing by changing the target data storage flag. The target data storage flag has its initial value set to off.

  First, the calculation unit 8 acquires the current value i (t) measured by the current value measurement unit 7 through the A / D conversion unit 8a, and the voltage value v (t) measured by the voltage value measurement unit 4 is A The temperature T (t) acquired through the / D conversion unit 8a and measured by the temperature measurement unit 5 is acquired through the A / D conversion unit 8a and sequentially stored in the memory 8b (S1).

  The calculation unit 8 first analyzes the current value i (t) stored in the memory 8b under the condition that the target data storage flag is off (S2: YES), and determines the current value i (t). It is determined whether or not has changed rapidly (transiently) (S3). At this time, for example, the calculation unit 8 determines whether or not the degree of change in the current value i (t) is equal to or greater than a predetermined first gradient, whereby the current value i (t) changes rapidly (transiently). It is good to determine whether or not. For example, when the vehicle is stopped, the current flowing through the load 2 is minimal. Then, since the current value i (t) becomes substantially constant, the calculation unit 8 determines NO in step S3 and exits the process shown in FIG.

  For example, when the vehicle suddenly starts while the vehicle is stopped, a large current flows through the driving load 2, so that the current increases rapidly and the degree of increase in the current value i (t) increases. On the equivalent circuit model 9, the capacitors C1... Cn are quickly charged. On the other hand, when the vehicle stops waiting for a signal, or when an ignition key switch (not shown) is turned off by the user and the engine is stopped, the current flowing through the driving load 2 is rapidly increased. The current value i (t) decreases and increases. On the equivalent circuit model 9, the charges accumulated in the capacitors C1... Cn are rapidly discharged. At this time, the calculation unit 8 determines that the change has occurred transiently in Step S3 and determines YES.

  When the condition of step S3 is satisfied, the calculation unit 8 determines whether or not the current value i (t) falls within a predetermined range during a predetermined period before the change timing of the current value i (t). Determine (S4). This determination condition indicates a processing condition for determining whether or not the current value i (t) is substantially constant during a predetermined period before the change timing.

  In the equivalent circuit model 9, different time constants τ1... Τn are determined for each RC parallel circuit 12a connected in series. The determination condition in step S4 is a process provided to determine whether or not the acquired data is appropriate as an analysis process prior to the parameter estimation process in the equivalent circuit model 9.

  The determination condition of step S4 is provided, for example, for determining whether or not the accumulated charge amount of the capacitors C1... Cn is settled to a charge amount within a predetermined range on the equivalent circuit model 9. Specifically, the determination condition in step S4 is, for example, that the capacitor (for example, Cn) of the RC parallel circuit 12a having the maximum time constant (for example, τn) among the time constants τ1. It is provided to determine whether or not the accumulated charge has settled within a predetermined range.

  When the determination condition in step S4 is not satisfied, the calculation unit 8 determines that the data is unsuitable for parameter estimation processing because the final accumulated charge amount of the capacitor of the equivalent circuit model 9 is unknown, and the estimation processing shown in FIG. Exit.

  When determining that the condition of step S4 is satisfied, the arithmetic unit 8 turns on and holds the target data storage flag in the memory 8b (S5). Then, the calculation unit 8 stores the current value i (t), the voltage value v (t), and the temperature T (t) that are captured thereafter as “target data” in the memory 8b (S6).

  Then, the calculation unit 8 determines whether or not the number of target data acquired after the change timing is equal to or greater than the first predetermined number (S7). For example, if the calculation unit 8 has acquired a data number equal to or greater than a first predetermined number (for example, several hundreds of seconds when converted to time), it is determined that the determination condition of step S7 is satisfied. The determination condition in step S7 is provided to determine whether or not the transient response characteristic is sufficiently settled in a steady state.

  For example, when a minimum current (≈0 A) is obtained after sufficient charges are accumulated in the capacitors C1... Cn in the equivalent circuit model 9, the determination condition in step S7 is that the time constant τ1. It is provided to determine whether or not the charge of the capacitor (for example, Cn) of the RC parallel circuit 12a having the maximum time constant (for example, τn) has been discharged and settled to a charge amount within a predetermined range.

  FIG. 4 is a timing chart showing in principle the current change and voltage change of the equivalent circuit model 9. FIG. 4 shows, as a reference example, a step response when the current value i (t) rapidly decreases to 0 [A] from the timing when the current value i (t) is a predetermined value for a certain time. The transient response characteristic of the equivalent circuit model 9 in this case can be expressed as the following equation (1) in principle when the current change timing is set to time zero.

このとき時間経過しｔ→０〜∞とすると、定常状態（Ｖ＝Ｖ_ＯＣＶ）に至るまでの差電圧ΔＶは下記（２）式のように表すことができる。

If the time elapses and t → 0 to ∞, the differential voltage ΔV until the steady state (V = V _OCV ) is reached can be expressed by the following equation (2).

この差電圧ΔＶは、等価回路モデル９の直流抵抗成分（Ｒ０＋Ｒ１＋…＋Ｒｎ）にかかる電圧となる。この直流抵抗成分（Ｒ０＋Ｒ１＋…＋Ｒｎ）にかかる電圧を算出することが等価回路モデル９のパラメータを算出するために重要となる。ステップＳ７の判定条件は、過渡応答特性が十分に定常状態に落ち着いていることを条件としているものである。演算部８は、このステップＳ７の判定条件を満たすと判定したときには、対象データ記憶フラグをオフとし（Ｓ８）、後述のステップＳ１２の処理に移行する。

This difference voltage ΔV is a voltage applied to the DC resistance component (R0 + R1 +... + Rn) of the equivalent circuit model 9. It is important to calculate the voltage applied to the DC resistance component (R0 + R1 +... + Rn) in order to calculate the parameters of the equivalent circuit model 9. The determination condition in step S7 is that the transient response characteristic is sufficiently settled in a steady state. When the calculation unit 8 determines that the determination condition of step S7 is satisfied, the calculation unit 8 turns off the target data storage flag (S8), and proceeds to the process of step S12 described later.

  In step S7, when the number of target data stored in the memory 8b is less than the first predetermined number (S7: NO), the calculation unit 8 rapidly (transiently) changes the current value i (t) again. It is determined whether or not (S9). The change direction of the current value i (t) may be an increase direction or a decrease direction. At this time, for example, the calculation unit 8 determines whether or not the current value i (t) has changed rapidly again (transiently) by determining whether or not the degree of change in the current value i (t) is equal to or greater than a predetermined second gradient. good. When the vehicle restarts from a state waiting for a signal, the current value i (t) flowing through the load 2 increases rapidly. Also, the current value i (t) flowing through the load 2 increases rapidly even when the vehicle suddenly accelerates from the normal traveling state.

  On the other hand, when the vehicle stops from the normal traveling state, the current value i (t) flowing through the load 2 rapidly decreases. In addition, the current value i (t) flowing through the load 2 rapidly decreases as the vehicle decelerates, and after it is determined YES in step S3, the current value i (t) flowing through the load 2 when the vehicle further stops. ) Is further decreased, YES is determined in step S9. The determination condition of step S9 is provided to determine the timing at which the current value i (t) rapidly changes in this way. The calculation unit 8 exits the process of FIG. 3 when the determination condition of step S9 is not satisfied.

  The calculation unit 8 returns to step S1 of FIG. 3 in the cycle T0 and repeats the process, and continuously acquires the measured values of the current value i (t), the voltage value v (t), and the temperature T (t) at intervals of the cycle T0. . At this time, once the target data storage flag is turned on, the calculation unit 8 determines NO in step S2, continues to store the measurement value as target data in step S6, and until the determination condition of step S7 or S9 is satisfied. These steps S1, S2, S6, S7, and S9 are repeated.

  The calculation unit 8 is on condition that the current value i (t) changes rapidly (transiently) again until the first predetermined number or more of the target data is acquired (S7: NO, S9: YES). The target data storage flag is turned off (S10). Then, the calculation unit 8 determines whether or not the number of data in the low frequency region (number of low frequency data) after the change timing in step S3 is greater than a second predetermined number compared to the number of data in the high frequency region (number of high frequency data). Determine (S11).

  When the sampling current value i (t) changes rapidly (transiently), the current value i (t) contains a large amount of high frequency components. After the rapid change, the voltage does not reach the steady state. It changes slowly, and many low frequency components are included in the subsequent time domain. Therefore, a time region immediately after the change timing can be considered as a high frequency region, and a time region after a certain timing thereafter can be considered as a low frequency region.

  In the equivalent circuit model 9, initial values of time constants τ1... Τn (resistance value × capacitance value of the capacitor) of the RC parallel circuit 12a are set in advance by experiment or simulation. Therefore, the high frequency region / low frequency region is divided in advance according to the size of the initial value of the time constants τ1... Τn of the RC parallel circuit 12a of the equivalent circuit model 9.

  FIG. 5A and FIG. 5B show examples of images for classifying into a high frequency region / low frequency region. In the present embodiment, as shown in FIG. 5 (a), the high frequency region RN1 / low frequency region RN2 is divided into two time regions, but it is not necessarily divided into two regions RN1 and RN2. As shown in FIG. 5 (b), for example, it may be divided into three or more regions RN1 to RN3, or may be divided into separate n regions for every time constant τ1... Τn of all RC parallel circuits 12a. . If two or more of the time constants τ1... Τn are within a predetermined time, the areas corresponding to the two or more time constants may be divided into the same area.

  When the arithmetic unit 8 samples the current value i (t) or the like at every predetermined period T0, the number of data in the low frequency region when the frequency changes gradually after that is larger than the number of data in the high frequency region immediately after the change timing. Become more.

  The processing in step S11 is provided to divide the processing contents according to the number of acquired data in the high frequency region and the low frequency region and their relationship. The process of step S11 is a process for determining in advance in the post-processing steps (S13 to S14) whether or not the data for the low frequency region RN2 is satisfactory even if the data is thinned out.

  In this processing method, the number of data changes according to the data thinning method in steps S12 and S13 described later. For example, when it is predetermined that the number of data in the low frequency region RN2 is thinned in half. Is preferably determined to satisfy the condition of step S11 when there is at least twice as many data as the number of data in the high-frequency region RN1.

  The calculation unit 8 estimates the parameters of the equivalent circuit model 9 in accordance with the estimation methods A to C according to the determination results of the number of data determination condition steps S7 and S11 (S12 to S14). To do.

  If the number of target data is greater than or equal to the first predetermined number in step S7, the computing unit 8 estimates the parameters of the equivalent circuit model 9 using the estimation method A (S12). As schematically shown in FIG. 6A, the calculation unit 8 equalizes the ratio of the number of data in the high-frequency region RN1 and the low-frequency region RN2 by thinning out sampling data in the low-frequency region RN2 in particular. To do. Here, the term “equivalent” indicates the ratio in which the number of data in each region is the same or the number of data is added to or subtracted from a predetermined margin.

  FIG. 10 shows an example in which the estimation process is performed without thinning out the data D in the low frequency region RN2. As shown in FIG. 10, when the estimation process is performed without thinning out the data D of the low-frequency region RN2, it has been found by the inventor's simulation that the fitting accuracy in the high-frequency region RN1 deteriorates (FIG. 10). 10 (see region RN1a in FIG. 10). This is because if the number of data in the low frequency region RN2 increases, the data in the low frequency region RN2 is easily affected, and the high frequency region RN1 cannot be fitted well.

  Therefore, in the present embodiment, the calculation unit 8 corrects the ratio between the number of data in the low frequency region RN2 and the number of data in the high frequency region RN1. For example, the calculation unit 8 makes the number of data in the low frequency region RN2 and the number of data in the high frequency region RN1 substantially the same number (within the same number or the number of additions / subtractions of the predetermined margin), thereby fitting accuracy within the entire frequency region. To keep in good condition.

  The data decimation method may be simply a predetermined time interval, but is low at time intervals (sampling intervals) T1, T2, T3,... According to a function that regularly increases monotonically (for example, a logarithmic function or a square root function). It is preferable to leave the data in the frequency region RN2 as valid data and discard the remaining data. Then, as shown in FIG. 6A, the valid data D can be determined by thinning out the data in the low frequency region RN2 so that the number of data decreases with time.

  Then, the calculation unit 8 estimates the parameters of the equivalent circuit model 9 by the nonlinear least square method using the effective data D after the thinning process is performed. When the calculation unit 8 estimates the parameters (resistance value, capacitance value) of the fitting function so as to minimize the square of the error with the fitting function, the data of almost the same number of data in the low frequency region RN2 and the high frequency region RN1. The parameters can be calculated so as to be within a substantially uniform error range in all regions.

  Further, if the calculation unit 8 performs processing so as to leave the data at the time intervals T1, T2, T3,... According to the above-described function to obtain effective data D, and thin out the data between them, the fitting parameter of the data in the low frequency region RN2 The parameters can be calculated so that the error range is substantially uniform between the high frequency region RN1 and the low frequency region RN2. FIG. 6B shows the valid data D and the fitting function F1 after processing at this time. Even in the low frequency region RN2, the degree of fitting in the high frequency region RN1 (particularly, the region RN1a) can be increased while keeping the error E1 sufficiently low in the extremely low frequency region.

  When it is determined that the number of data in the low frequency region RN2 is greater than the second predetermined number compared to the number of data in the high frequency region RN1, the calculation unit 8 estimates the parameters of the equivalent circuit model 9 using the estimation method B ( S13 in FIG.

  As schematically shown in FIG. 7, when the estimation method B is used, the calculation unit 8 reduces the sampling data of the low frequency region RN2b that has reached the low frequency region RN2 among the acquired data. However, a large amount of sampling data in the predetermined minimum frequency region RN2c is left. In this case, for example, the calculation unit 8 thins out the data D of the low frequency region RN2b that extends from the high frequency region RN1 to the low frequency region RN2 and sets the residual data as valid data D, for example, the lowest frequency region RN2c (near the final data) ) Is made valid data D without being thinned out. Then, the effective data D in the lowest frequency region RN2c can be made particularly dense. Similarly, in the case of the estimation method B, it is preferable that the ratio of the number of data in the high frequency region RN1 and the low frequency region RN2 is substantially equal.

  As described above, calculating the voltage ΔV applied to the DC resistance component (R0 + R1 +... + Rn) of the equivalent circuit model 9 is particularly important for calculating the parameters of the equivalent circuit model 9. For this reason, it is important to calculate the fitting parameter so that the voltage ΔV applied to the DC resistance component (R0 + R1 +... + Rn) can be calculated as accurately as possible.

  If the arithmetic unit 8 determines that only the number of data less than the first predetermined number can be acquired, for example, if the data D in the low frequency region RN2 is thinned monotonously as shown in FIG. There is a possibility that accurate fitting to the DC resistance component (very low frequency region) may not be possible. Therefore, the calculation unit 8 leaves a large amount of sampling data near the lowest frequency region RN2c in step S13. Then, it becomes possible to fit the DC resistance component (very low frequency region) as accurately as possible, and the parameter estimation accuracy can be improved.

  When the calculation unit 8 determines that the number of data in the low frequency region RN2 is not sufficiently larger than the number of data in the high frequency region RN1 and the number of data equal to or greater than the second predetermined number is not acquired, the calculation unit 8 uses the estimation method C. The parameters of the equivalent circuit model 9 are estimated (S14 in FIG. 3). In such a case, as shown in FIG. 8A, the absolute number of all the target data in the regions RN1 and RN2 itself is reduced, and the number of data in the low frequency region RN2 is extremely small, so that the data in the low frequency region RN2 It is not desirable to thin out D. However, in the simulation performed by the inventors, if the fitting function F2 is drawn as it is, the error E2 becomes large even at a very low frequency even in the low frequency region RN2, as indicated by a broken line in FIG. 8B.

Therefore, the calculation unit 8 increases the weighting coefficient k of the data in the low frequency region RN2 when calculating the sum of squares of errors, for example. For example, to simplify the explanation, the total number of target data is 14 points, and e1 to e14 are assumed to be errors between the fitting function and each sampling data. Here, it is assumed that errors in the high frequency region RN1 are e1 to e7, and errors in the low frequency region RN2 are e8 to e14. Then, as shown in the following equation (3), the sum of error squares Σe ² is

とし、高周波領域ＲＮ１側では重み付け係数を１とし、低周波領域ＲＮ２では重み付け係数ｋについて１を超える所定数とする。

The weighting coefficient is 1 on the high frequency region RN1 side, and the weighting coefficient k is a predetermined number exceeding 1 on the low frequency region RN2.

  Then, the error in the low-frequency region RN2 can be focused on the fitting function. In this example, only one weighting coefficient k is used, but different weighting coefficients (for example, k1, k2) may be used for the high frequency region RN1 and the low frequency region RN2. In this case, the plurality of weighting coefficients k1 and k2 are preferably set such that the weighting coefficient k2 on the low frequency region RN2 side is set higher than the weighting coefficient k1 on the high frequency region RN1 side.

  Then, the arithmetic unit 8 uses a nonlinear least square method, compares the processed effective data D with the pseudo response waveform by the parameter, and performs fitting so that the sum of the squares of the errors becomes a minimum value. Estimate the parameters of

  In this example, the calculation unit 8 performs an error weighting process for all data of the measurement voltage value v (t) in the low frequency region RN2, but the measurement voltage value v (( The error may be calculated after the data thinning process t), and the error weighting process may be performed.

The computing unit 8 calculates the open circuit voltage V _OCV using the parameters calculated using the estimation methods A to C in steps S12 to S14 (S15 in FIG. 3). Specifically, the calculation unit 8 acquires the voltage value v (t) applied to the equivalent circuit model 9 and substitutes the parameter into the equivalent circuit model 9 to calculate the voltage value v (t). The open circuit voltage V _OCV is calculated by subtracting from.

And the calculating part 8 estimates charge condition SOC according to the map of open circuit voltage _VOCV -charge condition SOC, and the information of temperature T (t) (S16). FIG. 9 shows an example of a map of the open circuit voltage V _{OCV -charged} state SOC stored in advance in the memory 8b in the computing unit 8, but the charged state SOC changes depending on the open circuit voltage V _OCV . The state of charge SOC can be calculated.

The features of the present embodiment will be summarized below using the above description as an example.
According to the present embodiment, the calculation unit 8 partially calculates the measurement voltage value v (t) for the measurement voltage value v (t) acquired after the change timing at which the measurement current value i (t) changes transiently. After the effective data D is determined by thinning out, the parameters of the equivalent circuit model 9 (resistance values of resistors R0... Rn, capacitors C1... Cn are set so as to minimize the sum of squares of errors by the non-linear least square method. Capacity value). In particular, the calculation unit 8 divides the high frequency region RN1 and the low frequency region RN2 set according to the time constants τ1. doing. As a result, the fitting function F1 as accurate as possible can be obtained, the parameters of the equivalent circuit model 9 can be estimated with high accuracy during actual traveling of the vehicle, and the state of charge SOC can be calculated with high accuracy.

  In addition, although the form which divided | segmented the frequency domain into RN1 and RN2 (and RN3) was shown, it is not limited to these two or three area | regions, For example, four or more frequency domain RN1, N2, RN3, ... The frequency regions RN1 and RN2 (and RN3...) May not be divided. Regardless of whether the area is divided or not, effective data is left in the low frequency area RN2 at a time interval (sampling interval) according to a function that regularly increases monotonically (eg, logarithmic function or square root function). D should be discarded and the remaining data discarded. When the area is divided, a different function may be used for each area, and the time interval is calculated according to the elapsed time from the timing when the measured current value i (t) changes transiently with the same function. You may do it. Even in such a case, the effective data D is determined while increasing the amount of thinning out the data D of the measured voltage value v (t) as time elapses from the timing when the measured current value i (t) changes transiently. Can do.

  The calculation unit 8 determines to decrease the effective data D from the high frequency region RN1 toward the low frequency region RN2 (and RN3...). Then, the influence on the fitting parameter by the effective data D on the high frequency region RN1 side can be increased, and the influence on the fitting parameter by the effective data D on the low frequency region RN2.

  For example, when the vehicle stops at a signal, the current value i (t) changes rapidly, and the calculation unit 8 obtains sampling data (voltage value v (t), current value i (t), etc.) for a first predetermined period. When the number of data more than the number (long time in terms of time (for example, several hundred seconds)) can be secured, the parameters of the fitting function F1 can be obtained with high accuracy, particularly on the low frequency region RN2 side, but the number of data in the high frequency region RN1 In some cases, the number of data in the low frequency region RN2 becomes too large.

  In such a case, the calculation unit 8 corrects the ratio of the number of data in the high frequency region RN1 and the low frequency region RN2 by thinning out the data in the low frequency region RN2. Thereby, the degree of influence on the fitting process by the effective data D in the low frequency region RN2 can be reduced, and the fitting accuracy of the high frequency region RN1 can be improved. Thereby, the parameter of the equivalent circuit model 9 can be calculated with high accuracy, and the state of charge SOC can be calculated with high accuracy.

  For example, when it is not possible to secure a long time (several hundreds of seconds) until the vehicle starts moving from a signal stop while actually traveling, when estimating (identifying) the parameters of the equivalent circuit model 9 during actual traveling, such a long time is required. May not be able to acquire the signal data. In such a case, for example, there is a possibility that the calculation unit 8 can acquire only the number of data less than the first predetermined number until the condition of step S7 is satisfied, and the parameter cannot be calculated with high accuracy.

  In such a case, the calculation unit 8 thins out and reduces the target data of the low frequency region RN2b that reaches the low frequency region RN2 from the high frequency region RN1 while the lowest frequency region RN2c (near the final data) of the data of the low frequency region RN2 ) Is left as a lot of effective data D so as not to be thinned out, and the effective data D in the lowest frequency region RN2c is made particularly dense. Then, when the calculation unit 8 performs the fitting process by the nonlinear least square method, the influence of the effective data D in the lowest frequency region RN2c out of the low frequency region RN2 can be increased, and the extremely low frequency component (DC resistance component) can be increased. ) Side error can be suppressed as much as possible. Thereby, the calculation accuracy of the parameters of the equivalent circuit model 9 can be improved, and consequently the state of charge SOC can be calculated with high accuracy.

  According to this embodiment, the calculation unit 8 sets an error weighting coefficient for the measured voltage value v (t) acquired after the change timing at which the measured current value i (t) changes transiently, and weights the measured voltage value v (t). Using the error weighted according to the coefficients, the parameters of the equivalent circuit model 9 (resistance values of resistors R0... Rn, capacitors C1... Cn are set so as to minimize the sum of squares of errors by the nonlinear least square method. Capacity value). In particular, the calculation unit 8 divides the high frequency region RN1 and the low frequency region RN2 set according to the time constants τ1... Τn of the RC parallel circuit 12a, and calculates error weighting coefficients in the regions RN1 and RN2 (and RN3). It is set. As a result, the fitting function F1 as accurate as possible can be obtained, the parameters of the equivalent circuit model 9 can be estimated with high accuracy during actual traveling of the vehicle, and the state of charge SOC can be calculated with high accuracy.

  In addition, although the form which sets the weighting coefficient of an error by dividing the frequency region into regions RN1 and RN2 (and RN3) is not limited to these two or three regions, for example, four or more frequencies The regions RN1, N2, RN3,... May be divided, and the frequency regions RN1 and RN2 (and RN3) may not be divided. Regardless of whether the region is divided or not, a function that decreases monotonically regularly (for example, a logarithmic function or the reciprocal of the square root function) has elapsed from the timing when the measured current value i (t) changes transiently. An error weighting coefficient may be set using a function that varies with time. Further, when the regions RN1, RN2 (, RN3,...) Are separated, different functions may be used for each region, and the measured current value i (t) changes transiently using the same function. An error weighting coefficient may be calculated according to the elapsed time from the timing.

  For example, when the above-described equation (3) is applied when the nonlinear square method is used, the error weighting coefficient k of the low frequency region RN2 may be set as a coefficient smaller than 1. Also in this case, a plurality of error weighting coefficients (for example, k1 and k2) may be used. The error weighting factor k2 in the low frequency region RN2 may be set smaller than the error weighting factor k1 in the high frequency region RN1.

  In particular, the calculation unit 8 determines that the number of data of the measured voltage value v (t) is less than the first predetermined number and that the number of data in the low frequency region RN2 is less than the second predetermined number compared to the high frequency region RN1. As a condition, weighting is performed so that at least part of the error in the low-frequency region RN2 is increased with respect to the error in the measured voltage value v (t) in the high-frequency region RN1. Specifically, for example, the calculation unit 8 uses the weighting coefficient k (> 1) to calculate the error of the data D in the low frequency region RN2 when calculating the sum of the squares of the errors to the fitting function of each data D. It is corrected.

  Then, the influence on the fitting parameter by the error (for example, e1 to e7) on the high frequency region RN1 side can be reduced, and the influence on the fitting parameter by the error (for example, e8 to e14) on the low frequency region RN2 side can be increased. As a result, the error on the DC resistance component (very low frequency component) side can be suppressed as much as possible, the parameter calculation accuracy can be improved, and the state of charge SOC can be calculated with high accuracy.

(Other embodiments)
The present invention is not limited to the above-described embodiment, and for example, the following modifications or expansions are possible.
The parameter calculation method of the above embodiment is, for example, the current value i flowing through the secondary battery 3 when the vehicle is stopped, when the vehicle is decelerated (rapid deceleration), when the engine is stopped, when the vehicle starts (rapid start), when the engine is started, etc. The parameters of the equivalent circuit model 9 can be calculated using the characteristics before and after the timing when (t) changes suddenly. When the current changes transiently (for example, stepped), if this transient change is repeated in a very short time, necessary data may not be acquired. In such a case, the calculation process of the state of charge SOC shown in FIG. 3 may not be performed. Further, for example, when applied to the process shown in FIG. 3, when the transient change is repeated in a short time, the calculation unit 8 determines NO in step S <b> 11, but in this case, the third predetermined number (<Second predetermined number) If more data is acquired, the process proceeds to step S14 to estimate a parameter using the estimation method C, and if only data less than the third predetermined number is acquired, the parameter is estimated. It is preferable that the process shown in FIG. Then, it is possible not to perform the calculation process of the state of charge SOC in step S16.

  In the above-described embodiment, the form of minimizing the total sum of the squares of errors by the nonlinear least square method is shown as the predetermined fitting method, but the present invention is not limited to this, and other types similar to the nonlinear least square method are shown. A fitting technique may be used. Nonlinear programming can be used as another fitting method.

  Although the RC parallel circuit group 12 in which the resistor R0 is connected in series is applied as the equivalent circuit model 9, it is not limited to this circuit configuration. For example, an equivalent circuit model to which other capacitor components or inductor components are added may be used.

  In the above-described embodiment, the measured voltage value v (t) after the measured current value i (t) changes transiently is divided into regions (RN1, RN2, RN3,...), And each region is processed individually. Although the form is shown, the processing may not be divided for each frequency domain (time domain).

  If the frequency domain is not divided, for example, in step S7 shown in FIG. 3, the calculation unit 8 replaces the process in step S12 with the condition that the number of target data is determined to be equal to or greater than the first predetermined number. The time interval is sampled by using a monotonically increasing function (for example, a logarithmic function or a square root function) as a function that changes according to the elapsed time from the timing at which the measured current value i (t) changes transiently. Effective data D may be determined as the interval.

  Further, a function (a simple decrease function in all frequency regions (all time regions) after the measured voltage value v (t) changes transiently regardless of whether or not the frequency regions RN1, RN2 (, RN3...) Are divided. For example, an error weighting coefficient corresponding to a logarithmic function or a reciprocal of the square root function) is linked to the measured voltage value v (t), or linked to the effective data D determined by thinning out the measured voltage value v (t). May be.

  In addition, whether the frequency regions RN1, RN2 (, RN3,...) Are divided or not, error weighting coefficients are calculated in all frequency regions (all time regions) after the measured voltage value v (t) changes transiently. Corresponding to the measured voltage value v (t), an equivalent circuit model is obtained by a predetermined fitting method (for example, the above-mentioned nonlinear least square method, nonlinear programming method, etc.) by partially thinning out an error weighted according to an error weighting coefficient. Nine parameters may be estimated.

  After the arithmetic unit 8 performs the weighting process on the errors of all the measured voltage values v (t) in the low frequency region RN2 (, RN3,...), The error that has been subjected to the weighting process is partially thinned out to obtain an effective error. And the parameters of the equivalent circuit model 9 may be estimated by a predetermined fitting method using this effective error. That is, for example, in the nonlinear least square method shown in the equation (3), for example, “e10” on the low frequency region RN2 side is an invalid error, and e1 to e7, e8 to e9, and e11 to e14 are effective errors. The sum may be obtained.

  In addition, the calculation unit 8 determines that the number of data of the measured voltage value v (t) is less than the first predetermined number, and determines that the number of data in the low frequency region RN2 is greater than or equal to the second predetermined number compared to the high frequency region RN1. On the condition, the parameter may be estimated by setting the error weighting coefficient for the data D by the following method instead of the parameter estimation method shown in step S13. For example, when the measured voltage value v (t) is acquired as shown in the upper diagram of FIG. 7, the calculation unit 8 does not thin out the data D, and the error occurs with respect to the data in the lowest frequency region RN2c in the low frequency region RN2. It is also possible to estimate the parameter of the equivalent circuit model (9) by setting the weighting coefficient of the above to be heavy. The data D may be thinned out after setting the error weighting coefficient to be heavy.

  In addition, the code | symbol with the parenthesis attached | subjected to the claim attaches | subjects the code | symbol corresponding to the component of this-application specification, and gives an example of the component. Therefore, the invention according to the present application is not limited to the elements indicated by the reference numerals attached to the constituent elements of the claims, and can be variously expanded in terms of the claims or the equivalents thereof.

  In the drawing, 1 is an equivalent circuit parameter estimation device, 4 is a voltage value measurement unit, 7 is a current value measurement unit, 8 is a calculation unit (estimation unit), 9 is an equivalent circuit model, 11 is an internal impedance element, and 12 is RC. A parallel circuit group, 12a indicates an RC parallel circuit.

Claims (16)

A current value measuring unit (7) for measuring a current flowing in a secondary battery (3) mounted on a vehicle such as a hybrid vehicle or an electric vehicle;
A voltage value measuring unit (4) for measuring a voltage applied to the secondary battery;
In the state where the plurality of RC parallel circuits (12a) having different time constants for the secondary battery are replaced with the equivalent circuit model (9), the measured current data and the voltage value measurement of the current value measuring unit (7) An estimation unit (8) for estimating a parameter of the equivalent circuit model (9) of the secondary battery (3) according to measured voltage data of the unit (4),
The estimation unit (8) determines effective data by partially thinning out the measurement voltage data acquired after the change timing at which the measurement current data by the current value measurement unit (7) changes transiently, A parameter estimation device for an equivalent circuit of a secondary battery for a vehicle, wherein the parameter of the equivalent circuit model (9) is estimated using the effective data by a predetermined fitting technique. The equivalent circuit parameter estimation apparatus for a vehicle secondary battery according to claim 1,
The said estimation part (8) determines the said effective data by using the time interval according to the function which increases monotonically regularly as a sampling interval, The parameter estimation apparatus of the equivalent circuit of the secondary battery for vehicles characterized by the above-mentioned. The parameter estimation device for an equivalent circuit of the vehicle secondary battery according to claim 1 or 2,
The estimation unit (8) sets a frequency region (RN1, RN2, RN3,...) Corresponding to a time constant of the RC parallel circuit (12a), changes the sampling interval for each setting region, and the effective data. A parameter estimation device for an equivalent circuit of a secondary battery for a vehicle, wherein In the vehicle secondary battery equivalent circuit parameter estimation apparatus according to claim 3,
The estimator (8) determines to reduce the effective data from the high frequency region (RN1) to the low frequency region (RN2, RN3,...) Of the frequency region. A parameter estimation device for an equivalent circuit of a secondary battery. In the parameter estimation apparatus of the equivalent circuit of the secondary battery for vehicles as described in any one of Claims 1-4,
A first judgment is made as to whether or not the number of data greater than or equal to the first predetermined number has been acquired from the timing at which the measured current data from the current value measuring section (7) has transiently changed for the measured voltage data of the voltage value measuring section (4). 1 determination means (8),
The estimation unit (8)
A frequency region (RN1, RN2, RN3,...) Corresponding to a time constant of the RC parallel circuit (12a) is set, and the effective data is determined by changing a sampling interval for each setting region,
On the condition that the first determination means (8) determines that the number of data equal to or greater than the first predetermined number has been acquired, the data of the low frequency region is thinned and the number of effective data in the low frequency region The parameter of the equivalent circuit of the vehicle secondary battery, wherein the ratio of the effective data in the high frequency region is corrected and the parameter of the equivalent circuit model (9) is estimated using the effective data after the processing Estimating device. In the parameter estimation apparatus of the equivalent circuit of the secondary battery for vehicles as described in any one of Claims 1-4,
A first judgment is made as to whether or not the number of data greater than or equal to the first predetermined number has been acquired from the timing at which the measured current data from the current value measuring section (7) has transiently changed for the measured voltage data of the voltage value measuring section (4). 1 determination means (8),
The estimation unit (8)
A frequency region (RN1, RN2, RN3,...) Corresponding to a time constant of the RC parallel circuit (12a) is set, and the effective data is determined by changing a sampling interval for each setting region,
On the condition that the first determination means determines that the number of data equal to or greater than the first predetermined number has not been acquired, the data in the low frequency region is thinned and the number of effective data in the low frequency region and the high frequency The ratio of the effective data in the region is corrected to make the data in the lowest frequency region (RN2c) dense, and the parameters of the equivalent circuit model (9) are estimated using the effective data after the processing. A parameter estimation device for an equivalent circuit of a vehicle secondary battery. The parameter estimation device for an equivalent circuit of the vehicle secondary battery according to any one of claims 1 to 6,
The estimation unit (8) partially thins out the measurement voltage data to determine effective data, sets an error weighting coefficient for the effective data, and uses an error weighted according to the weighting coefficient. An apparatus for estimating a parameter of an equivalent circuit of a secondary battery for a vehicle, wherein the parameter of the equivalent circuit model is estimated by a predetermined fitting method. A current value measuring unit (7) for measuring a current flowing in a secondary battery (3) mounted on a vehicle such as a hybrid vehicle or an electric vehicle;
A voltage value measuring unit (4) for measuring a voltage applied to the secondary battery;
In the state where the plurality of RC parallel circuits (12a) having different time constants for the secondary battery are replaced with the equivalent circuit model (9), the measured current data and the voltage value measurement of the current value measuring unit (7) An estimation unit (8) for estimating a parameter of the equivalent circuit model (9) of the secondary battery (3) according to measured voltage data of the unit (4),
The estimation unit (8) sets an error weighting factor for the measurement voltage data acquired after a change timing at which the current measurement data by the current value measurement unit (7) changes transiently, and sets the weighting factor to the weighting factor. A parameter estimation device for an equivalent circuit of a secondary battery for a vehicle, wherein the parameter of the equivalent circuit model (9) is estimated by a predetermined fitting method using an error weighted accordingly. The parameter estimation device for an equivalent circuit of the vehicle secondary battery according to claim 7 or 8,
The estimation unit (8) sets the weighting coefficient according to a function that regularly monotonously decreases when setting the weighting coefficient of the error, and estimates the parameter of the equivalent circuit of the vehicular secondary battery apparatus. The parameter estimation device for an equivalent circuit of the vehicle secondary battery according to any one of claims 7 to 9,
The estimation unit (8) sets frequency regions (RN1, RN2, RN3,...) Corresponding to the time constant of the RC parallel circuit (12a), and sets the error weighting coefficient for each setting region. An apparatus for estimating a parameter of an equivalent circuit of a secondary battery for a vehicle. In the vehicle secondary battery equivalent circuit parameter estimation apparatus according to claim 10,
The estimation unit (8) is characterized in that the error weighting coefficient is set to decrease from the high frequency region (RN1) to the low frequency region (RN2, RN3,...) Of the frequency region. A parameter estimation device for an equivalent circuit of a vehicular secondary battery. In the vehicle secondary battery equivalent circuit parameter estimation device according to claim 8,
The estimation unit (8) sets frequency regions (RN1, RN2, RN3,...) Corresponding to the time constant of the RC parallel circuit (12a), and changes and sets the error weighting coefficient for each setting region. ,
A first judgment is made as to whether or not the number of data greater than or equal to the first predetermined number has been acquired from the timing at which the measured current data from the current value measuring section (7) has transiently changed for the measured voltage data of the voltage value measuring section (4). 1 determination means,
The estimator (8) is configured so that the first determination means determines that the first predetermined number is larger than the first predetermined number, and the high frequency region (RN1) to the low frequency region (RN2, RN3,...) Of the frequency region. A parameter estimation device for an equivalent circuit of a secondary battery for a vehicle, characterized in that an error weighting coefficient is reduced as it goes to. In the vehicle secondary battery equivalent circuit parameter estimation device according to claim 8,
The estimation unit (8) sets frequency regions (RN1, RN2, RN3,...) Corresponding to the time constant of the RC parallel circuit (12a), and changes and sets the error weighting coefficient for each setting region. ,
First determination means for determining whether or not a number of data greater than or equal to a first predetermined number has been acquired from a timing at which measurement current data by the current value measurement unit has transiently changed with respect to measurement voltage data of the voltage value measurement unit;
It is determined whether or not the number of data in the low frequency region is greater than a second predetermined number compared to the high frequency region from the timing when the measured current data by the current value measuring unit changes transiently with respect to the measured voltage data of the voltage value measuring unit. Second determining means for performing,
The estimation unit (8) determines that the number of data is less than a first predetermined number by the first determination unit and the number of data in a low frequency region is less than a second predetermined number compared to a high frequency region by the second determination unit. The parameter of the equivalent circuit model (9) is estimated by weighting so that at least a part of the error in the low frequency region is made heavier with respect to the error in the high frequency region. An apparatus for estimating a parameter of an equivalent circuit of a vehicular secondary battery. In the vehicle secondary battery equivalent circuit parameter estimation device according to claim 8,
First determination means (8) for determining whether or not the number of data greater than or equal to a first predetermined number has been acquired from the timing at which the measured current data by the current value measuring section has changed transiently for the measured voltage data of the voltage value measuring section When,
It is determined whether or not the number of data in the low frequency region is greater than a second predetermined number compared to the high frequency region from the timing when the measured current data by the current value measuring unit changes transiently with respect to the measured voltage data of the voltage value measuring unit. Second determining means for performing,
The estimation unit (8)
A frequency region (RN1, RN2, RN3,...) Corresponding to a time constant of the RC parallel circuit (12a) is set, and the error weighting coefficient is changed and set for each setting region.
The first determination means determines that the number of data greater than or equal to the first predetermined number has not been acquired, and the second determination means determines that the number of data in the low frequency region is greater than or equal to the second predetermined number compared to the high frequency region. On the condition that the parameter of the equivalent circuit model (9) is estimated by increasing the weighting factor of the error in the measured voltage data in the lowest frequency region (RN2c) of the low frequency region (RN2). A parameter estimation device for an equivalent circuit of a vehicle secondary battery. In the parameter estimation apparatus of the equivalent circuit of the secondary battery for vehicles as described in any one of Claims 8-14,
The estimation unit (8) determines an effective error by partially thinning out an error weighted according to the weighting coefficient, and uses the effective error to determine a parameter of the equivalent circuit model (9) by a predetermined fitting technique. A parameter estimation device for an equivalent circuit of a vehicular secondary battery, characterized in that In the parameter estimation apparatus of the equivalent circuit of the secondary battery for vehicles as described in any one of Claims 1-15,
An apparatus for estimating a parameter of an equivalent circuit of a secondary battery for a vehicle, wherein the predetermined fitting technique is such that a total sum of squares of errors is set to a minimum value by a non-linear least square method.

Images