DE102019212909A1 - Method for detecting a fault in a battery system as well as battery system and motor vehicle - Google Patents

Abstract

The invention relates to a method for detecting a fault in a battery system (11), several battery components (13), each of which has at least one component-specific monitoring unit (14), and a computing device (15) being provided in the battery system (11). The invention provides that the computing device (15) receives a respective monitoring signal (17) from each of the at least one component-specific monitoring unit (14) from each of the battery components (13), so that several monitoring signals (17) from different ones of the battery components (13) are received. are present in the computing device (15), and a monitoring value of a statistical measure formed from the monitoring signals (17) is checked for plausibility by means of a predetermined correlation-based classification method (18).

Description

The invention relates to a method for detecting a fault in a battery system. The method uses the fact that a plurality of battery components are present in the battery system, each of which has at least one component-specific monitoring circuit or monitoring unit, the monitoring signals of which are correlated with one another. The invention also relates to a battery system by means of which the method can be carried out, and to a motor vehicle with such a battery system.

A battery system can be based on a parallel and serial connection of individual battery cells. As a rule, the total current of all interconnected battery cells is measured, and the temperature is determined at several points in battery modules, in each of which some of the battery cells are combined. The cell voltage of all individual battery cells is measured individually in each battery cell or on a string of battery cells connected in parallel. The measurements are intended to reliably monitor battery cells, in particular those based on lithium-ion technology, for critical states, such as overvoltage or undervoltage, for example. The measurements can be evaluated by a central battery management system (BMS) or in individual control circuits of the battery modules or the battery cells. The evaluation usually monitors the battery cells on the basis of the available voltage, current and temperature limits.

In the case of collective measurements of current and temperature at only a few measuring points and due to the simple comparison of the measurement results with threshold values, it has proven to be difficult to detect errors in a battery system at an early stage, the influence of which increases only relatively slowly over time (for example over several hours or Days away). For example, it is difficult to detect errors relating to an increasing internal resistance, a creeping internal short circuit or a decrease in a cell capacity for energy storage before damage occurs to the battery system.

Such a battery system with individual voltage measurement on each battery cell and a collective measurement for the current intensity in individual strings of serially connected battery cells is for example from the DE 11 2017 000 969 T5 known. A controller receives the measured values from all sensors and evaluates them with regard to the status of the battery system. An early detection of creeping errors, the influence of which is gradually increasing, is not described.

From the DE 10 2010 047 960 A1 a battery system with a sensor network is known in order to be able to detect the field voltages of individual battery cells with high accuracy. It is not described how an early detection of creeping errors can take place from this.

From the DE 10 2014 220 062 A1 a battery system is known which has an evaluation unit for each individual battery cell, by means of which it is determined by means of a probabilistic approach which battery cells can be used for the current load case. The battery cells can then be switched on or off individually so that only the battery cells identified as suitable are active. Error detection to the effect that it could be recognized why a battery cell is unsuitable is not described.

From the DE 10 2015 002 827 A1 It is known that a monitoring unit can be provided for each individual battery cell for recording cell-specific measured values, for example current, voltage and / or temperature. There is no evaluation to the effect of being able to recognize creeping errors at an early stage.

The invention is based on the object of providing fault detection for faults in individual battery components in a battery system.

The object is achieved by the subjects of the independent claims. Advantageous embodiments of the invention are described by the dependent claims, the following description and the figures.

The invention provides a method for detecting a fault in a battery system. The method is based on the fact that a plurality of battery components are provided in the battery system, each of which has at least one component-specific monitoring circuit or monitoring unit. Such a battery component can be, for example, a single battery cell, or it can be a battery module in which some of the battery cells of the battery system are combined. A monitoring unit can, for example, have a measuring circuit with one sensor or a plurality of sensors. A monitoring unit can have an electronic circuit with at least a microcontroller and / or at least one microprocessor, for example to operate a state model of the monitored battery component.

The method can be carried out by a computing device which is also part of the battery system. It can be based on at least one microcontroller and / or at least one microprocessor. The processing unit can be a central processing unit for the battery system and this central processing unit can receive monitoring signals from all battery cells. The computing device thus receives a respective monitoring signal of the respective at least one component-specific monitoring unit of the respective battery component from each of the battery components. Such a monitoring signal can describe, in analog or digital form, a respective temporal course of a monitoring variable (e.g. a state variable or measured variable). Since the computing device receives at least one monitoring signal from each of the battery components, there are a total of several monitoring signals from different battery components in the computing device. The computing device now uses these monitoring signals to calculate at least one monitoring value of a predetermined statistical measure (e.g. a correlation) on the basis of some or all of the monitoring signals and to calculate an expectation interval or an expectation corridor for the at least one monitoring value on the basis of correlation data stored in the computing device to pretend. The statistical measure depends in particular on a degree of correlation of the monitoring signals. It is not the exact time course of the at least one monitoring signal that is checked, but a statistical behavior for which an expectation interval or expectation corridor is specified, within which the respective monitoring value lies if there is no error, i.e. in the case of the absence of the error. The expectation corridor is specified on the basis of correlation data stored in the computing device. To calculate and check the at least one monitoring value, the computing device uses a correlation-based classification method. Even if it is not possible to determine an exact course for the error-free case from at least one of the monitoring signals, the correlation data are nevertheless used to estimate or specify in which expectation corridor, i.e. within which limits or within which interval, the at least one monitoring value that was calculated for the statistical measure should lie or should run if there is no fault in the battery system. The statistical measure is a correlation-based statistical measure, that is to say a statistical variable which describes a statistical relationship between different ones of the monitoring signals, for example their correlation coefficients.

It can now be checked whether the at least one monitoring value actually lies or runs within the expectation corridor. The test result is referred to here as the classification result. The computing device thus determines the classification result which indicates whether the at least one monitoring value deviates from the expectation corridor determined for this monitoring value in accordance with a predetermined deviation criterion. If, according to the classification result, the deviation criterion is met (i.e. there is a deviation from the expected corridor according to the deviation criterion), the computing device generates an error signal that signals that a faulty battery component is present, i.e. a battery component has a fault.

The method is based on said correlation data. These can be, for example, a respective correlation coefficient between two monitoring signals. The calculation of a correlation coefficient is known per se. If the time course of the monitoring signals and the correlation coefficient is now available, the expectation corridor can be used to identify from the correlation data whether the monitoring signals behave in accordance with the underlying statistical measure as is to be expected in the error-free case. If two monitoring signals are correlated when the battery system is in a fault-free state, so that the correlation coefficient for the fault-free state is known, then, if the fault state of the battery system is unknown, a time curve of one of the monitoring signals can be measured and then checked using the correlation coefficient to determine whether the statistical behavior is in the expectation corridor. If the at least one monitoring value lies outside the expectation corridor, then a changed correlation coefficient results, this can be interpreted as an indication of an error.

As already described, the correlation data used or required can be determined from the monitoring signals of a fault-free battery system. These can be generated, for example, by means of a fault-free prototype of the battery system to be monitored or after the battery system has been completed. In a further embodiment, different faults can also be introduced into the prototype in a targeted manner one after the other in order to learn this faulty behavior and then to classify between these faults.

The invention has the advantage that a change in the correlation of the monitoring signals with one another can already be detected and this is used as an indication of an error. A so-called insidious error that prevents continued operation of the Battery system allows, but increases over time until a failure occurs, can be recognized early on, even before the failure of the battery system or one of the battery components, from a changed statistical behavior of the monitoring signals with regard to their correlation.

The invention also encompasses embodiments which result in additional advantages.

In one embodiment, said deviation criterion (on the basis of which the deviation of the at least one monitoring value from the respective expectation corridor is detected) comprises that the respective monitoring value to be checked runs outside the expectation corridor for at least a predetermined minimum period of time and / or by more than a predetermined tolerance value from the expectation corridor deviates. Observing a minimum period of time has the advantage that a statistically improbable but nevertheless permissible circuit state of the battery system is not misinterpreted as an error. The minimum duration can be in a range with a lower limit of 0.5 seconds to 2 minutes on the one hand to an upper limit of 5 minutes to 1 hour on the other hand.

The expectation corridor can be defined as a tolerance interval or as a threshold value and / or on the basis of a statistical hypothesis test. The deviation criterion then includes that the respective monitoring value lies outside a predetermined tolerance range and / or above or below a threshold value.

In one embodiment, said classification method comprises at least one of the following: an outlier detection, an artificial neural network, a principal component analysis (PCA), an independent component analysis (ICA - independence analysis). This classification method has proven to be reliable when it comes to the detection of a discrepancy between at least two monitoring signals for which a relationship between the time courses can only be described in a statistical sense on the basis of correlation data, i.e. if a description of a deterministic, functional relationship cannot possible or too complex. Since monitoring signals can deviate from one another even in fault-free operation, the classification methods described enable a reliable differentiation between a deviation due to non-critical influences, such as temperature differences between the battery cells and / or a measurement noise, on the one hand, and a critical deviation due to an actual fault in one of the battery components on the other hand possible.

As already stated, a monitoring signal is to be understood in particular as a time sequence of individual signal values, namely measured values or status values or parameter values, that is to say generally a series or sequence of signal values. A current signal value can therefore be taken from each monitoring signal at a given point in time. If the current signal value of these monitoring signals is taken from a predetermined set of monitoring signals at a given measurement time, then a vector can be formed from these individual values so that an N-dimensional vector is obtained with N signal values, which is described here as a signal vector. N is greater than 1, in particular greater than 2 or 3.

The correlation data can describe the expected statistical behavior in the error-free case for such a signal vector. Accordingly, one embodiment provides that the at least one monitoring value comprises a paired covariance or correlation coefficient of the monitoring signals and / or a T ² value and / or squared prediction error value, SPE value, formed from the monitoring signals. The deviation criterion then includes in particular that a threshold value comparison for the covariance or the correlation coefficient and / or a Student T distribution and / or a T ² value and / or squared prediction error, SPE, and / or a statistical hypothesis test is determined becomes. It is therefore possible to check for the current signal vector from signal values whether it corresponds to a given statistic or with what probability it corresponds to this statistic. Suitable calculation methods for the values mentioned can be found, for example, in the publication D ražen Slišković, Ratko Grbić, Željko Hocenski, “Multivariate statistical process monitoring,” Tehnicki Vjesnik, vol. 19, 1 (2012), pp. 33-41, 2012 and the publication S. Yin, SX Ding, A. Haghani, H. Hao, and P. Zhang, "A comparison study of basic data-driven fault diagnosis and process monitoring methods on the benchmark Tennessee Eastman process," Journal of Process Control, vol. 22, no.9, pp. 1567-1581, 2012 , can be taken. Error detection can then be implemented by comparing the threshold value with a respectively predetermined threshold value. The threshold value can, for example, correspond to a significance limit with a percentage that indicates the probability with which the current signal vector was generated by a statistical process that would result in a fault-free battery system or (vice versa) how improbable it is that the signal vector is from such a process Process originates.

A particularly preferred embodiment provides for this that several predetermined Measurement times or checking times, the current signal values of the monitoring signals are recorded or read out and the current signal values are combined to form said signal vector and this signal vector is converted into principal components or main components by means of PCA base vectors that are specified by the correlation data. The coefficients of the main components are thus calculated in a manner known per se by means of the PCA. For some or all of these principal components, said T ² value is calculated as a monitoring value for the Student T-distribution and / or for some or all of the principal components said SPE value is calculated as a monitoring value. For the respective value (T ² value and / or SPE value), a threshold value comparison is carried out with a respective predetermined threshold value and the deviation criterion is met if the respective threshold value is exceeded for a predetermined number of checking times. The respective signal vector can therefore be checked for a single threshold value (T ² value or SPE value) or for two threshold values (T ² value and SPE value). This test has proven to be particularly robust against false tripping (false positive) and against overlooked errors (false negative).

In one embodiment, at least one of the following is signaled as said fault: a rising internal resistance, a rising contacting resistance, a falling storage capacity, an internal short circuit in the faulty battery component; a cell error (internal / external short circuit, general parameter change); an insulation failure; a deviation from the usual behavior (e.g. larger SOC or temperature differences than specified according to a standard behavior). This type of error can first be determined, before the respective battery component fails, by means of a deviation in the correlation of signal values from different battery components.

In one embodiment, the fault is localized (i.e. not only its detection, but also the indication of the location or faulty battery component) by recognizing which of the monitoring signals causes the deviation criterion to be met and which battery component whose monitoring unit has generated this monitoring signal, identified as the faulty battery component. In other words, if the deviation criterion is met, it is traced back which of the monitoring signals caused the deviation. Since this monitoring signal comes from a monitoring unit of one of the battery components, it can be assumed that this battery component is the faulty one.

In at least one embodiment, the at least one monitoring unit of each of the battery components generates a respective monitoring signal with signal values (i.e. sensor values and / or status values at individual monitoring times) relating to at least one of the following operating variables: an electrical voltage, an electrical current, a temperature, an internal resistance. If the monitoring signal does not relate to measured values but, for example, model parameters or status data, the monitoring data can relate to: parameter data at least one of the following parameters: internal resistance, capacitance, open circuit voltage, cell time constants; and / or status data of at least one of the following statuses: charge status, polarization voltage, polarization current, aging status. These types of monitoring data have proven to be reliable in terms of evaluation using a correlation-based classification method.

The invention also provides a battery system which can be monitored for a fault by means of an embodiment of the method according to the invention. The battery system is intended for a motor vehicle, for example as a traction battery or high-voltage battery (HV - electrical voltage greater than 60 volts). A plurality of battery components are provided in the battery system in a manner known per se, each of which has at least one component-specific monitoring unit. The battery system also has a computing device which is set up to carry out the embodiment of the method according to the invention. The computing device can be based on at least one microprocessor and / or at least one microcontroller. The computing device can have a data memory in which program data or program instructions are stored which, when executed by the computing device, cause the computing device to carry out the embodiment of the method according to the invention.

The invention also provides a motor vehicle in which an embodiment of the battery system according to the invention is provided. The motor vehicle according to the invention is preferably designed as a motor vehicle, in particular as a passenger vehicle or truck, or as a passenger bus or motorcycle.

The invention also includes the combinations of the features of the described embodiments.

• 1 eine schematische Darstellung einer Ausführungsform des erfindungsgemäßen Kraftfahrzeugs mit einer Ausführungsform des erfindungsgemäßen Batteriesystems;
• 2 ein Diagramm mit einem schematisierten Verlauf einer Fehlergröße über der Zeit;
• 3 ein Diagramm mit schematisierten Verläufen mehrerer Überwachungssignale über der Zeit;
• 4. eine Skizze zur Veranschaulichung einer Anwendung einer Hauptkomponentenanalyse auf Signalwerte;
• 5 Diagramme zur Veranschaulichung einer korrelationsbasierten Klassifikationsmethode mit einem Abweichkriterium auf Basis eines T²-Werts und eines SPE-Werts;
• 6 ein Flussschaudiagramm zur Veranschaulichung einer Ausführungsform des erfindungsgemäßen Verfahrens; und
• 7 ein Flussschaudiagramm zur Veranschaulichung einer Ausführungsform des erfindungsgemäßen Verfahrens mit einer Fehlerlokalisation.

Exemplary embodiments of the invention are described below. This shows:

• 1 a schematic representation of an embodiment of the motor vehicle according to the invention with an embodiment of the battery system according to the invention;
• 2 a diagram with a schematic course of an error size over time;
• 3rd a diagram with schematized courses of several monitoring signals over time;
• 4th . a sketch to illustrate an application of a principal component analysis to signal values;
• 5 Diagrams to illustrate a correlation-based classification method with a deviation criterion based on a T ² value and an SPE value;
• 6th a flow chart to illustrate an embodiment of the method according to the invention; and
• 7th a flow chart to illustrate an embodiment of the method according to the invention with an error localization.

The exemplary embodiments explained below are preferred embodiments of the invention. In the exemplary embodiments, the described components of the embodiments each represent individual features of the invention which are to be considered independently of one another and which also further develop the invention in each case independently of one another. Therefore, the disclosure is also intended to include combinations of the features of the embodiments other than those illustrated. Furthermore, the described embodiments can also be supplemented by further features of the invention already described.

In the figures, the same reference symbols denote functionally identical elements.

1 shows a motor vehicle 10 , which can be a motor vehicle, in particular a passenger car or truck, or a passenger bus. In the motor vehicle 10 can be a battery system 11 be provided, for example as a traction battery or high-voltage battery for a drive of the motor vehicle 10 can be designed. The battery system 11 can have a supply voltage of more than 60 volts in the motor vehicle 10 provide. To the battery system 11 can at least one vehicle component 12th be connected to using the energy from the battery system 11 to be operated and / or to be included in the battery system 11 feed in electrical energy. A possible vehicle component can be: an electric machine, an air conditioning compressor, a charger.

About said supply voltage in the battery system 10 can generate or provide the battery system 11 multiple battery components 13th have, which can be, for example, battery modules each with a plurality of battery cells or individual battery cells. An electrical connection of the battery components 13th with each other and with the at least one vehicle component 12th is in 1 not shown for the sake of clarity. The interconnection can be designed according to the prior art.

Every battery component 13th can each have at least one monitoring circuit or monitoring unit 14th each having, for example, at least one sensor for measuring measured values and / or an electronic circuit for monitoring or detecting a respective state of the battery component, for example a state of charge (SOC), and / or for determining at least one parameter value for a model the battery component, for example an internal resistance, can be designed.

In the battery system 11 can be a computing device 15th be provided, the at least one microprocessor and / or at least one microcontroller 16 may have that with the respective at least one monitoring unit 14th of the battery components 13th for a transmission of a respective observation signal or monitoring signal 17th from the respective monitoring unit 14th can be coupled. The coupling can be wired or radio-based, as is known per se from the prior art.

The computing device 15th can be based on the received monitoring signals 17th detect if any of the battery components 13th has an error. This can be done in the computing device 15th a correlation-based classification method 18th be implemented.

To illustrate the from the computing device 15th Feasible method for the detection of the error is referred to in the following to the other figures.

2 illustrates an exemplary possible failure 19th , which can consist, for example, in the fact that, over time t, a contacting resistance R, which is between one of the battery components 13th and the said interconnection can result, increases over time, which is shown here by a graph of a time course 20th of Contacting resistance R is shown. It is shown how the contact resistance R increases to 4 milliohms. It should be noted that the time course is shown in a multiple of 10 ⁴ seconds, that is, the error 19th becomes larger over a relatively long period of time of more than an hour and thus gains influence.

The error grows here within a time interval 21 from zero to a critical value of, for example, 4 milliohms.

3rd illustrates how within the critical time interval 21 the course of the monitoring signals 17th can result. An example is here as a monitoring signal 17th a time curve of a current I is shown. The monitoring signals shown 17th can from the different battery components 13th have been received. Because of the bug 19th can change the course over time 22nd result in a discrepancy that cannot be identified by a conventional signal comparison due to the error 19th caused can be seen.

4th and 5 illustrate a possible correlation-based classification method 18th to get the bug 19th from the time courses 22nd of the monitoring signals 17th to recognize. It is shown how over the time t at respectively specified checking times (for example every second or every 10 seconds or every 30 seconds or every minute) from the then available current signal values of the monitoring signals 17th the current signal value 23 used and the current signal values 23 to a signal vector 24 can be summarized. On the signal vector 24 Principal Component Analysis, PCA, can be used, resulting in weighting factors or coefficient values for the principal components 25th that result together again as a vector 26th can be interpreted. The main components of the PCA can be described by correlation data C, by means of which the main component vectors can be described or defined.

The main components 25th can now all or only some of them at least one statistical verification process 27 be taken as a basis. A statistical verification process 27 may consist of all or some of the main components 25th a T ² value can be calculated for a T ² statistic and a threshold value comparison 28 can be carried out with a predetermined threshold value T ² -b, which can be, for example, a significance limit, for example 5% significance limit. It is shown how within the time interval 21 the rate for exceeding the threshold value increases, i.e. a positive exceeding or an exceeding of the threshold value in the threshold value comparison 28 is recognizable.

As a second statistical verification process 27 An SPE value (SPE - Squared Prediction Error) can be calculated on the basis of some of the main components or all of the main components and here also in a threshold value comparison 29 a comparison can be carried out with a threshold value SPEb, which, for example, can also be calculated using a significance limit, for example 5% significance limit. Here, too, it can be seen that in the duration of the time interval 21 if the threshold value is exceeded by the threshold value comparison 29 is recognized.

The bottom diagram shows how an AND operation indicates that both the threshold comparison 28 as well as the threshold comparison 29 signal that a threshold value has been exceeded, which is then signaled as a detected error (symbolized by a plus symbol), while if none or only one of the threshold values is exceeded, no error is detected (symbolized by a minus symbol). Exceeding both threshold values is a criterion for deviation 30th represent, that is, the current signal values 23 differ significantly in statistical behavior from the behavior that is to be expected in the error-free case.

Is the deviation criterion 30th fulfilled, an error signal 31 which indicates that there is an error. The error signal 31 can also only be generated under the condition that, for several or a predetermined number of checking times, both threshold values are exceeded in the threshold value comparison 28 , 29 is recognized.

Instead of the correlation-based classification method 18th according to 4th Another method can also be used, such as is known, for example, from process monitoring for monitoring processes, such as can be carried out, for example, in industrial plants for production.

6th summarizes the steps of the procedure again. In one step S10 can monitor signals 17th are received, each of which can contain a time sequence of sensor values and / or cell state values and / or cell parameter values as a signal value. In one step S11 a process monitoring method can be carried out, for example the one in 4th illustrated correlation-based classification method 19th . This results in one step S12 a statistical evaluation, for example with the described T ² and SPE Statistics of the signal values on which the method is based 23 . This check leads to a classification result 32 or represents the classification result 32 in one step S13 it is decided whether the error signal 31 should be generated, i.e. whether the deviation criterion 30th is satisfied.

7th illustrates a further development of the procedure in which, in the event that in step S13 the deviation criterion 30th is fulfilled and thus an error signal 31 should be generated, this leads to that in one step S14 the existence of an error is assumed or taken as a basis and in one step S15 for the monitoring signals 17th it is determined which of the monitoring signals 17th with its signal value 23 has caused the deviation criterion 30th has been fulfilled.

That monitoring signal can then be selected which has ensured that the deviation criterion over a predetermined period of time, that is to say for one or more checking times 30th was fulfilled. In one step S16 the monitoring signal can thus be selected that has significantly made the greatest contribution over time to that in the threshold value comparisons 28 , 29 the threshold value was exceeded in each case or the deviation criterion in general 30th has been fulfilled. In one step S17 can then use the most probable monitoring signal recognized in this way 17th used to power the associated battery component 13th from which the monitoring signal 17th originates to be identified as a faulty battery component.

A particularly preferred exemplary embodiment is described below.

In the SmartCell concept (battery components with monitoring unit) voltage, temperature and current sensors are provided on each individual cell. In this way, these signals can be monitored for each individual cell. By considering the entirety of these measurement data, cells with unusual behavior (i.e. behavior that deviates from the majority of cells) should now be localized. Fault detection algorithms can thus identify cell defects that are not visible on the basis of the signals from a single cell measurement, but only by comparing the signals from all cells of the battery system. The number of detectable errors increases (e.g. increasing internal resistance, decreasing capacitance, internal short circuits). At the same time, these errors in the overall battery system can also be traced back to individual battery cells.

Core elements are the acquisition of current, temperature and voltage signals as well as other signals acquired at the cell level and the detection of unusual behavior by means of, for example, classification methods (Robust Outlier Detection, Artificial Neural Networks, Principal Component Analysis, etc.) and then the detection of cell defects on the basis of the classification result.

3rd shows the current curves of 12 cells in a 3s4p connection (3 serial connections of 4 cells connected in parallel) during driving. After a while, a battery cell shows an increased contact resistance ( 2 ). 3rd shows a section of the current profile in the presence of the increased contact resistance. The change could not be detected using conventional processes in battery management systems (eg threshold-based).

The comparison of the battery cells with each other (correlation-based approach) enables the change to be detected (cf. 5 below). In the 4th and 5 The illustrated signals and detection algorithms are only to be seen as examples of the idea. The invention can also be applied to other types of errors.

In addition to sensor values (e.g. voltage, current, temperature), parameters (e.g. internal resistance, capacitance, open-circuit voltage, cell time constants) and states (e.g. state of charge, polarization voltages or currents, state of aging) can also be used for the data-driven classification process.

The classification procedure is based in particular on Principle Component Analysis (PCA). However, this is only one possible procedure. In principle, other processes from the field of process monitoring can be used, including - Adjustments to the standard PCA: Dynamic PCA (DPCA), Kernel PCA (KPCA), Multiscale PCA (MSPCA), Multiblock PCA (MBPCA) - Independent Component Analysis (ICA) - Partial Least Squares (PLS) - Fisher Discriminant Analysis (FDA) - Subspace approach - Support Vector Machines (SVM) - Neural networks.

6th and 7th show again a somewhat more general description of the sequence of the method for detecting errors and for localizing errors.

Overall, the examples show how the invention can provide error detection and error localization by means of a sensor system.

QUOTES INCLUDED IN THE DESCRIPTION

This list of the documents listed by the applicant was generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Patent literature cited

• DE 112017000969 T5 [0004]
• DE 102010047960 A1 [0005]
• DE 102014220062 A1 [0006]
• DE 102015002827 A1 [0007]

Non-patent literature cited

• ražen Slišković, Ratko Grbić, Željko Hocenski, “Multivariate statistical process monitoring,” Tehnicki Vjesnik, vol. 19, 1 (2012), pp. 33-41, 2012 [0021]
• S. Yin, S. X. Ding, A. Haghani, H. Hao, and P. Zhang, "A comparison study of basic data-driven fault diagnosis and process monitoring methods on the benchmark Tennessee Eastman process," Journal of Process Control, vol. 22, no. 9, pp. 1567-1581, 2012 [0021]

Claims (10)

A method for detecting a fault in a battery system (11), wherein a plurality of battery components (13), each of which has at least one component-specific monitoring unit (14), and a computing device (15) are provided in the battery system (11), characterized in that the computing device (15) receives a respective monitoring signal (17) from the respective at least one component-specific monitoring unit (14) from each of the battery components (13), so that several monitoring signals (17) from different battery components (13) are received in the computing device (15) are present, and by means of a predetermined correlation-based classification method (18) on the basis of a respective time curve (22) of some or all of the monitoring signals (17) in each case at least one monitoring value of a predetermined statistical measure and for the at least one monitoring value on the basis of in the computing device (1 5) stored correlation data (C) an expectation corridor within which the respective monitoring value must lie in the absence of the error is specified and a classification result (32) is determined which indicates whether the respective monitoring value is determined according to a predetermined deviation criterion (30) Expectation corridor deviates, and if, according to the classification result (32), the deviation criterion (30) is met, an error signal (31) is generated which signals that a battery component (13) has an error. Procedure according to Claim 1 wherein the deviation criterion (30) comprises that the respective monitoring value deviates from the expected corridor at least for a predetermined minimum period of time and / or by more than a predetermined tolerance value. Method according to one of the preceding claims, wherein the classification method (18) comprises at least one of the following: an outlier detection, an artificial neural network, a principal component analysis, PCA; an independent component analysis, ICA. Method according to one of the preceding claims, wherein the at least one monitoring ^{value comprises a paired correlation coefficient of the monitoring signals and / or a T 2} value and / or squared prediction error value, SPE value formed from the monitoring signals, and the deviation criterion (30) respective threshold value comparison (28, 29) for a Student T distribution and / or T ² value and / or Squared Prediction Error, SPE, and / or a statistical hypothesis test. Method according to one of the preceding claims, wherein at predetermined checking times in each case - a respective current signal value (23) of the monitoring signals (17) is recorded and - the current signal values (23) are combined to form a signal vector (24) and - by means of PCA base vectors, which are predetermined by the correlation data (C) for a PCA, principal component analysis, the signal vector (24) is converted into principal components (25) and - for some or all of the principal components (25) a T ² value and / or an SPE Value is calculated as the respective monitoring value and a threshold value comparison (28, 29) is carried out for the respective value with a respective predetermined threshold value and the deviation criterion (30) is met if the respective threshold value is exceeded for a predetermined number of checking times. Method according to one of the preceding claims, wherein at least one of the following is signaled as the error (19): an increasing internal resistance, an increasing contacting resistance (R), a decreasing storage capacity, an internal short circuit in the faulty battery component; a cell defect; an insulation failure; a deviation from a predetermined standard behavior. Method according to one of the preceding claims, wherein for a localization of the error (19) it is determined by which of the monitoring signals (17) the deviation criterion (30) is met, and that battery component (13) whose monitoring unit (14) this monitoring signal (17 ) has generated when the faulty battery component (13) is identified. Method according to one of the preceding claims, wherein the at least one monitoring unit (14) of each of the battery components (13) generates its respective monitoring signal (17) with signal values (23) relating to at least one of the following operating variables: an electrical voltage, an electrical current, a temperature ; and or regarding parameter data, at least one of the following parameters: internal resistance, capacitance, open circuit voltage, cell time constants; and / or regarding status data at least one of the following statuses: State of charge, polarization voltage, polarization current, state of aging. Battery system (11) for a motor vehicle (10), with several in the battery system (11) Battery components (13), each of which has at least one component-specific monitoring unit (14) and a computing device (15) are provided, characterized in that the computing device (15) is set up to carry out a method according to one of the preceding claims. Motor vehicle (10) with a battery system (11) Claim 9 .

Images