Battery information processing device and battery information processing method

Abstract

This invention provides a battery information processing device that can maintain degradation diagnostic quality while suppressing the amount of transmitted data, etc., by selecting or downsampling transmitted data, stored data, or processed data, taking into account the magnitude of quantization errors in the various data necessary for diagnosing degradation of battery packs. [Solution] The system includes a voltage change calculation unit that calculates the voltage change ΔV of the battery pack based on the input voltage time series, a voltage change estimation unit that estimates the voltage change ΔVest of the battery pack based on the input voltage time series, and a time series selection unit that selects and outputs the input voltage time series. The time series selection unit selects either the value before or after the change in the voltage time series, or the average value thereof, when the difference between the voltage change ΔV calculated by the voltage change calculation unit and the voltage change ΔVest estimated by the voltage change estimation unit is smaller than a predetermined value.

Description

Technical Field

[0001] The present invention relates to a battery information processing apparatus and a battery information processing method for suppressing the amount of data to be transmitted / stored / processed at a stage prior to deterioration diagnosis of a battery pack or the like.

Background Art

[0002] When the usage period of a secondary battery becomes long, battery deterioration such as a decrease in the Ah capacity of the battery or an increase in the resistance of the battery occurs. Therefore, in equipment or devices incorporating a secondary battery, it is necessary to replace the deteriorated battery at an appropriate time. Therefore, in an electric vehicle (hereinafter referred to as "EV") or a stationary battery system equipped with a battery pack, it is also necessary to replace the deteriorated battery pack at an appropriate time.

[0003] Here, in the technical field of EVs and stationary battery systems, techniques for monitoring the deterioration of a battery pack or diagnosing an abnormality in the battery pack based on various battery information have been put into practical use. The battery information transmitted and used in this case is generally voltage information of a battery cell or a battery pack, current information flowing through the battery pack, and temperature information of the battery.

[0004] These battery information are quantized by an AD converter to become digital data. However, if the quantized digital data is directly transmitted, stored, or processed, the communication load, storage capacity, and processing load increase. Therefore, in order to suppress the communication load and the like, it is effective to thin out the digital data so that the quality of deterioration monitoring and abnormality detection of the battery pack does not deteriorate.

[0005] Therefore, the abstract of Patent Document 1 states that the problem of "simplifying the system configuration to accurately acquire measurement data during long-term charge-discharge tests, and accurately acquiring detailed data of transient regions when switching between charge, discharge, and rest modes" is solved by "the control CPU 31 rapidly samples the detection signals of charge-discharge current and voltage with an AD converter and converts them into digital data. For these digital data, data is acquired for a predetermined fixed time in the vicinity of transient regions, while long-term digital data is acquired in parallel with this, and the same timing information is added to each of these digital data to associate them and transmit them to the computer 100 via Ethernet at high speed. On the computer side, the digital data are associated with each other based on the timing information added to the transmitted digital data and acquired."

[0006] Thus, in Patent Document 1, data is acquired only for a predetermined fixed time in the vicinity of the transient region, and in parallel with this, digital data over a long period of time is thinned out and acquired, thereby suppressing the amount of digital data transmitted. [Prior art documents] [Patent Documents]

[0007] [Patent Document 1] Japanese Patent Publication No. 2010-112771 [Overview of the Initiative] [Problems that the invention aims to solve]

[0008] However, the technology described in Patent Document 1 selects or downsamples digital data without considering the magnitude of the quantization error that inevitably occurs during quantization in the AD converter. Therefore, using digital data with large quantization errors may degrade the quality of charge / discharge tests.

[0009] Therefore, the present invention aims to provide a battery information processing device and a diagnostic battery information processing method that can maintain the quality of degradation diagnosis while suppressing the amount of transmitted data, stored data, or processed data by selecting or downsampling transmitted data, stored data, or processed data, taking into account the magnitude of the quantization error of various data necessary for diagnosing degradation of battery packs. [Means for solving the problem]

[0010] To solve the above problems, the battery information processing device of the present invention includes a voltage change calculation unit that calculates the voltage change ΔV of the battery pack based on an input voltage time series, a voltage change estimation unit that estimates the voltage change ΔVest of the battery pack based on an input voltage time series, and a time series selection unit that selects and outputs an input voltage time series. The time series selection unit selects either the value before or after the change in the voltage time series, or the average value thereof, when the difference between the voltage change ΔV calculated by the voltage change calculation unit and the voltage change ΔVest estimated by the voltage change estimation unit is smaller than a predetermined value.

[0011] Furthermore, the battery information processing device of the present invention includes a temperature change calculation unit that calculates the temperature change ΔT of the battery pack based on an input temperature time series, a voltage change estimation unit that estimates the temperature change ΔTest of the battery pack based on an input temperature time series, and a time series selection unit that selects and outputs an input temperature time series. The time series selection unit selects either the value before or after the change in the temperature time series, or the average value thereof, when the difference between the temperature change ΔT calculated by the temperature change calculation unit and the temperature change ΔTest estimated by the temperature change estimation unit is smaller than a predetermined value. [Effects of the Invention]

[0012] According to the battery information processing device and the diagnostic battery information processing method of the present invention, by selecting or thinning out transmission data, stored data, or processed data in consideration of the magnitude of quantization errors of various data necessary for diagnosing the deterioration of a battery pack, it is possible to maintain the deterioration diagnosis quality while suppressing the amount of transmission data, stored data, or processed data.

[0013] Problems, configurations, and effects other than those described above will be clarified by the description of the following embodiments.

Brief Description of the Drawings

[0014] [Figure 1] Functional block diagram of an EV system to which the battery information processing device of Example 1 is applied. [Figure 2] Functional block diagram of the battery information processing device of Example 1. [Figure 3A] Example of quantization when the quantization specification is truncation. [Figure 3B] Example of quantization when the quantization specification is rounding up. [Figure 3C] Example of quantization when the quantization specification is rounding. [Figure 4] Logic table used by the selection determination unit of Example 1. [Figure 5] Example of an equivalent circuit model of a battery. [Figure 6] Example of a battery model table. [Figure 7] Example of an OCV table.

Modes for Carrying Out the Invention

[0015] Hereinafter, embodiments of the battery information processing device and the battery information processing method according to the present invention will be described with reference to the drawings.

Examples

[0016] Referring to FIGS. 1 to 7, the battery information processing device 1 according to Example 1 of the present invention will be described.

[0017] <Battery Monitoring System 2> First, the battery monitoring system 2 to which the battery information processing device 1 of this embodiment is applied will be explained using the functional block diagram of Figure 1. As shown in the figure, the battery monitoring system 2 is a system in which the EV 21 and the server 22 are connected in a communicative manner. Note that either wireless communication or wired communication may be used for communication between the two. Using a wireless communication network such as a mobile phone network allows communication between the two even while the EV 21 is in motion. On the other hand, using a wired communication network allows communication between the two to be established at specific locations, such as near an EV charger.

[0018] < <ev21>> EV21 includes a battery pack 21a, a BMS (Battery Management System) 21b, and a transmission unit 21c. Each will be briefly described below.

[0019] The battery pack 21a is an energy source that supplies power to the drive motor of EV21 and the like, and incorporates a plurality of battery cells.

[0020] The BMS 21b is a system that outputs digital data used for deterioration monitoring and abnormality diagnosis of the battery pack 21a. Specifically, it is a system that generates and outputs time-series digital data by quantizing analog information such as the voltage of the battery pack 21a, the voltage of each battery cell, the current flowing through the battery pack 21a, and the temperature of each battery cell at a predetermined interval (for example, 100 ms) using an AD converter. Hereinafter, the time-series digital data of voltage, current, and temperature will be referred to as voltage time series, current time series, and temperature time series, respectively. Needless to say, in order to obtain the above various analog information, voltage sensors, current sensors, temperature sensors, etc. are arranged at appropriate parts of EV21 or the battery pack 21a.

[0021] This BMS 21b is an example of the application target of the battery information processing device 1. The BMS 21b to which the battery information processing device 1 is applied can suppress the amount of data transmitted to the transmission unit 21c by thinning out a part of the time-series digital data or reducing the number of Bits of the digital data.

[0022] The transmission unit 21c is a functional unit that transmits the digital data output by the BMS 21b to the server 22. This transmission unit 21c is an example of the application target of the battery information processing device 1. The transmission unit 21c to which the battery information processing device 1 is applied can suppress the amount of data transmitted to the server 22 by thinning out a part of the time-series digital data or reducing the number of Bits of the digital data.

[0023] <<Server 22>> Server 22 comprises a receiving unit 22a, storage 22b, and battery diagnostic unit 22c. Each of these is outlined below.

[0024] The receiving unit 22a is a functional unit that transmits the time-series digital data received from the transmitting unit 21c to the storage unit 22b. This receiving unit 22a is one example of an application of the battery information processing device 1. A receiving unit 22a to which the battery information processing device 1 is applied can suppress the amount of data transmitted to the storage unit 22b (i.e., the amount of data stored in the storage unit 22b) by decimating some of the time-series digital data or reducing the number of bits in the digital data.

[0025] The storage unit 22b is a functional unit that stores time-series digital data received from the receiving unit 22a, and specifically includes a hard disk drive, semiconductor memory, and the like.

[0026] The battery diagnostic unit 22c is a functional unit that performs degradation monitoring, anomaly diagnosis, and failure prediction of the battery pack 21a by analyzing time-series digital data stored in the storage 22b. Specifically, it is a functional unit realized by the execution of a predetermined program by the computing device such as the CPU or SoC of the server 22.

[0027] This battery diagnostic unit 22c is one example of an application for the battery information processing device 1. The battery diagnostic unit 22c to which the battery information processing device 1 is applied can reduce the analysis load of digital data by decimating some of the time-series digital data or reducing the number of bits in the digital data before analyzing the digital data. The diagnostic results output by the analysis of this battery diagnostic unit 22c include, for example, the battery's SOHQ (State of Health Capacity), SOHR (State of Health Resistance), Wh-SOH, etc. In addition, examples of battery failures that the battery diagnostic unit 22c can predict include failures such as a sudden drop in voltage due to a micro-short, a sudden decrease in capacity, and a sudden increase in resistance (sudden voltage fluctuation when constant current is applied).

[0028] In Figure 1, the battery monitoring system 2 is shown to monitor the battery pack 21a built into the EV21. However, the battery monitoring system 2 is not limited to this, and may also monitor battery packs built into HEVs (Hybrid Electric Vehicles) or stationary battery systems.

[0029] <Battery Information Processing Device 1> Figure 2 is a functional block diagram of the battery information processing device 1, which is implemented using software. As shown in the figure, the battery information processing device 1 includes a voltage change calculation unit 11, a quantization specification recording unit 12, a voltage change estimation unit 13, a selection determination unit 14, and a time series selection unit 15, and outputs digital data for subsequent transmission / storage / processing. The details of each unit will be explained sequentially below. Although Figure 2 illustrates a situation in which a voltage time series is input and output, the battery information processing device 1 may also be configured to input and output a temperature time series instead of a voltage time series. In that case, it goes without saying that the battery information processing device 1 can be configured by replacing the voltage change calculation unit 11 and the voltage change estimation unit 13 in Figure 2 with a temperature change calculation unit and a temperature change estimation unit.

[0030] <<Voltage Change Calculation Unit 11>> The voltage change calculation unit 11 calculates the change amount ΔV of the voltage time series input to the battery information processing device 1. That is, if the input voltage one time step ago was 4.12V and the input voltage at the current time is 4.14V, it outputs 0.02V as the change amount ΔV.

[0031] <<Quantization Specification Recording Unit 12>> The quantization specification recording unit 12 records the specification of the quantization step size δ for the quantization process, and the specification of whether the quantization process is truncation, rounding, or rounding up.

[0032] <<Voltage Change Estimation Unit 13>> The voltage change estimation unit 13 estimates how many mV the battery voltage will change based on past voltage time series, current time series, and temperature time series. This can be estimated from the time series using, for example, an ARMA model (autoregressive moving average model), or it can be estimated from a battery model if one exists. This value is denoted as ΔVest. Details of this estimation method will be described later.

[0033] <<Selection Determination Unit 14>> The selection determination unit 14 determines how to select the voltage time series input to the battery information processing device 1 and pass it on to the subsequent transmission / storage / processing stage. The determination here changes depending on whether the quantization specification is truncation M1, rounding M2, or rounding up M3. The premise of this determination is explained using Figures 3A to 3C. In each figure, the horizontal axis represents the sampling time and the vertical axis represents the value. Furthermore, although the characteristics described below are those when quantizing the true value D0, which is analog data, similar characteristics exist when requantizing digital data using a larger quantization step δ to reduce the number of bits in the digital data.

[0034] Figure 3A shows an example of quantization when the quantization specification is truncation M1. As a result of truncating the true value D0 of the trigonometric function by quantization processing in units of 0.01V, digital data D1 is always generated that is smaller than the true value D0. In this example, when the true value D0 is increasing (ΔV>0), the error with the true value D0 is minimized when the digital data D1 increases, and when the true value D0 is decreasing (ΔV<0), the error with the true value D0 is maximized when the digital data D1 decreases.

[0035] Figure 3B shows an example of quantization when the quantization specification is rounding M2. The true value D0 of the trigonometric function is rounded using a quantization process in units of 0.01V, resulting in the generation of digital data D2 that is slightly above or below the true value D0. In this example, the error with the true value D0 is minimized by using the average value of the digital data D2 before and after the change.

[0036] Figure 3C shows an example of quantization when the quantization specification is rounded up to M3. As a result of rounding up the true value D0 of the trigonometric function by 0.01V units in the quantization process, digital data D3 is always generated that is larger than the true value D0. In this example, when the true value D0 is increasing (ΔV>0), the error with the true value D0 is maximized when the digital data D3 is increasing, and when the true value D0 is decreasing (ΔV<0), the error with the true value D0 is minimized when the digital data D3 is decreasing.

[0037] From the above explanation, it can be seen that the time at which the error between the true value D0 and the quantized digital data is minimized differs depending on the quantization specification.

[0038] Next, the decision logic in the selection decision unit 14 will be explained using the logic table in Figure 4. In Figure 4, data rows No. 1 to 5 are prepared for truncation M1, data rows No. 6 to 10 are prepared for rounding M2, and data rows No. 11 to 15 are prepared for rounding up M3.

[0039] Data column C1 is a data column used to select data rows according to the quantization specifications.

[0040] Data sequence C2 is a data sequence for extracting data rows corresponding to the relationship between the voltage change ΔV calculated by the voltage change calculation unit 11 and the quantization step width δ recorded in the quantization specification recording unit 12.

[0041] Data column C3 is a data column for extracting data rows that correspond to the relationship between the magnitude of the |ΔV-ΔVest| value, calculated using the voltage change ΔV calculated by the voltage change calculation unit 11 and the voltage change ΔVest estimated by the voltage change estimation unit 13, and the judgment threshold ε.

[0042] These data sequences C1 to C3 allow us to identify the data rows corresponding to the quantization specification and the relationships between ΔV, δ, ΔVest, and ε.

[0043] Additionally, data column C4 is the reliability data column, data column C5 is the selected data data column, and data column C6 is the data time data column.

[0044] For example, if the quantization specification is truncation M1 and the voltage change ΔV is 0, data row No. 3 is selected. In this situation, since there is no change in the voltage time series, outputting the current voltage value overlaid on the output of the previous voltage value does not provide any particular informational value. Therefore, under the conditions of data row No. 3, the selection determination unit 14 outputs a logic that does not select the voltage time series sampled this time (see data column C5). As a result, the battery information processing device 1 can decimate a portion of the input voltage time series and suppress the amount of data in the output voltage time series. Note that increasing the quantization step size δ increases the amount of voltage time series to be decimated, regardless of whether the quantization target is analog or digital data. Therefore, it is desirable to increase the quantization step size δ to an extent that does not cause degradation during the final diagnostic processing.

[0045] Furthermore, if the quantization specification is truncation M1, the voltage change ΔV is greater than or equal to δ, and |ΔV-ΔVest|<ε, then data row No. 5 is selected. In the situation where this data row is selected, if the quantized data has changed and the voltage change is minimal, the value after the voltage change can be judged to be reliable because it has changed due to the truncation of the quantized data (see data column C4). For this reason, the data after the voltage change and the time after the change are adopted (see data columns C5 and C6). This makes it possible to avoid outputting data before the voltage change, which has a large quantization error.

[0046] Furthermore, if the quantization specification is truncation M1, the voltage change ΔV is greater than or equal to δ, and |ΔV-ΔVest|≧ε, then data row No. 4 is selected. In this situation, the voltage change estimate is deemed unreliable because it has become an unexpected value (see data column C4). In this case, the data may not be selected, or even if selected, it may be flagged as unreliable and transmitted / stored (see data columns C5 and C6).

[0047] On the other hand, if the voltage change ΔV is less than or equal to -δ, the voltage after the change is unreliable, but the voltage before the change is reliable because it is the value just before truncation. Therefore, if the quantization specification is truncation M1, the voltage change ΔV is less than or equal to -δ, and furthermore |ΔV-ΔVest|<ε, then data row No.2 is selected. In other words, since the value hardly changes, the data before the change is adopted, and the data time is set to before the data change (see data columns C5 and C6). This makes it possible to avoid outputting data after a voltage change, which has a large quantization error.

[0048] The above explanation applies when the quantization specification is truncation M1. However, in the case of rounding up M3, the selection logic is reversed due to the mechanism explained in Figures 3A and 3C. Furthermore, in the case of rounding to the nearest integer M2, due to the mechanism explained in Figure 3B, the true value is crossed before and after the value change, so the average of the values ​​before and after and the average of the time points before and after are adopted.

[0049] Here, since the voltage change ΔV has quantization step variations, the ΔV used in ΔV-ΔVest may be filtered with a low-pass filter. The judgment threshold ε may be δ / 2, δ, or a small constant. Also, the processing of data column C3 to identify data rows based on |ΔV-ΔVest| may be omitted.

[0050] <<Time Series Selection Section 15>> The time series selection unit 15 selects data to be output to the next stage from the input voltage time series according to the logic table in Figure 4. As a result, if the battery information processing device 1 of this embodiment is applied to the BMS 21b or the transmission unit 21c, the amount of data can be reduced while maintaining the quality of the data transmitted to the next stage; if it is applied to the reception unit 22a, the amount of data can be reduced while maintaining the quality of the data stored in the storage 22b; and if it is applied to the battery diagnostic unit 22c, the amount of data can be reduced while maintaining the quality of the data being processed.

[0051] <<Details of the voltage change estimation unit 13>> Next, we will explain the details of the voltage change estimation unit 13. Here, we will explain the cases where the voltage change estimation unit 13 has a battery model table and where it does not.

[0052] <<<If you have a battery model>>> If a battery model is available, the equivalent circuit model of the battery can be represented as shown in Figure 5. In this case, the battery voltage V(t) at time t can be calculated using Equation 1, where SOC(t) represents the battery's charge level [%].

[0053]

number

[0054] Vpk(t) is the k-th polarization voltage Vp and is approximated by Equation 2. This is because the polarization voltage Vpk(t) in Figure 5 is due to the parallel CR circuit. Δt is the sampling time step.

[0055]

number

[0056] Here, the difference in battery voltage from the previous time, i.e., the voltage change ΔV(t) = V(t) - V(t-1), is calculated using Equation 3. However, Equation 3 is an approximation formula assuming that the change in SOC at time 1 is negligible. ΔI(t) in Equation 3 represents I(t) - I(t-1). τk becomes ck × rk, and its dimension is [s], representing the time constant.

[0057]

number

[0058] From Equation 3, ΔV(t) can be estimated by calculating the values ​​of rk, τk, and r0 from the battery model table (index is SOC) and the estimated values ​​of SOC and each polarization voltage vpk for reference to the battery model table using Equation 2 every hour. A hypothetical example of this battery model table is shown in Figure 6. The battery model table may also include an index for temperature, or an index for current.

[0059] The SOC(t) can be calculated using the value output from BMS21b, or it can be estimated using a Kalman filter. If estimating with a Kalman filter, a battery model table is required, but since it contains the values ​​for τk, rk, and r0, this table can be used.

[0060] <<<If you do not have a battery model>>> Next, we will discuss voltage estimation when a battery model table is not available. In this case, the voltage at the next time point can be estimated from past current time series and voltage time series using AR (autoregression), ARMA models, or recurrent neural networks.

[0061] This is because the equivalent circuit model of the battery has the configuration shown in Figure 5, meaning that V(t) can be expressed by the difference equation in Equation 4. The coefficients and constants in Equation 4 can be expressed using the circuit parameters, OCV, in Figure 5, but this is omitted here. Furthermore, the coefficients and constants in Equation 4 can be considered constant values ​​for short periods of time (for example, a time step of about 100) during which the State of Cycles (SOC) can be assumed to remain unchanged.

[0062]

number

[0063] Equation 4 is an AR model, and the coefficients and constants can be determined by using the values ​​of the quantized voltage time series and current time series for V and I. These coefficients and constants can be determined by creating the Yule-Walker equation using past V and I time series and solving the system of linear equations. The length of these coefficients can be, for example, the past 100 steps, which is the range in which the SOC does not change much, or the time range of the current SOC and α%. The timing can be every step, or for example, every 10 steps. Alternatively, instead of Equation 4, a difference equation with the time difference of current and voltage as the variable can be used as Equation 5. Furthermore, the coefficients and constants in Equation 4 can be calculated and memorized in advance (however, this cannot account for degradation). The number of polarizations can be determined by the Akaike Information Criterion (hereinafter referred to as "AIC"), or it can be a fixed value beforehand.

[0064]

number

[0065] Furthermore, equations 4 and 5 may be identified using an autoregressive moving average model (hereinafter referred to as the "ARMA model") with delayed noise added, or they may be treated nonlinearly as a neural network.

[0066] <<Details of Battery Diagnostic Unit 22c>> Next, the diagnostic process in the battery diagnostic unit 22c will be explained. This can be defined solely by the timing of voltage changes, or the data can be reproduced at each sampling time. In the latter case, the acquired data can be linearly interpolated or spline interpolated. The diagnostics here are used to calculate the SOC inversely when the OCV settles to a constant value, to calculate resistance, and to calculate capacitance. Since the resistance and capacitance calculations are diagnostic values, the user is notified at this time. An example of resistance and capacitance calculation will be explained.

[0067] The resistance is used to calculate SOHR (the ratio of the DC resistance r0 to its value when new, converted to 25°C). This is done by calculating ΔV / ΔI and using the 25°C converted resistance R25 for the SOC(t) at that time. An example of the 25°C conversion formula is given in Equation 4. B may be a pre-set value or may be determined from the physical properties. Equation 6 uses Arrhenius's law. In Equation 6, temp refers to temperature [°C]. Alternatively, a conversion table for resistance by temperature and SOC may be used.

[0068]

number

[0069] Here, ΔV is calculated based on the quantized voltage and is therefore affected by quantization errors. Also, from Equation 3, the polarization voltage component is affected by the reciprocal of ΔI / Δt. That is, when ΔI is small, the error in SOHR becomes larger due to the influence of polarization. In this case, the error is smaller if the current change is large and the processing is performed at a time when the voltage is reliable. Therefore, the resistance is calculated as ΔV / ΔI when |ΔI| is above a preset threshold and the reliability data sequence C4 in Figure 4 is "present". Then, the resistance converted to 25°C is obtained using Equation 6. When calculating as SOHR, a resistance table R25_table(soc) for new products at 25°C is prepared and calculated using Equation 7.

[0070]

number

[0071] Next, determining the OCV after a sufficient amount of time has elapsed is important for estimating the battery's state of charge (SOC). The method for doing this is described below. As time progresses, the OCV changes because the battery's polarization voltage approaches zero, but it will eventually settle at a constant value.

[0072] Therefore, there are two methods: "A. Measuring OCV when a sufficient amount of time has elapsed" and "B. Measuring the OCV time series and estimating the value when a sufficient amount of time has elapsed." We will explain each of these methods.

[0073] First, in the case of "A. OCV measurement timing after sufficient time has elapsed," this is defined as the "Yes" timing in the reliability column of Figure 3. Then, the quantized value is processed as shown in Figure 3 and the voltage is read. Using this voltage value Vs, the SOC is calculated by working backward from the OCV(SOC) table and the diagnosis is performed.

[0074] Next, we will explain "B. Estimated value obtained by measuring the OCV time series after a sufficient amount of time has elapsed." For B, it is necessary to collect the voltage time series. This collection timing is specified as "Yes" in the reliability column of Figure 3. In this case, the voltage time series will no longer be arranged at equal intervals. We will explain how to handle this case. Generally, a battery circuit consists of multiple CR parallel circuits (polarization) connected in series. Therefore, the polarization time series is expressed as the sum of multiple exponential functions. This is written in Equation 8. Equation 8 is the equation for the voltage time series after the current becomes 0.

[0075]

number

[0076] We need to find OCV(SOC) (which is a constant here) and vp1(0),...,vpk(0),τ1,...,τk that minimize the sum of squares (Equation 9) of the difference between Equation 8 and the reliable voltage data V(t(n)) at time t(n) (the voltage processed as "present" in Figure 4 using the reliability data sequence C4 in Figure 4). Here, we need a number of data points such that n > 2k+1.

[0077]

number

[0078] To find the OCV(SOC), vp1(0), ..., vpk(0), τ1, ..., τk that minimizes Equation 9, one can use Newton's method to make the gradient of Equation 7 zero, or one can use a differential evolution algorithm or the steepest descent method. The obtained OCV(SOC) will be the final OCV, and based on this value, SOC can be found from the OCV(SOC) table (an example of an OCV table is shown in Figure 7). Alternatively, instead of Equation 6, the coefficients and constants can be found using the AR model or ARMA model, and the values ​​when the voltage becomes steady can be adopted. In this case, the model in Equation 4 is replaced by setting the current = 0.

[0079] Here, the number of polarizations can be determined by AIC, or it can be a predetermined fixed number (for example, 3).

[0080] In the case of an EV, if we have the voltage time series before and after charging, and the current during charging, it is possible to determine the battery's SOHQ. This can be done using Equation 10.

[0081]

number

[0082] Q: Charged electricity [Ah] Qmax: Battery rated capacity [Ah] Next, there are cases where sufficient voltage time series data before and after charging cannot be obtained, and it becomes necessary to determine SOHQ and SOHR from the voltage time series during charging. The process in this case is described below. This is a method for determining SOHQ and SOHR from past current time series and voltage time series. Here, we assume that the shape of the resistance (a function of SOC) does not change due to degradation, or that it can be ignored, and that the resistance after degradation is the resistance when new multiplied by SOHR / 100. From this, for example, using the current time series and voltage time series during charging or driving, we obtain Equation 11, and the Lp norm of the difference between the estimated time series and the actual time series is used as the evaluation value, and we find the SOHR and SOHQ that minimize Equation 11. p can be 1, 2, or ∞ (maximum value selection). N is the number of data points.

[0083]

number

[0084] Here, the estimated voltage time series (t; SOHQ, SOHR) can also be expressed as Equation 12.

[0085]

number

[0086] Here, if an estimate of SOC(t;SOHQ) is available using the Kalman filter, that value should be used, and OCV should be estimated from the OCV table. If there is no estimated value for SOC, the SOC is given in the form of equation 13.

[0087]

number

[0088] In Equation 13, Q(t) is the current integral time series [Ah], and Qmax represents the current battery capacity [Ah]. The values ​​of vp1, vp2, ... in Equation 12 may also be expressed as Equation 2 using a table of polarization resistance and polarization time constant (Figure 6). Alternatively, the data excluding transient data (for example, data for a certain period when the current changes and becomes constant, e.g., 500s) may be used, and the calculation vpk(SOC) = rk(SOC) × I may be performed. In this case, the data for a certain period when the current changes and becomes constant, e.g., 500s, is excluded from Equation 11. The target of the norm calculation in Equation 11 (essentially, which time t value is used in the calculation) may be defined as the timing at which the data shown in Figures 3A to 3C is considered correct, or, as mentioned above, the correct data may be interpolated and used.

[0089] The timing of the diagnosis can be when the data that appears correct in Figures 3A-3C is received, or when charging or driving is completed. Defining the timing of the data in this way improves accuracy by excluding data with large errors and speeds up the optimization calculation.

[0090] Next, instead of using Equation 11, you can convert the resistance based on the voltage time series and current time series and compare it with the resistance table in Equation 14, or with the SOHR estimate in Equation 15. The estimated SOHR is calculated as Estimated Resistance Time Series (t;SOHQ) ÷ Table Resistance (t;SOHQ) × 100 in Equation 15.

[0091]

number

[0092]

number

[0093] The estimated resistance time series can also be expressed as Equation 14, which is the difference between OCV and actual voltage divided by the current. Here, if the absolute value of the current is smaller than a predetermined value, the corresponding data is excluded from the calculations in Equations 14 and 15.

[0094]

number

[0095] SOC(t;SOHQ) can also be expressed as Equation 13. For optimization to minimize the Lp norm, the initial SOC, SOHR, and SOHQ can be determined using a quasi-Newton method or a differential evolution algorithm. The target of the norm calculation in Equations 14 and 15 (essentially, which time t value to use in the calculation) can be defined as the timing at which the data shown in Figure 2 is thought to be correct, or, as mentioned above, the data that is thought to be correct can be interpolated and used.

[0096] The timing of the diagnosis can be when the data that appears correct in Figure 2 is received, or when charging or driving is completed. Defining the timing of the data in this way improves accuracy by excluding data with large errors and speeds up the optimization calculation.

[0097] In this case, for the resistor, temperature information can be used as a temperature compensation, and the correction in Equation 6 may also be applied.

[0098] As described above, the battery information processing device of this embodiment can maintain degradation diagnosis quality while suppressing the amount of transmitted data, stored data, or processed data by selecting or downsampling transmitted data, stored data, or processed data, taking into account the magnitude of the quantization error of the various data necessary for battery pack degradation diagnosis. [Examples]

[0099] Next, the battery information processing device 1 of Example 2 will be described. Note that common points with Example 1 will be omitted from the explanation.

[0100] This embodiment describes the data selection when temperature data is quantized. This method is similar to the voltage data processing described in Example 1, but the battery temperature estimation method changes. This method is described below.

[0101] The temperature can be determined using the diffusion equation, but this is impractical from a computational standpoint because it is a partial differential equation. For this reason, as in this embodiment, a thermal model can be used, with the typical battery temperature θ(t)(°C) expressed as equations 17 and 18. Here, θa(t) represents the ambient temperature (°C). P(t)(W) is the heat source and is expressed by the equation |(battery voltage(t)-OCV(SOC(t)))×current(t)|. C is the heat capacity (J / K) and Y is the thermal conductance (W / K) constant.

[0102]

number

[0103]

number

[0104] Equations 17 and 18 allow us to numerically determine the battery temperature θ(t) if θa(t), C, and Y are given. Therefore, if the ambient temperature is measured separately and C and Y are set in advance, the temperature can be estimated. In reality, C and Y are often unknown. In this case, since equations 17 and 18 are linear differential equations, X can be approximated by the difference equation in equation 19.

[0105]

number

[0106] Since Equation 19 is an AR model, the coefficients a1 and b0 can be determined from the past time series of X and P. Next, what we want to find is the temperature difference ΔX(t) = X(t) - X(t-1). In this case, we get Equation 20.

[0107]

number

[0108] By using Equation 20, information on the outside temperature becomes unnecessary. Here, we assume that the outside temperature does not change suddenly. Alternatively, using Equation 20, we can estimate the next temperature difference by finding the coefficient from the past temperature difference and the time series of P.

[0109] In practice, the thermal circuit model may consider multiple temperature points, and C and Y in Equation 18 may be represented as matrices. Therefore, we can consider the difference equation in Equation 21 as the actual solution.

[0110]

number

[0111] As an AR model for Equation 21, the coefficients of Equation 21 can be identified and the temperature difference predicted. Also, because it is a battery pack, it is often possible to obtain temperature information at multiple points within the pack. In this case, the cells within the battery pack are close to each other and influence one another. Therefore, in Equation 21, ΔX can be treated as the temperature vector of multiple measured points, and a1, a2, ... can also be treated as vectors. The above is an AR model, but it can also be solved as an ARMA model assuming that the noise has autocorrelation, or as a recurrent neural network assuming nonlinearity. [Examples]

[0112] Next, the battery information processing device 1 of Example 3 will be described. Note that common points with Example 1 will be omitted from the explanation.

[0113] If the quantization specification is truncation M1, instead of the logic in Figure 3, you may uniformly add the quantization step size δ / 2. Also, if the quantization specification is rounding up M3, instead of the logic in Figure 3, you may uniformly subtract the quantization step size δ / 2. After subtracting or adding the quantization step size δ / 2, you may use the data as is, or you may perform the same processing as when rounding. [Examples]

[0114] Next, the battery information processing device 1 of Example 4 will be described. Note that common points with Example 1 will be omitted from the explanation.

[0115] In this embodiment, the reliability of the estimate is also taken into consideration. In short, a message is displayed as a way to handle cases where the estimate is unreliable. The user can choose between two modes: one where the calculation includes unreliable data, and another where only reliable data is extracted and used for the calculation. When displaying information to the user, it is also possible to notify them that the calculation was performed using reliable data (with limited data) or unreliable data with a large amount of data. [Explanation of Symbols]

[0116] 1. Battery Information Processing Device 11 Voltage change calculation unit, 12 Quantization specification recording unit, 13 Voltage change estimation unit, 14. Judgment Selection Unit, 15. Time series selection section, 16 Information Processing Department 2. Battery monitoring system, 21 EVs, 21a battery pack, 21b BMS, 21c Transmitter, 22 servers, 22a Receiver, 22b storage, 22c Battery Diagnostic Unit

Claims

1. A battery information processing device applied to a battery monitoring system that monitors battery packs, A voltage change calculation unit that calculates the voltage change ΔV of the battery pack based on the input voltage time series, A voltage change estimation unit that estimates the voltage change ΔVest of the battery pack based on the input voltage time series, It has a time series selection unit that selects and outputs an input voltage time series, The battery information processing device is characterized in that, when the difference between the voltage change ΔV and the voltage change ΔVest is smaller than a predetermined value, it selects either the value before or after the change in the voltage time series, or the average value thereof.

2. In the battery information processing device according to claim 1, The battery information processing device is characterized in that the time series selection unit does not output the input voltage time series when the voltage change ΔV is 0.

3. In the battery information processing device according to claim 1, The battery information processing device is characterized in that the voltage change estimation unit estimates the voltage change ΔVest from past current time series and current by assuming a voltage circuit model of the battery pack and preparing voltage circuit parameters.

4. In the battery information processing device according to claim 1, The battery information processing device is characterized in that the voltage change estimation unit considers the voltage circuit model of the battery pack as a difference equation, identifies the parameters of the difference equation from past current time series and voltage time series, and estimates the voltage change ΔVest from those parameters and the difference equation.

5. In the battery information processing device according to claim 1, The aforementioned time series selection unit, when the quantization specification is truncation, If the voltage change ΔV is negative, select the voltage value and time before the change in the voltage time series. A battery information processing device characterized by selecting the voltage value and time after the change in the voltage time series if the voltage change ΔV is positive.

6. In the battery information processing device according to claim 1, The aforementioned time series selection unit, when the quantization specification is rounding, A battery information processing device characterized by selecting the average value of the voltage value and time before and after the change in the voltage time series if the voltage change ΔV is negative or positive.

7. In the battery information processing device according to claim 1, The aforementioned time series selection unit, when the quantization specification is rounded up, If the voltage change ΔV is negative, select the voltage value and time after the change in the voltage time series. A battery information processing device characterized by selecting the voltage value and time before the change in the voltage time series if the voltage change ΔV is positive.

8. A battery information processing device applied to a battery monitoring system that monitors battery packs, A temperature change calculation unit calculates the temperature change Δθ of the battery pack based on the input temperature time series, A voltage change estimation unit that estimates the temperature change Δθest of the battery pack based on the input temperature time series, It has a time series selection unit that selects and outputs an input temperature time series, The battery information processing device is characterized in that, when the difference between the temperature change Δθ and the temperature change Δθest is smaller than a predetermined value, it selects either the value before or after the temperature time series changes, or the average value thereof.

9. In the battery information processing device according to claim 8, The battery information processing device is characterized in that the time series selection unit does not output the input temperature time series when the temperature change Δθ is 0.

10. In the battery information processing device according to claim 8, The battery information processing device is characterized in that the temperature change estimation unit estimates the temperature change Δθest from past temperature time series and heat generation amount by assuming a thermal circuit model of the battery pack and preparing thermal circuit parameters.

11. In the battery information processing device according to claim 8, The battery information processing device is characterized in that the temperature change estimation unit considers the thermal circuit model of the battery pack as a difference equation, identifies the parameters of the difference equation from past temperature heat generation time series, and estimates the temperature change Δθest from those parameters and the difference equation.

12. In the battery information processing device according to claim 8, The aforementioned time series selection unit, when the quantization specification is truncation, If the temperature change Δθ is negative, select the temperature value and time before the change in the temperature time series. A battery information processing device characterized by selecting the temperature value and time after the change in the temperature time series if the temperature change Δθ is positive.

13. In the battery information processing device according to claim 8, The aforementioned time series selection unit, when the quantization specification is rounding, A battery information processing device characterized by selecting the average value of the temperature value and time before and after the change in the temperature time series if the temperature change Δθ is negative or positive.

14. In the battery information processing device according to claim 8, The aforementioned time series selection unit, when the quantization specification is rounded up, If the temperature change Δθ is negative, select the temperature value and time after the change in the temperature time series. A battery information processing device characterized by selecting the temperature value and time before the change in the temperature time series if the temperature change Δθ is positive.

15. A battery information processing method applied to a battery monitoring system that monitors battery packs, A voltage change calculation step that calculates the voltage change ΔV of the battery pack based on the input voltage time series, A voltage change estimation step in which the voltage change ΔVest of the battery pack is estimated based on the input voltage time series, It includes a time series selection step that selects and outputs an input voltage time series, The battery information processing method is characterized in that, in the time series selection step, when the difference between the voltage change ΔV and the voltage change ΔVest is smaller than a predetermined value, either the value before or after the change in the voltage time series, or the average value thereof, is selected.

16. In the battery information processing method of claim 15, A battery information processing method characterized in that, in the time series selection step, if the voltage change ΔV is 0, the input voltage time series is not output.

17. A battery information processing method applied to a battery monitoring system that monitors battery packs, A temperature change calculation step that calculates the temperature change Δθ of the battery pack based on the input temperature time series, A voltage change estimation step in which the temperature change Δθest of the battery pack is estimated based on the input temperature time series, It includes a time series selection step that selects and outputs an input temperature time series, A battery information processing method characterized in that, in the time series selection step, if the difference between the temperature change Δθ and the temperature change Δθest is smaller than a predetermined value, one of the values ​​before or after the change in the temperature time series, or the average value thereof, is selected.

18. In the battery information processing method of claim 17, A battery information processing method characterized in that, in the time series selection step, if the temperature change Δθ is 0, the input temperature time series is not output.

Images