CN117723994A

CN117723994A - Method and device for detecting anomalies in a device battery of a technical device

Info

Publication number: CN117723994A
Application number: CN202311205406.9A
Authority: CN
Inventors: 李实�
Original assignee: Robert Bosch GmbH
Current assignee: Robert Bosch GmbH
Priority date: 2022-09-19
Filing date: 2023-09-18
Publication date: 2024-03-19
Also published as: DE102022209795A1

Abstract

A method for identifying a critical fault in a device battery based on an assessment of the operational parameter trend of a large number of device batteries, having the steps of: -providing a plurality of operating parameter profiles of the device batteries during a predefined load mode, respectively, the load mode indicating a defined time profile of the load parameters of the device batteries, -identifying at least one of the device batteries as abnormal according to at least one predefined abnormality criterion, wherein the at least one abnormality criterion evaluates the operating characteristics in relation to the respective operating characteristics of all remaining device batteries; -identifying at least one device battery as faulty if the relevant device battery is identified as abnormal for each abnormality criterion for a predetermined number of evaluations for the successively given load patterns; -identifying a severe fault of a specific device battery from the frequency distribution of the number of fault identifications of all device batteries and from the number of fault identifications of the specific device battery.

Description

Method and device for detecting anomalies in a device battery of a technical device

Technical Field

The present invention relates to a method for diagnosing a device battery for a technical device, in particular for diagnosing a device battery by means of abnormality detection and for detecting a serious fault.

Background

The energy supply of electrical devices and machines, such as, for example, electrically drivable motor vehicles, which are operated independently of the electrical network is usually carried out with a device battery or a vehicle battery. These device batteries or vehicle batteries provide electrical energy for operating the device.

The device battery decays during its lifetime and according to its load or use. So-called aging results in a continuous decrease of the maximum power capacity or storage capacity. The aging state corresponds to a measure for indicating aging of the accumulator. Conventionally, new device batteries can have a 100% state of aging (SOH-C in terms of their capacity) that decreases significantly over the lifetime of the device battery. The degree of ageing (temporal change in the ageing state) of the device battery depends on the individual load of the device battery, that is to say, for the vehicle battery of the motor vehicle, on the behavior of the driver, the external environmental conditions and the type of vehicle battery. In addition, in addition to this typical periodic aging, malfunctions or functional malfunctions may also occur in the battery cells of the device battery.

In technical installations operating with batteries, for safety reasons, in particular in the case of high energy densities, the device battery used must be monitored for faults in a defined manner. If a battery cell, a unit consisting of a plurality of battery cells, or the entire device battery fails, the technical device may become unable to function properly in response to the failure occurring, and the safety of the technical device and the user may also be compromised in the event of a functional failure that leads to a strong temperature rise. It is therefore necessary to inform the abnormality of the battery cells of the device battery in advance and reliably.

For monitoring the system battery from a large number of systems, the operating variable data are usually continuously recorded and transmitted as an operating variable profile block by block to a central unit outside the system. In order to evaluate operating variable data, in particular in a physical or electrochemical cell model based on differential equations, the operating variable data is sampled in the form of a trend with a relatively high time resolution (sampling rate) of between 1Hz and 100Hz, for example, and the cell state is ascertained therefrom by means of a time integration method.

For example, it is possible to use and adjust an electrochemical cell model for a device battery having a similar battery cell or a similar cell chemical cell in the central unit, so that the battery state can be ascertained and monitored.

Publication WO2011/077540A1 discloses a fault recognition device for a battery pack having a plurality of series-connected battery cells. A plurality of identification units are provided, which are each associated with a plurality of battery cells, wherein each identification unit is designed to compare the voltage output of the battery cell associated therewith with a predetermined nominal voltage. The current of the plurality of battery cells is detected by means of a current detector, wherein the plurality of detection units output a fault detection signal that indicates whether the battery pack has a battery cell that outputs a voltage that is reduced such that it is lower than the nominal voltage when discharging. The fault monitoring device is used for monitoring whether an internal resistance fault occurs based on the fault identification signal and the current value, wherein the internal resistance fault indicates that one of the battery single cells has the internal resistance which is increased so that the internal resistance exceeds an upper limit value.

Publication CN 110350258B discloses an anomaly identification based on voltage and current detection, temperature and smoke sensing.

Publication DE 10 2014 204 956 A1 or US2017/0082693A1 discloses a method for detecting anomalies in a battery cell, wherein events are detected by a short-circuit sensor device in a signal of the terminal voltage of the battery cell, which events have successive edges at time intervals of microseconds, and wherein the events detected by the short-circuit sensor device are transmitted to a battery management system.

Publication US 2009/012859 A1 discloses a method for detecting an abnormality of a battery pack including a secondary battery composed of at least one cell and a voltage detection circuit for measuring a cell voltage of the secondary battery, wherein the cell voltage is measured and it is judged if a predefined abnormality judgment condition in terms of an internal short circuit of the cell and an abnormality of the voltage detection circuit is satisfied in the case of using the measured cell voltage, and wherein if the abnormality judgment condition is satisfied, it is determined that the internal short circuit of the cell and/or the abnormality of the voltage detection circuit has occurred.

Publication "abnormality detection during lithium ion battery authentication test (Anomaly Detection During Lithium-ion Battery Qualification Testing)" by 2018 of the publication of Saxena, M.Kang, Y.Xing, and M.Pecht, published in 2018IEEE International prediction and health management conference (International Conference on Prognostics and Health Management, ICPHM), discloses an abnormality identification method based on identification of a loss of battery capacity on pages 1-6.

Most of the methods of the prior art are based on rule-based evaluation of the measurements of the battery management system in order to determine anomalies. These methods have the disadvantage that they are generally based on a comparison of standard measurements of the battery management system or values derived therefrom with normal ranges or abnormal thresholds. Since the measured voltage/resistance is very sensitive with respect to temperature, SOC, measurement noise and even with respect to load conditions, it is difficult to determine a reliable comparison value for anomaly identification. Furthermore, the hardware overhead is expensive in the prior art solutions. Abnormal cells may change during operation of the vehicle due to cell-to-cell differences, different load conditions experienced by each cell, and cell balance.

Disclosure of Invention

According to the invention, a method for detecting anomalies in a device battery of a technical device, and a corresponding device according to the parallel claims are provided according to claim 1.

Further embodiments are found in the dependent claims.

According to a first aspect, a method for identifying a critical fault in a device battery based on an assessment of the temporal operating parameter trend of a large number of device batteries is proposed, the method having the following steps:

providing a plurality of operating parameter profiles of the device battery during predetermined load modes, respectively, which indicate a defined temporal profile of the load parameters of the device battery, in particular of the battery current,

-identifying at least one of the device batteries as abnormal based on at least one predefined abnormality criterion, wherein the abnormality criterion evaluates an operating characteristic in relation to a corresponding operating characteristic of all remaining device batteries;

-identifying the at least one device battery as faulty in a fault identification if for evaluation of a predetermined number of consecutive predetermined load patterns the relevant device battery is identified as abnormal in respect of each of the at least one abnormality criterion;

-identifying a severe fault of a specific one of the device batteries from a frequency distribution of the number of fault identifications of all device batteries and from the number of fault identifications of the specific device battery.

Device batteries for energy intensive applications often have a large number of battery cells. The battery cells are electrically connected and thus can supply high currents and high voltages for the operation of the technical installation. In order to ensure the operation of the technical installation and the safety of its users, the method described above provides for determining anomalies in the installation battery or in the battery cells of the installation battery. This enables countermeasures to be taken in time or warnings to be given to the user so that the user can leave the dangerous scope of action of the device battery if necessary. The method provides that the running parameter trend of each battery cell of the battery of the device is evaluated. This is preferably done in a central unit remote from the device in order to save resources in the technical device.

In order to detect anomalies, the operating variable profile is detected in a predefined load pattern during a time period. The load mode, for example, presets the course of the battery current during the duration. The load pattern is known, for example, precisely from the charging characteristic curve during the charging process, so that the duration during the charging phase is particularly suitable for a subsequent evaluation of the anomaly detection.

The detected operating parameter profiles include cell voltage of the battery cells, battery current at the module level, and battery temperature. The operating parameter trend is filtered and outliers are removed to remove outliers, remove unsynchronized points, and fill in data gaps.

At least one operating characteristic is now evaluated on the basis of the operating parameter trend detected during the duration of the predefined load mode and evaluated on the basis of the corresponding abnormality criteria.

Further, the at least one anomaly criterion can include at least one of:

evaluating the cell voltage after a defined load step of a predefined load pattern with respect to the cell voltages of all the device batteries after the defined load step of the predefined load pattern,

-evaluating the extremum of the quotient of the charge difference and the voltage difference with respect to the extremum of the quotient of the charge difference and the voltage difference of all the device batteries; and is also provided with

-evaluating the temperature difference between the start time of the load mode and the end of the load mode with respect to the temperature difference between the start time of the load mode and the end of the load mode for the battery temperatures of all the device batteries.

Thus, as an operating characteristic, the at least one abnormality criterion can be evaluated, for example, for the cell voltage of the battery cell, the extremum of the quotient dQ/dU, and the temperature difference of the battery temperature before and after the operation in the load mode.

The three anomaly criteria mentioned above and possibly further anomaly criteria are preferably evaluated jointly.

The anomaly criteria are evaluated in the central unit based on a frequency distribution of operational characteristics across a large number of device batteries. For this purpose, the frequency distribution of the respective operating characteristics is determined on all the device batteries.

The device battery can now be classified as abnormal by an abnormality criterion according to each operating characteristic, wherein a large number of characteristics of the device battery are given an evaluation metric. Thus, abnormal device batteries can be identified by means of a clustering method such as, for example, DBSCAN, k-means, etc., based on the operation characteristics of the respective device batteries, respectively. Further, abnormal device batteries can be identified by evaluation of kurtosis and/or skewness with respect to the assumed normal distribution. In addition, the device battery whose operating characteristic value is below and/or above a corresponding operating characteristic threshold value, which can be specified by a lower and/or upper score value, can be identified as an abnormal device battery.

As a result, the following description is obtained for each device battery and for each associated operating characteristic, namely: the relevant device battery is not abnormal with respect to the checked operating characteristics, that is to say has an abnormal character.

The device battery is now identified and identified as faulty by: the device battery has been identified as abnormal by using the abnormality criteria for all operating characteristics and for a predefined (minimum) number of consecutive checks. The point at which these device batteries are identified as faulty corresponds to fault identification. The fault identification is performed for each device battery, thereby generating a frequency distribution of the fault identification over a large number of device batteries.

It can be provided that the identification of a critical fault of a specific one of the device batteries is carried out on the basis of a threshold number of said fault identifications, which is derived from a frequency distribution of the number of fault identifications of all device batteries determined to be faulty. The threshold number of times can be determined in particular by a predefined quantile value, such as, for example, a 95% quantile.

Furthermore, a severe fault of a specific one of the device batteries can be identified from a voltage difference threshold exceeding a maximum voltage difference of the cell voltage with respect to the average cell voltage, wherein the voltage difference threshold is derived from a frequency distribution of the maximum voltage differences of all device batteries determined to be faulty.

Accordingly, the frequency distribution of the individual cell voltages (maximum or minimum cell voltages) of all battery cells relative to the maximum voltage difference between the average cell voltages is determined on all device batteries, and in particular before the balancing method is used. Particularly in the case where the average cell voltage deviates from the individual cell voltages by a relatively high degree, the severity of the abnormality found is higher than in the case where the voltage difference is small.

The frequency distribution of the number of fault detection times is obtained on all the device batteries, and furthermore, the frequency distribution of the maximum voltage difference is obtained especially on all the device batteries identified as faulty. The device battery is identified as faulty, the number of times of fault identification of the device battery exceeds a certain threshold number of times and the maximum voltage difference of the device battery exceeds a certain voltage difference threshold, and a corresponding warning is issued to the user or measures are taken to control the operation of the corresponding device battery.

For example, the specific threshold number of times and/or the specific voltage difference threshold can be determined by a respectively predefined quantile limit (quantile), for example a quantile limit, for example a 95% quantile limit, with respect to the respective frequency distribution. The quantile limit can be predefined, for example, by the manufacturer of the battery cells.

The above-described method is thus enabled to identify a device battery that is faulty in severity and must cause a warning to the user and/or a limitation to the operation of the technical device by taking into account one or more anomaly criteria and taking into account the results of fault recognition in a large number of device batteries and evaluation of at least one operating characteristic. In this case, a reference is predefined by the operating characteristics of the device battery of a large number of devices as to when an abnormality detection should be evaluated as a serious abnormality.

Drawings

The embodiments are explained in detail below with the aid of the figures. Wherein:

FIG. 1 shows a schematic diagram of a system for implementing anomaly identification for a device battery of a large number of vehicles of a fleet;

FIG. 2 shows a flow chart illustrating a method for identifying a faulty device battery;

fig. 3a to 3c show load patterns for the charging process as references for checking the anomaly criteria and the resulting frequency distribution of the operating characteristic values based on the respective anomaly criteria; and is also provided with

Fig. 4a and 4b show the frequency distribution of the number of fault identifications of a faulty vehicle battery or the frequency distribution of the maximum voltage difference of the battery cells of a faulty vehicle battery.

Detailed Description

The method according to the invention is described below with the aid of a vehicle battery as a device battery in a large number of motor vehicles as the same type of device. For this purpose, the operating variable trend is received in the central unit and evaluated there. For this purpose, the central unit evaluates the anomaly criteria.

The above examples represent a large number of fixed or mobile devices with grid-independent energy supply, such as for example vehicles (electric vehicles, mopeds, etc.), equipment, machine tools, household appliances, IOT devices, etc., which are in connection with a central unit (cloud) outside the device via a corresponding communication connection (e.g. LAN, internet).

Fig. 1 shows a system 1 for collecting fleet data in a central unit 2 for implementing a method for monitoring the functions of a vehicle battery and for identifying anomalies. Fig. 1 shows a fleet 3 with a plurality of motor vehicles 4.

One of the motor vehicles 4 is shown in more detail in fig. 1. The motor vehicle 4 has a vehicle battery 41 with a battery cell 45, an electric drive motor 42 and a control unit 43. The control unit 43 is connected to a communication device 44, which is suitable for transmitting data between the respective motor vehicle 4 and the central unit 2 (so-called cloud).

The control unit 43 is in particular designed to detect operating variables of the vehicle battery 41 with a high time resolution, such as, for example, between 1Hz and 50Hz, such as, for example, 10Hz, and to transmit them to the central unit 2 via the communication device 44.

The motor vehicle 4 transmits operating variables F to the central unit 2, which indicate at least the variables characterizing the operation of the vehicle battery. The operating variables F can indicate, for the vehicle battery, not only at the group level, at the module level and/or at the cell level, the instantaneous battery current, the instantaneous battery voltage, the instantaneous battery temperature and the instantaneous State of Charge (SOC). The operating variable F is detected as an operating variable trend in a rapid time grid from 0.1Hz to 50Hz and can be transmitted periodically to the central unit 2 in uncompressed and/or compressed form. For example, in order to minimize the data traffic to the central unit 2, the time series can be transmitted to the central unit 2 block by block at intervals of 10min to several hours, using a compression algorithm.

The central unit 2 has a data processing unit 21 in which a part of the method described below can be carried out, and a database 22 for storing operating parameter trends and the like. The central unit 2 is designed to receive the operating variable trend and to perform an anomaly detection for each of the connected vehicle batteries.

Fig. 2 shows a flow chart of a method for detecting a faulty vehicle battery 41, which is carried out in the central unit 2. The method is based on an evaluation of a large number of vehicle batteries 41 of the vehicle 4, which transmit the behavior of the operating variables, such as the cell voltage of all battery cells, the battery current, the battery temperature and the state of charge, to the central unit 2 at least during a predefined known load mode.

In step S1, the operating variables are received and in step S2 it is checked whether a predefined load pattern is present for a predefined duration of, for example, between 5min and 30min, preferably 10 min.

The load pattern is predefined and corresponds, for example, to a load pattern of the vehicle battery 41 during charging with a predefined charging profile. The load pattern should be the same for all vehicles 4 or vehicle batteries 41 so that anomalies in the battery cells can be similarly identified according to anomaly criteria. The operation of the charging profile can be signaled separately accordingly or determined for the relevant vehicle battery 41 by monitoring the operating parameter profile of each of the vehicle batteries 41. The load pattern is assumed in this regard to be a temporal progression of the battery current.

If the predefined load pattern is detected in the operating variable profile of at least one vehicle battery 41 for a predefined duration (alternative: yes), the method is continued in step S3. Otherwise, the process jumps back to step S1.

In step S3, the operating variable trend is filtered, subjected to outlier detection, and/or data gaps are filled in, according to methods known per se.

In step S4, one or more abnormality criteria are now checked for each vehicle battery 41 for which a predefined load pattern has been identified, based on the operating characteristics derived from the operating parameter profiles of the predefined load pattern.

First, as an example for an anomaly criterion, the cell voltages of all battery cells 45 of all vehicle batteries 41 for which a predefined load pattern has been determined are detected after one or more specific load steps S, and the frequency distribution of the frequency H of the generated cell voltages Ucell is determined for each of the specific load steps, respectively. For example, the corresponding frequency distribution H is shown for two load steps S in the load mode of fig. 3 a. In particular, the cell voltage Ucell is detected during a charging process according to a predefined charging profile immediately after a battery current increase. The frequency distribution H of the cell voltages of all the vehicle batteries 41 is derived for each specific load step.

As another example of evaluation for the abnormality criteria, the extremum of the quotient dQ/dU in the constant current phase of the specific load mode can also be detected for each vehicle battery 41 (each time when the specific load mode has been determined) as shown in fig. 3 b. The extreme value dQ/dU can be ascertained for all relevant vehicle batteries 41 accordingly and evaluated in the corresponding frequency distribution H of the determined values.

Furthermore, the temperature difference Δt of the battery temperature can be ascertained before and after the operation of the predetermined load mode of the vehicle battery 41 (each time when a specific load mode has been determined), and evaluated by means of a frequency distribution on all vehicle batteries 41, as shown in fig. 3 c. The temperature difference Δt can be ascertained for all vehicle batteries 41 accordingly, and evaluated in the respective frequency distribution of the determined values.

In step S5, the frequency distribution H is now evaluated individually, for example, in terms of deviations from the average characteristics of all vehicle batteries 41 of the fleet 3. The evaluation involves each of the operating characteristics considered according to the abnormality criteria so as to identify the vehicle battery 41 as abnormal if each of the abnormality criteria is satisfied. Thereby, it is possible to detect and identify the abnormal vehicle battery 41 in each operation characteristic. For this purpose, for example, clustering algorithms can be used with respect to the running characteristics of the anomaly criteria, such as, for example, DBSCAN (density-based clustering algorithm), K-means clustering, etc.

Methods of detecting a faulty vehicle battery by kurtosis and slope of the frequency distribution, such as, for example, by isolated forest analysis, can also be used.

If the relevant operating characteristic exceeds and/or falls below one or more operating characteristic thresholds associated with the respective operating characteristic, an abnormal vehicle battery can also be identified in terms of one or more of the operating characteristics. The respective operating characteristic threshold value is derived from a fractional observation of the respective frequency distribution H associated with the respective operating characteristic. For example, a lower quantile value, e.g., 5% quantile, can be defined for a lower threshold of the relevant operating feature and/or an upper quantile value, e.g., 95% quantile, can be defined for an upper threshold of the relevant operating feature.

The operating characteristic threshold can also be predefined based on a frequency distribution corresponding to the expected failure rate of the battery manufacturer.

In step S6, if the vehicle battery 41 is identified as abnormal according to a predefined abnormality criterion in a successive check of a predetermined number of times for all operating characteristics, a faulty vehicle battery is now identified. For each of the faulty vehicle batteries 41 thus identified, the fault counter is incremented for each instance of the faulty vehicle battery 41 identified.

For all faulty vehicle batteries 41 that were previously identified at least once, the number of faulty identifications is evaluated accordingly in a frequency distribution.

In fig. 4a, the frequency of the fault detection of the faulty vehicle battery 41 is seen in the frequency distribution.

In addition, in step S7, if the number of times of failure recognition is recognized for the vehicle battery 41 exceeds a predetermined threshold number of times, which is derived from the frequency distribution of the number of times of failure recognition for all the device batteries determined to be failed, a serious failure can be recognized. The threshold number of times can be determined in particular by a predefined quantile value, such as, for example, a 95% quantile.

Additionally, as a criterion for the existence of a serious failure of the vehicle battery 41, the maximum voltage difference can be evaluated. For this purpose, the maximum voltage difference between the cell voltage and the average cell voltage is ascertained for the checked vehicle battery 41 and checked against a predefined voltage difference threshold. If the maximum voltage difference exceeds a predefined voltage difference threshold, a critical fault is identified.

The voltage difference threshold can be derived from the frequency distribution of the maximum voltage differences of all the vehicle batteries 41 determined to be faulty. The frequency distribution can now be evaluated by the described threshold comparison on all faulty vehicle batteries, as shown in fig. 4 b.

The predefined threshold value can be determined from the frequency distribution by a predefined quantile value, such as, for example, a 95% quantile.

In particular, before the balancing process is performed, a voltage difference in absolute terms should be produced between the average value of the cell voltages of the battery cells of the respective vehicle battery 41 and the maximum and/or minimum cell voltage of the particularly faulty battery cell.

In step S8, a vehicle battery 41 is now identified which is above a respective predetermined threshold value both in the frequency distribution with respect to the number of fault recognitions and in the frequency distribution with respect to the maximum voltage difference.

If both criteria are met, the relevant vehicle battery 41 is identified as a vehicle battery with a serious fault, and in step S9 a respective warning is issued to the user of the vehicle or vehicles with the respective vehicle battery 41 and/or the operation of the respective vehicle battery 41 is restricted.

Claims

1. A computer-implemented method for identifying a severe fault in a device battery (41) based on an assessment of an operational parametric trend of a large number of device batteries, the method having the steps of:

providing (S1) a plurality of operating parameter profiles of the device battery (41) during a predefined load mode, which indicates a defined temporal profile of the load parameters of the device battery (41), in particular of the battery current,

-identifying (S5) at least one of the device batteries (41) as abnormal according to at least one predefined abnormality criterion, wherein the at least one abnormality criterion evaluates an operating characteristic in relation to a respective operating characteristic of all remaining device batteries (41);

-identifying (S6) the at least one device battery (41) as faulty in a fault identification if the relevant device battery (41) is identified as abnormal for each of the at least one abnormality criteria for a predetermined number of evaluations for consecutive predetermined load patterns;

-identifying (S8) a severe fault of a specific one of the device batteries (41) from a frequency distribution of the number of fault identifications of all device batteries (41) and from the number of fault identifications of the specific device battery (41).

2. The method of claim 1, wherein the at least one anomaly criterion comprises at least one of:

-evaluating the extremum of the quotient of the charge difference and the voltage difference with respect to the extremum of the quotient of the charge difference and the voltage difference of all device batteries (41); and is also provided with

-evaluating the temperature difference between the start time of the load mode and the end of the load mode with respect to the temperature difference between the start time of the load mode and the end of the load mode for the battery temperatures of all the device batteries (41).

3. Method according to claim 1 or 2, wherein the device battery (41) is identified as abnormal based on the result of a clustering method of the operational features of the at least one abnormality criterion, in particular based on DBSCAN or k-means clustering.

4. Method according to claim 1 or 2, wherein the device battery (41) is identified as abnormal based on the results of an isolated forest method for analyzing the kurtosis and/or skewness of the frequency distribution of the operational features of the at least one abnormality criterion over all device batteries (41).

5. The method according to claim 1 or 2, wherein the device battery (41) is identified as abnormal according to a density function determined from a frequency distribution of operational characteristics of the at least one abnormality criterion over all device batteries (41) and at least one operational characteristic threshold.

6. The method according to any of claims 1 to 5, wherein the identification of the critical faults of a specific one of the device batteries (41) is carried out according to a threshold number of times of the fault identification, which is derived from a frequency distribution of the number of times of fault identification of all device batteries (41) determined to be faulty.

7. Method according to any of claims 1 to 6, wherein the identification of a severe fault of a specific one of the device batteries (41) is implemented according to a voltage difference threshold of a maximum voltage difference of the cell voltage and an average cell voltage, wherein the voltage difference threshold is derived from a frequency distribution of the maximum voltage differences of all device batteries (41) determined to be faulty.

8. Apparatus for carrying out one of the methods according to any one of claims 1 to 7.

9. Computer program product comprising instructions which, when the program is executed by at least one data processing mechanism, cause the data processing mechanism to perform the steps of the method according to any one of claims 1 to 7.

10. A machine readable storage medium comprising instructions which, when executed by at least one data processing mechanism, cause the data processing mechanism to perform the steps of the method according to any one of claims 1 to 7.