DE102022213590B4 - Method and device for determining a predicted battery state for the purpose of anomaly prediction of a device battery in a technical device using edge computing - Google Patents

Abstract

The invention relates to a method for determining at least one predicted internal battery state of a specific device battery (41) of a technical device (4) and for detecting an anomaly of the specific device battery (41) based on the at least one predicted internal battery state.

Description

Technical area

The invention relates to electrical devices operated independently of the mains, in particular electrically driven motor vehicles, in particular electric vehicles or hybrid vehicles, with device batteries and also to measures for determining a predicted state of health (SOH) of the device battery. The invention relates to methods and devices for determining a predicted internal battery state of a device battery in a technical device by evaluating the behavior of the device batteries of a large number of similar device batteries. The invention also relates to methods for robustly determining a predicted internal battery state in the technical device for an anomaly prediction.

Technical background

The energy supply for electrical devices and machines that operate independently of the mains, such as electrically powered vehicles, is usually provided by device batteries or vehicle batteries. These supply electrical energy to operate the devices.

Device batteries degrade over their service life and depending on their load or use. This degradation is called cyclical ageing and leads to a continuously decreasing maximum power or storage capacity. The ageing state corresponds to a measure of the ageing of device batteries. According to the convention, a new device battery can have an ageing state (in terms of its capacity, SOH-C) of 100%, which decreases noticeably over the course of its service life. A measure of the ageing of the device battery (change in the ageing state over time) depends on an individual load on the device battery, i.e. in the case of vehicle batteries in motor vehicles, on the usage behavior of a driver, external environmental conditions and the vehicle battery type.

In order to monitor device batteries from a large number of devices, operating variable data is usually recorded continuously and transmitted in blocks as operating variable curves to a central unit external to the device. In the case of device batteries with a large number of battery cells, the operating variables can be recorded at cell level and transmitted to the central unit in compressed form. To evaluate the operating variable data, in particular to determine battery states in models based on differential equations, the operating variable data is sampled with a high temporal resolution (sampling rates) of, for example, between 1 and 100 Hz and an internal electrochemical battery state is determined from this using a time integration method.

Based on the internal battery conditions, the device battery can be monitored to detect anomalies. If the battery conditions can be predicted, a predictive detection of an impending anomaly can be performed.

The document EN 10 2020 215 297 A1 discloses a computer-implemented method for predicting a modeled aging state of an electrical energy storage device having at least one electrochemical unit.

The document EN 10 2020 212 282 A1 discloses a computer-implemented method for parameterizing an electrochemical battery model of a particular battery of a device.

Disclosure of the invention

According to the invention, a method and a device for internally determining at least one predicted internal battery state of a device battery of a technical device using edge computing and for detecting an anomaly based on the at least one predicted internal battery state according to claim 1 and a corresponding system according to the independent claim are provided.

Further embodiments are specified in the dependent claims.

• - Bereitstellen des mindestens einen inneren Batteriezustands der bestimmten Gerätebatterie durch Auswerten eines elektrochemischen Batteriemodells basierend auf Betriebsgrößenverläufen der bestimmten Gerätebatterie, wobei das elektrochemische Batteriemodell basierend auf Betriebsgrößenverläufen einer Vielzahl von Gerätebatterien parametrisiert und/oder trainiert ist;
• - Bereitstellen eines mithilfe des mindestens einen inneren Batteriezustand parametrisierten Batterie-Dynamik-Modells, wobei das Batterie-Dynamik-Modell zumindest einen Batteriestrom auf eine Batteriespannung abbildet;
• - Anpassen oder Korrigieren des mindestens einen inneren Batteriezustands mithilfe eines Beobachtermodells basierend auf einem Spannungsunterschied zwischen einer gemessenen Batteriespannung der bestimmten Gerätebatterie und einer mithilfe des Batterie-Dynamik-Modells modellierten Batteriespannung der bestimmten Gerätebatterie zu einem oder mehreren Zeitpunkten;
• - Prädizieren eines prädizierten Betriebsgrößenverlaufs mithilfe eines weiteren Batteriedynamikmodell, das mit dem korrigierten mindestens einen inneren Batteriezustand parametrisiert ist;
• - Ermitteln eines prädizierten Verlaufs des mindestens einen inneren Batteriezustands abhängig von dem prädizierten Betriebsgrößenverlauf mithilfe des elektrochemischen Batteriemodells, das die Zeitreihenintegration ausgehend von dem mindestens einen korrigierten inneren Batteriezustand (Z') zu einem aktuellen Zeitpunkt vornimmt.

According to a first aspect, a method for determining at least one predicted internal battery state of a specific device battery of a technical device and for detecting an anomaly of the specific device battery based on the at least one predicted internal battery state, comprising the following steps:

• - Providing the at least one internal battery state of the specific device battery by evaluating an electrochemical battery model based on operating variable profiles of the specific device battery, wherein the electrochemical battery model is parameterized and/or trained based on operating variable profiles of a plurality of device batteries;
• - Providing a battery dynamics model parameterized using the at least one internal battery state, wherein the battery dynamics model maps at least one battery current to a battery voltage;
• - adjusting or correcting the at least one internal battery state using an observer model based on a voltage difference between a measured battery voltage of the specific device battery and a battery voltage of the specific device battery modeled using the battery dynamics model at one or more points in time;
• - Predicting a predicted operating variable curve using a further battery dynamics model which is parameterized with the corrected at least one internal battery state;
• - Determining a predicted course of the at least one internal battery state depending on the predicted operating variable course using the electrochemical battery model, which carries out the time series integration starting from the at least one corrected internal battery state (Z') at a current point in time.

The state of health (SOH) is the key value for indicating the remaining battery capacity or remaining battery charge of device batteries. The state of health is a measure of the aging of the device battery and thus a measure for characterizing the residual value of the battery. In the case of a device battery or a battery module or a battery cell, the state of health can be specified as the capacity retention rate (SOH-C). The capacity retention rate SOH-C, i.e. the capacity-related state of health, is specified as the ratio of the measured instantaneous capacity to an initial capacity (at the time of commissioning) of the fully charged battery and decreases with increasing aging. Alternatively, the state of health can be specified as an increase in the internal resistance (SOH-R) with respect to an internal resistance at the start of the service life (time of commissioning) of the device battery. The relative change in the internal resistance SOH-R increases with increasing aging of the battery.

The aging state of a device battery with regard to cyclic aging is usually not measured directly. This would require a series of sensors inside the device battery, which would make the production of such a device battery costly and complex and would increase the space required. In addition, everyday measuring methods for directly determining the aging state in energy storage devices are not yet available on the market.

For capacity reasons, device batteries of a large number of devices are therefore monitored in a central unit external to the device. To do this, the devices transmit temporal operating variables of the device batteries' operating variables to the central unit, with an electrochemical battery model being executed in an aging model in the central unit in order to determine an aging state and current internal electrochemical states. Depending on the model used, time series of operating variables are continuously recorded as operating variable curves, such as battery current, battery temperature, state of charge and/or battery voltage, and transmitted to the central unit in blocks and, if necessary, in compressed form. The operating variable curves are evaluated there so that, based on one or more aging state models, a device-specific state, specified by internal electrochemical battery states and, if necessary, other variables, can be calculated/determined.

In addition, the operating variables from the large number of device batteries can be evaluated using statistical methods in order to parameterize another electrochemical performance battery model, e.g. by fitting. The model parameters also correspond to internal electrochemical battery states, which correspond to physical states. The internal electrochemical battery states that can be determined by the aging model and the performance model correspond to one another or can be converted into one another.

• - eine Menge an zyklisierbarem Lithium,
• - einen Volumenanteil der Anode,
• - einen Volumenanteil der Kathode,
• - eine Reaktionsrate oder einen Reaktionskoeffizient bzgl. einer Anode,
• - ein Diffusionskoeffizient der Anode,
• - eine Schichtdicke bzgl. der Anode,
• - eine Reaktionsrate oder einen Reaktionskoeffizienten bzgl. der Kathode,
• - einen Diffusionskoeffizient der Kathode,
• - eine Porosität in der Anode,
• - eine Porosität in der Kathode,
• - ein Bruggemann-Koeffizient in der Anode,
• - ein Bruggemann-Koeffizient in der Kathode,
• - eine Elektrolytkonzentration,
• - einen Kontaktwiderstand,
• - eine mechanische Partikelbelastung.

The electrochemical battery states may include one or more of the following electrochemical state variables as internal battery states:

• - an amount of cyclable lithium,
• - a volume fraction of the anode,
• - a volume fraction of the cathode,
• - a reaction rate or reaction coefficient with respect to an anode,
• - a diffusion coefficient of the anode,
• - a layer thickness regarding the anode,
• - a reaction rate or reaction coefficient with respect to the cathode,
• - a diffusion coefficient of the cathode,
• - a porosity in the anode,
• - a porosity in the cathode,
• - a Bruggemann coefficient in the anode,
• - a Bruggemann coefficient in the cathode,
• - an electrolyte concentration,
• - a contact resistance,
• - mechanical particle loading.

Additional parameters for electrochemical, mechanical or physical modeling may also be included.

The electrochemical battery model, e.g. in the form of an aging model, comprises a differential equation system which, based on differential equations parameterized via model parameters, models internal battery states, in particular equilibrium states and possibly kinetic states, using a time integration method and provides a relationship between operating variables of the battery cells of the device battery, namely a battery current, a battery voltage, a battery temperature and a state of charge of the device battery. Such electrochemical battery models are known, for example, from the publications US2016/023,566 , US2016/023,567 and US2020/150,185 known. The aging state of the respective device battery can be determined approximately as a linear combination of the internal electrochemical states. The electrochemical battery model can be evaluated in the central unit or in a control unit for an individual device battery in order to determine the internal battery states.

A possible electrochemical battery model can be provided, for example, in the form of a hybrid aging state model, which corresponds to a combination of a physical aging model with a data-based model. In the hybrid aging state model, a physical aging state can be determined using a physical or electrochemical aging model and this can be subjected to a correction value that results from a data-based correction model, in particular by addition or multiplication.

The physical aging model is based on electrochemical model equations, which characterize electrochemical states of a non-linear differential equation system in terms of aging reactions in a known manner, continuously calculate them depending on the temporal operating variable profiles according to a time integration method and map them to the physical aging state for output, as SOH-C and/or as SOH-R. The calculations can typically be carried out in the central unit (cloud) at intervals of predetermined evaluation periods, e.g. once a week.

Furthermore, the correction model of the hybrid data-based aging state model can be designed with a probabilistic or artificial intelligence-based probabilistic regression model, in particular a Gaussian process model, and can be trained to correct the aging state obtained by the physical aging model. For this purpose, there is therefore a data-based correction model of the aging state for correcting the SOH-C and/or at least one more for correcting the SOH-R. Possible alternatives to the Gaussian process are other supervised learning methods, such as those based on a random forest model, an AdaBoost model, a support vector machine or a Bayesian neural network.

It can be provided that the aging state model is provided in a central unit remote from the device, which comprises the physical aging model which is based on a differential equation system and is designed to calculate an aging state by means of a time integration method, wherein the physical aging model is fitted based on operating variable profiles of a plurality of device batteries of the same type in order to assign an aging state to a temporal operating variable profile of the specific device battery.

To monitor the device battery, the internal states of the physical aging model (differential equation system) of the hybrid aging state model in the central unit can be further developed based on the operating variable curves, e.g. starting from a commissioning point (start of service life)/reference point in time using the time integration method, from which an aging state can be determined based on the model. This requires that operating variable curves are continuous. The physical aging model is parameterized based on highly precise aging states SOH at regular points in time.

As a further model for determining a highly accurate aging state, a basic model can be provided, according to which a SOH-C measurement is carried out by Coulomb counting or by forming a temporal current integral during the charging process, which is divided by the change in the state of charge between the beginning and the end of the relevant charging and/or discharging phase. The calibration is advantageously carried out on the no-load voltage characteristic in rest phases in order to also calculate the state of charge curve in the central unit. A sufficiently reliable indication of the state of age can be obtained, for example, if the vehicle battery is charged from a defined relaxed state under reproducible load and ambient conditions of a fully discharged charging state is brought to a fully charged state.

The maximum charge recorded in this way can be related to an initial maximum charging capacity of the vehicle battery. Resistance-related aging states (SOH-R values) can also be calculated using voltage changes related to a current change. These are usually related to a defined time interval as well as defined environmental conditions and the energy flow direction of the system. Aging states determined in this way cannot be determined at any time and are usually used as labels for fitting the physical aging model and for training the correction model.

It can be provided that the calculation of at least one electrochemical battery state is carried out with the help of an observer, which can include, for example, a Luenberger observer, a Kalman filter or a particle filter. Observations of a known battery performance model, preferably executed in the central unit, are used with the aid of fitting the model equations. Kinetic model parameters and equilibrium parameters are determined by a fitting process based on operating variable profiles evaluated over time in the dynamic or static case. These model parameters correspond to physical battery states of the device battery and can be fused with the states of the physical aging model using suitable data fusion methods.

The fitting process can be triggered in the central unit, e.g. when a sufficiently long rest phase is detected (fitting of equilibrium parameters) or a sufficiently dynamic driving phase is detected (fitting of kinetic parameters).

The hybrid aging state model can now be parameterized or trained using the operating variables of all device batteries connected to the central unit. In particular, the physical aging model is fitted to the operating variables of the large number of device batteries at regular points in time, such as every three months. This gives model parameters of the electrochemical/physical aging model that specify equilibrium parameters and/or kinetic parameters. The correction model can then be trained on the residue of the electrochemical aging model based on the large number of operating variables of the large number of device batteries.

It can be intended that individual device batteries or compatible devices, such as charging stations or charging cables, are used for data acquisition or for data preprocessing or data aggregation. Advantageously, the model parameterization of the electrochemical battery model/physical aging model and the battery performance model with several labels that were recorded or derived on the basis of decentrally recorded measurements takes place in a central unit.

The at least one internal battery state can be determined by evaluating an operating variable profile of a specific device battery as a latent state in the differential equation system and, if necessary, merged with model parameters of a battery performance model. The at least one internal battery state is transmitted to the device with the specific device battery and used there to parameterize a battery dynamics model.

The battery dynamics model enables a simulation of the electrical behavior of the device battery, such as EP3435102A1 known, for example, to determine a battery voltage from a battery current. The at least one internal battery state determined on the basis of a large number of device batteries and the operating variable profile of the specific device battery is further evaluated in the device.

Based on the battery current and battery temperature curves measured in the respective technical device, a corresponding curve of the battery voltage and the state of charge of the device battery (through charge accumulation) can then be modeled, so that a modeled battery voltage and a modeled state of charge are always available at a given point in time.

Due to manufacturing tolerances and other influences, the behavior of the device batteries differs from one another, so that a correction of at least one internal battery state can be made using an observer model. The observer model can be used to compare the current measured battery voltage with the modeled current battery voltage and to adjust the at least one internal battery state, which is used as a model parameter in the battery dynamics model, depending on the voltage difference. The result is a corrected/adjusted at least one internal battery state that accurately describes the behavior of the specific device battery.

The observer model can be implemented, for example, as a Luenberger observer or using an extended Kalman filter or a particle filter and corrects the at least one internal battery state depending on the voltage difference between the modeled battery voltage and the measured battery voltage. The observer model is designed in such a way that, based on several voltage differences at different times, it explains the observation errors of the voltage differences in the most plausible way by correcting/adjusting the at least one internal battery state. This enables the battery-specific description of production variations, series variations and sudden events, such as a previous accident involving the technical device.

Furthermore, a fault of the device battery can be determined based on the predicted course of the at least one internal battery state, wherein in particular the fault of the device battery is determined based on a threshold value comparison of a lower limit and/or upper limit of a confidence range of the at least one internal battery state at a future point in time with a predetermined threshold value.

Furthermore, the confidence interval of the at least one corrected internal battery state can be determined by evaluating the operating variable curves of the plurality of device batteries for the corresponding at least one internal battery state, in particular according to the law of large numbers.

The corrected/adjusted at least one internal battery state can be evaluated in terms of its confidence. The confidence σ can be assumed to be constant or fixed based on test or validation results, which evaluate the scatter or accuracy of the kinetic or equilibrium parameters for different operating variables. Furthermore, based on the "number of fits", the law of large numbers (root n) can be applied to refine the effectively resulting uncertainty, characterized by the standard deviation σ _res , or to reduce the resulting uncertainty if the number of observations n is sufficiently high: ${\sigma }_{res}=\sigma /\sqrt{n}$

The evaluation can, for example, be gradient-based and evaluate a quantile of a gradient in a predetermined period of time, such as within the last 3 hours.

The labels for stochastic evaluation (e.g. SEl thickness) can come from laboratory measurements in which several battery cells with different aging histories, such as cyclic aging and calendar aging, and with different load histories, which can include e.g. a temperature history, a discharge stroke history, a current throughput, a charging power, a discharging power and other stress factors, are evaluated and measured.

Alternatively or additionally, a parallel probabilistic model can be evaluated, which can provide an uncertainty or a confidence. Optionally, an additive correction of the probabilistic model can also take place here, e.g. implemented as a Gaussian process.

The adjusted/corrected at least one internal battery state at a current point in time is now used as a starting point to determine a predicted course of the at least one internal battery state using the time series integration of the predicted operating variable course and the electrochemical battery model.

For this purpose, the predicted operating size trends are determined using a usage pattern model that derives usage behavior from the historical operating size trends or parts thereof. The usage pattern model determines realistic artificial operating size trends that cause cyclical aging of the device battery that essentially corresponds to the cyclical aging caused by previous use.

Subsequently, anomaly detection can be performed based on the predicted internal battery states and their confidences.

Anomaly detection can be rule-based or data-based to evaluate at least one internal battery state and its confidence at a specific time within the prediction horizon to detect a fault or proper operation of the vehicle battery.

According to a further aspect, an apparatus for carrying out one of the above methods is provided.

Short description of the drawings

• 1 eine schematische Darstellung eines Systems zum Bereitstellen von fahrer- und fahrzeugindividuellen Betriebsgrößen zur Bestimmung eines Alterungszustands einer Fahrzeugbatterie;
• 2 eine schematische Darstellung eines funktionalen Aufbaus des verteilten Systems der Zentraleinheit und eines Steuergeräts in dem technischen Gerät; und
• 3 eine Darstellung eines Flussdiagramms zur Veranschaulichung eines Verfahrens zur Ermittlung eines prädizierten Verlaufs eines Alterungszustands;
• 4 eine Darstellung eines prädizierten Verlaufs eines inneren Batteriezustands basierend auf dem mit korrigierten inneren Batteriezustand parametrisierten Batterie-Dynamik-Modell und den prädizierten Betriebsgrößenverläufen.

Embodiments are explained in more detail below with reference to the attached drawings. They show:

• 1 a schematic representation of a system for providing driver and vehicle-specific operating variables for determining an ageing state of a vehicle battery;
• 2 a schematic representation of a functional structure of the distributed system of the central unit and a control device in the technical device; and
• 3 a representation of a flow chart to illustrate a method for determining a predicted course of an aging state;
• 4 a representation of a predicted course of an internal battery state based on the battery dynamics model parameterized with the corrected internal battery state and the predicted operating variable curves.

Description of embodiments

In the following, the method according to the invention is described using vehicle batteries as device batteries in a large number of motor vehicles as similar devices. For this purpose, an aging state model is operated in the central unit and used to calculate an aging state. In the central unit, the aging state model is continuously updated based on operating variables of the vehicle batteries from the vehicle fleet or retrained on the basis of reference measurements in order to obtain internal battery states related to the respective aging state as a latent state of the differential equation system. These each represent the average internal battery state across the entire plurality of vehicle batteries for the specific aging state. Furthermore, the internal battery states that were determined based on the currently trained aging state model are communicated to the vehicles and used there to parameterize a battery dynamics model. In the vehicles, the model parameters of the battery dynamics model and thus the internal battery states are adapted to the actual individual battery state of the vehicle battery in question based on an observer. The adapted internal battery states and the adapted battery dynamics model are used to predict the aging state.

The above example is representative of a large number of stationary or mobile devices with a grid-independent energy supply, such as vehicles (electric vehicles, pedelecs, etc.), systems, machine tools, household appliances, IOT devices and the like, which are connected to a central unit external to the device (cloud) via a corresponding communication connection (e.g. LAN, Internet). Furthermore, it is also possible for the device to be in communication with the device battery and to take over the connection between the device battery and the central unit, such as a charging device or a charging cable.

1 shows a system 1 for collecting fleet data in a central unit 2 for creating and operating as well as evaluating an aging state model. The aging state model is used to determine an aging state of battery cells of the vehicle battery in a motor vehicle based on temporal operating variable profiles. 1 shows a vehicle fleet 3 with several motor vehicles 4.

One of the motor vehicles 4 is in 1 shown in more detail. The motor vehicles 4 each have a vehicle battery 41, 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 a central unit 2 (a so-called cloud). The vehicle battery 43 has a plurality of battery cells 45.

The control unit 43 is particularly designed to record operating variables with a high temporal resolution, such as between 1 and 50 Hz, such as 10 Hz, and to transmit them to the central unit 2 via the communication device 44.

The motor vehicles 4 send the operating variables F to the central unit 2, which at least indicate variables that influence the aging state of the vehicle battery 41 and are required to determine the internal states of the battery cells 45. In the case of a vehicle battery, the operating variables F can indicate a current battery current, a current battery voltage, a current battery temperature and a current state of charge (SOC: State of Charge), both at pack, module and/or cell level. The operating variables can be regularly transmitted to the central unit 2 in uncompressed and/or compressed form. For example, the time series can be transmitted to the central unit 2 in blocks at intervals of 10 minutes to several hours using compression algorithms in order to minimize data traffic to the central unit 2.

The central unit 2 has a data processing unit 21 in which the method described below can be carried out and a database 22 for storing data points, model parameters, states and the like.

In the central unit 2, an electrochemical battery model, such as an aging model, is implemented, which is a differential equation system and is based on time series integration of operating variables in order to determine current internal battery states when evaluating an operating variable. Such electrochemical battery models are, for example, state of the art in the publications US2016/023,566 , US2016/023,567 and US2020/150,185 The electrochemical battery model can also be fitted to the operating variables as a battery performance model in order to determine internal operating states as model parameters.

The aging state model can be parameterized by reliable, high-precision measurements of the aging state of individual vehicle batteries 41 in the fleet 3, so that an evaluation with operating variable curves results in internal battery states as internal states of the corresponding differential equation system.

The operating variable data thus enable modeling of one or more current internal battery states through model evaluation.

The aging state model can be partially data-based as a hybrid model. The aging state model is suitable as an electrochemical battery model and can be used regularly, i.e. e.g. after the respective evaluation period has elapsed, to determine the current aging states of the relevant vehicle battery 41 of the assigned vehicle fleet 3 based on the temporal progression of the individual operating variables (since commissioning or a point in time of a defined state of the respective vehicle battery) and the operating characteristics determined from this. In other words, it is possible to determine an aging state of the relevant vehicle battery 41 based on the progression of the operating variables of one of the vehicle batteries 41 of the motor vehicles 4 of the assigned vehicle fleet 3 and the operating characteristics resulting from these progressions of the operating variables. At the same time, the evaluation of the differential equation system results in internal battery states that are assigned to the individual vehicle battery.

The state of health (SOH) is the key value for indicating the remaining battery capacity or remaining battery charge. The state of health is a measure of the aging of the vehicle battery or a battery module or a battery cell and can be specified as the capacity retention rate (SOH-C) or as the increase in internal resistance (SOH-R). The capacity retention rate SOH-C is specified as the ratio of the measured instantaneous capacity to an initial capacity of the fully charged battery. The relative change in internal resistance SOH-R increases as the battery ages.

2 shows, by way of example, a schematic functional structure of an embodiment of a function carried out by the central unit 2 and a control unit 43 in a vehicle 4 with a data-based aging state model 9 that is constructed in a hybrid manner. The aging state model 9 comprises a physical aging model 5 and a data-based correction model 6.

The physical aging model 5 is a non-linear, mathematical model based on differential equations. Evaluating the physical aging model of the aging state model with operating variable profiles, in particular since the beginning of the service life of the vehicle battery 41, results in an internal state of the system of physical differential equations being established that corresponds to an electrochemical internal battery state of the vehicle battery 41. Since the physical aging model is based on physical and electrochemical laws, the model parameters of the physical aging model are variables that indicate physical properties.

The time series of the operating variables F of the vehicle battery 41 (when considering the entire battery) are therefore directly included in the physical aging state model 5, which is preferably designed as an electrochemical model and models corresponding internal electrochemical internal battery states, such as layer thicknesses (e.g. SEI thickness), changes in the cyclable lithium due to anode/cathode side reactions, rapid consumption of electrolytes, slow consumption of electrolytes, loss of the active material in the anode, loss of the active material in the cathode, etc...., using nonlinear differential equations and a multidimensional state vector.

The physical aging model 5 thus corresponds to an electrochemical battery model of the vehicle batteries and the cell chemistry. The aging model 5 determines a physically based aging state SOHph of dimension of at least one depending on the temporal operating variable profiles F and additionally as latent internal states internal battery states, which in particular can be mapped linearly or non-linearly to a capacity maintenance rate (SOH-C) and/or an internal resistance increase rate (SOH-R) in order to provide these as an aging state (SOH-C and SOH-R).

However, the model values for the physical ageing state SOHph provided by the physical ageing model 5 are inaccurate in certain situations and it may therefore be provided to correct them with a correction value k. The correction value k is provided by the data-based correction model 6, which is trained using training data sets from the vehicles 4 of the vehicle fleet 3 and/or using laboratory data.

The correction model 6 receives operating characteristics M on the input side, which are determined from the courses of the operating variables F and can also include one or more of the internal electrochemical states of the differential equation system of the physical model. Furthermore, the correction model 6 can receive the physical aging state SOHph obtained from the physical aging model 5 on the input side. The operating characteristics M of the current evaluation period are generated in a feature extraction block 8 based on the time series of the operating variables F. The operating characteristics M also include the internal states from the state vector of the electrochemical physical aging model and advantageously the physical aging state SOHph.

From the operating variables F, operating characteristics M can be generated in the central unit 2 for each vehicle fleet 3 or, in other embodiments, in the respective motor vehicles 4, which relate to an evaluation period. The evaluation period for determining the state of aging can be a few hours (e.g. 6 hours) to several weeks (e.g. one month). A typical value for the evaluation period is one week.

The operating characteristics M (for individual battery consideration or overall battery consideration) can, for example, include characteristics related to the evaluation period and/or accumulated characteristics and/or statistical variables determined over the entire service life to date. In particular, the operating characteristics can, for example, include: electrochemical states, such as. B. SEI layer thickness, change in cyclable lithium due to anode/cathode side reactions, rapid uptake of electrolyte solvents, slow uptake of electrolyte solvents, lithium deposition, loss of active anode material and loss of active cathode material, information on impedances or internal resistances, histogram features such as temperature versus state of charge, charge current versus temperature and discharge current versus temperature, in particular multi-dimensional histogram data regarding battery temperature distribution versus state of charge, charge current distribution versus temperature and/or discharge current distribution versus temperature, current throughput in ampere hours, total accumulated charge (Ah), an average capacity increase during a charge process (in particular for charges where the charge increase is above a threshold proportion [e.g. 20% ΔSOC] of the total battery capacity), charge capacity and an extreme value (e.g. B. Maximum) of the differential capacity during a measured charging process with a sufficiently large change in the state of charge (smoothed curve of dQ/dU: change in charge divided by change in battery voltage) or the accumulated driving performance. These quantities are preferably converted in such a way that they characterize the real usage behavior as best as possible and normalize it in the feature space. The operating characteristics M can be used in whole or in part for the procedure described below.

To determine a corrected aging state SOH to be output, the outputs SOHph, k of the physical aging model 5 and the data-based correction model 6, which is preferably designed as a Gaussian process model, are applied to one another. In particular, these can be added in a summation block 7 or otherwise multiplied (not shown) in order to obtain the modeled aging state SOH to be output for a current evaluation period. In the addition case, the confidence of the Gaussian process can also be used as the confidence of the corrected aging value SOH of the hybrid model to be output. The confidence or the confidence value of the Gaussian process model thus characterizes the modeling uncertainty of the mapping of operating characteristic points to an aging state.

The hybrid aging state model is trained in the central unit 2 using a training block 10. For this purpose, training data sets are defined that assign operating size trends to an empirically or model-based determined aging state as a label.

For example, to determine an ageing state, a base model can be provided as a label for training the hybrid or data-based ageing state model, according to which a SOH-C measurement is carried out by Coulomb counting or by forming a temporal current integral during the charging process, which is divided by the change in the state of charge between the beginning and the end of the relevant charging and/or discharging phase. A sufficiently reliable indication of the ageing state can For example, this can be obtained when the vehicle battery is brought from a fully discharged charging state to a fully charged state during a charging process under reproducible load and ambient conditions. The maximum charge recorded in this way can be related to an initial maximum charging capacity of the vehicle battery. Resistance-related aging states (SOH-R values) can also be calculated using voltage changes related to a current change. These are usually related to a defined time interval as well as defined ambient conditions and the energy flow direction of the system.

The determination of an aging state as a label can be carried out in a manner known per se by evaluating the operating variable curves with an additional aging model in the vehicle or in the central unit 2 under defined load and ambient conditions of a label generation, such as in a workshop, on a test bench or a diagnostic or label generation mode, which represents an operating mode and guarantees compliance with predetermined operating conditions of the vehicle battery, such as constant temperature, constant current and the like. Other models can be used to determine the aging state. The training of the data-based correction model can be carried out in the conventional way based on the training data sets.

The physical aging model 5 can be fitted to the operating variable curves in rest phases under the control of the training block 10, e.g. using a least square method or similar. The physical aging model 5 can model equilibrium states and be described by model parameters. The model parameters can be re-parameterized at regular intervals, particularly when operating variable curves are available for a defined period of time of at least several (e.g. three) hours. This data can be collected for similar vehicle batteries 41 by evaluation in the central unit 2 and the adjustment or re-parameterization can be carried out there.

Such model parameters of the physical aging model 5 can, for example, include a scalar indication of the cyclable lithium (value relates to the cathode capacity), the proportion of cyclable lithium at the beginning of the battery's service life (scalar value), a volume proportion of the anode and a volume proportion of the anode at the beginning of the service life of the vehicle battery 41. The aging state SOH can be approximately determined as a linear combination of the internal equilibrium states.

The evaluation of operating variable curves of a specific vehicle battery with the hybrid aging state model 9, which is based on fleet data of operating variable curves of a large number of vehicle batteries of different calendar ages and different aging states, now leads to individual internal battery states Z, which result from the evaluation of the differential equation system.

The internal battery states Z can be transmitted directly to the vehicle in question or, alternatively, the internal battery states can be determined in a second way from the operating variables using another battery model, such as a battery performance model, and these can then be merged or combined with the previously obtained internal battery states using suitable data fusion methods, e.g. by weighted averaging and the like. The internal battery states determined in this way are determined on the basis of a large number of vehicle batteries and represent the average battery state across the entire vehicle fleet for a specific calendar age.

The respective internal battery states Z determined with the respective operating variable profiles in the hybrid aging state model are transmitted to the connected vehicles 4 of the vehicle fleet 3 at predetermined times, in particular after each update or each retraining of the aging state model. There, these are taken into account in a battery dynamics model 12 implemented there, which can be designed as a fractional model, in order to determine a profile of the state of charge and the battery voltage U _mod from the individual operating variable profiles I _mess (t), T _bat (t) of the battery current I _mess and the battery temperature T _bat measured in the vehicle. The battery dynamics model 12 is designed in a manner known per se to simulate the electrical behavior of the vehicle battery 41 based on the internal battery states Z. In particular, the battery dynamics model 12 is designed to determine a battery voltage U _mod from a battery current I _mess , a battery temperature T _Bat and a state of charge SOC. The state of charge is usually continuously monitored in a battery management system and determined by temporal charge integration of the course of the flowing battery current.

Furthermore, an observer model 13, e.g. a Luenberger observer, is provided, which for one or more points in time the modelled Battery voltage U _mod is compared with a currently measured battery voltage U _mess and a voltage difference is evaluated such that the obtained internal battery states Z are adapted to adapted internal battery states Z'. The observer model 13 therefore adapts the internal battery states Z depending on the observation of the voltage difference. This makes it possible to determine battery-specific deviations of the internal battery states Z from those based on the aging state model 9 parameterized/trained across all fleet data, so that deviations in battery behavior from the average design type of vehicle batteries 41 due to manufacturing tolerances and series variations are taken into account. The result is adapted internal battery states Z'.

These can be displayed to a driver accordingly and take into account both the fleet-based observations from the multitude of similar vehicle batteries 41 and the individual characteristics of the vehicle battery 41 of the vehicle in question.

The adjusted internal battery states Z' determined in this way can now be used for evaluation in a further battery dynamics model 14. The further battery dynamics model 14 and subsequent evaluations can be carried out either in the vehicle 4 or after the determined adjusted internal battery states Z' have been transmitted back to the central unit 2. The further battery dynamics model 14 can, as previously described, convert battery current and battery temperature curves of the state of charge into corresponding battery voltage and state of charge curves, with the further battery dynamics model 14 being parameterized based on the determined adjusted internal battery states Z'.

The further battery dynamics model 14 is used to convert predicted load size curves, namely curves of the battery current I _p and the battery temperature T _p, into corresponding curves of the predicted battery voltage U _p and the predicted state of charge SOC _P.

The predicted load size curves can be artificially generated in the vehicle 4 or in the central unit 2 based on individual usage characteristics using a usage pattern model 15. For this purpose, for example, a driver's usage behavior or driving style is analyzed and realistic load size curves, i.e. curves of the battery current and the battery temperature, are derived from this for a future period. This can be done, for example, by stringing together predetermined driving cycles, rest cycles and charging cycles, each with a predetermined time duration and with certain frequencies, depending on a usage behavior defined by usage-related variables (load profile), so that load size curves assigned to the cycles result from their stringing together. The resulting overall load size curve then corresponds to a realistic curve of the battery current and the battery temperature that is realized by the specific user or driver starting from the current point in time. In this way, an overall load size curve is preferably calculated up to a predetermined prediction horizon, i.e. a future point in time or a period of time in the future.

For example, the usage pattern model 15 can be designed based on data and, based on a characterization of the usage behavior of the user or the driving behavior or driving style of the driver, provide a sequence of predefined driving cycle profiles, rest cycle profiles and charging cycle profiles (each of which is assigned a time section of a load size curve), from which an entire load size curve can be determined, as described above. The data-based usage pattern model 15 can thus be trained to assign a sequence of driving cycle profiles, rest cycle profiles and charging cycle profiles depending on usage-related variables, such as an average ampere hour (Ah) throughput, an average discharge current, an average charging current and the like.

In a further hybrid aging state model 16, which can be implemented in the vehicle 4 or in the central unit 2, the predicted load size curve I _P , T _p and the predicted curves of the battery voltage U _p and the state of charge SOP determined therefrom can now be evaluated in a manner known per se by evaluating the physical aging model 5 and the data-based correction model 8 as described above up to the specified prediction horizon in order to obtain a predicted aging state SOH or a curve of a predicted aging state SOH . At the same time, a temporal evolution of the internal battery states is obtained, which result as latent states when evaluating the aging state model 9.

For example, in 4 the temporal development of an internal battery state is shown using the example of the SEI layer thickness of a specific vehicle battery. In a past time period A, the course of the SEI layer thickness can be seen, and in a predicted time period B, the course of the SEI layer thickness along with the corresponding end confidence interval. The corrected/adjusted at least one internal battery state can be evaluated in terms of its confidence. The confidence σ can be assumed to be constant or fixed, for example, based on test or validation results, which evaluate the scatter or the accuracy of the kinetic or equilibrium parameters that correspond to the internal battery states for different operating variables of different vehicle batteries 41. Furthermore, based on the "number of fits", the law of large numbers (root n) law can be applied to refine the effectively resulting uncertainty, characterized by the standard deviation σ _res , or to reduce the resulting uncertainty with a sufficiently high number of observations n: ${\sigma }_{res}=\sigma /\sqrt{n}$

The evaluation can, for example, be gradient-based and evaluate a quantile of a gradient in a predetermined period of time, such as within the last 3 hours.

The labels for stochastic evaluation (e.g. SEI thickness) can come from laboratory measurements in which several battery cells with different aging histories, such as cyclic mounting and calendar aging, and with different load histories, which can include, for example, a temperature history, a depth of discharge history, a current throughput, a charging power, a discharging power and other stress factors, are evaluated and measured.

Alternatively or additionally, a parallel probabilistic model can be evaluated, which can provide an uncertainty or a confidence. Optionally, an additive correction of the probabilistic model can also take place here, e.g. implemented as a Gaussian process.

Based on the predicted courses of the internal battery states determined in this way, a data-based or rule-based anomaly detection model 17 can be operated. The anomaly detection model 17 evaluates at one or more points in time at which predicted internal battery states, possibly together with their confidences, are available in order to detect the presence of an anomaly. The respective confidence of the respective predicted internal battery state can be taken into account. For example, with a rule-based anomaly detection model 17, an anomaly can be detected if a lower confidence limit of one or more predicted internal battery states to be evaluated is above a predetermined threshold value or if an upper confidence limit of one or more predicted internal battery states to be evaluated is below a predetermined threshold value.

In 3 is a flow chart illustrating a method that is carried out in the central unit 2 and the vehicles 4.

In step S1, operating variables F of the respective vehicle batteries 41 are first recorded in each of the vehicles, including the battery temperature, the battery current, the battery voltage and the state of charge. The operating variable profiles are recorded with a high temporal resolution of, for example, 10 Hz and provided in the central unit 2. The operating variables correspond to the battery current, the battery temperature, the battery voltage and, if applicable, the state of charge. Data processing is carried out, including outlier detection, filtering and data cleansing, if applicable, interpolation or the like.

In step S2, a hybrid aging state model is trained in the central unit 2 based on the collected operating variable profiles of the plurality of vehicle batteries 41. The aging state model is trained, as previously described, by fitting the electrochemical aging model 5 and by training the correction model 8 on the resulting residue. The hybrid aging state model 9 can be fitted and trained at predetermined regular time intervals of, for example, between 3 and 6 months, based on operating variable profiles labeled with reliably determined aging states.

The labels for the training process can be derived with high precision from an evaluation of certain processes, such as a charging process under predetermined conditions. For example, a basic model can be provided to determine an aging state, according to which a capacity-related aging state is determined by Coulomb counting or by forming a temporal current integral during the charging process, which is divided by the change in the state of charge between the start and end of the relevant charging and/or discharging phase.

When the hybrid aging state model 9 is evaluated with operating variable curves for an individual vehicle, internal battery states arise as latent states in the differential equation system, which can be interpreted as physical states. These internal battery states Z are then transmitted to the relevant individual vehicle 4 in step S3. This can be done, for example, after each reparameterization of the aging state model 9 for each of the vehicles 4 in the vehicle fleet 3.

In step S4, the internal battery states Z determined in this way are used in the vehicle to parameterize the battery dynamic model 12. The battery dynamic model is designed to determine a battery voltage U _mod and a state of charge SOC based on load variables, namely the battery current M _iess and the battery temperature T _Bat . The internal battery states Z can include, for example, an amount of available cyclable lithium, a volume fraction of the cathode, a volume fraction of the anode and the like.

In step S5, the battery dynamics model 12 is executed and, using the measured battery temperature T _Bat and the measured battery current I _{mess ,} a corresponding voltage response is calculated as a modeled battery voltage U _mod .

In step S6, the modeled battery voltage U _mod is compared with a battery voltage U _mess measured in the vehicle battery 41 and a corresponding voltage difference is determined as a voltage residue.

In the following step S7, the observer model 13 is applied, which uses the determined voltage difference, and corrects the internal battery states Z in order to explain the observation error in the most plausible way. The observer model 13 can be designed as a Luenberger observer, which determines corresponding changes in the internal battery states Z determined in the central unit 2 from several time-staggered queries. This results in a corresponding adaptation of the internal battery states Z to the individual vehicle battery 41.

This step can be embedded or calculated in the vehicle 4. This step can ensure that the local model now describes boundary conditions and influences that cannot be described via the central aging state model 9 in the central unit 2, e.g. due to battery-specific variables such as series variation and manufacturing tolerances, which can only be determined locally (e.g. due to the voltage residue).

The adjusted internal battery states Z' can now be used for predictive anomaly detection. The predictive anomaly detection can be carried out in the vehicle 4 or after appropriate feedback in the central unit 2.

In step S8, the adapted internal battery states Z' can be used to parameterize the further battery dynamics model 14. This serves to predict operating variable curves based on predicted load variable curves.

For this purpose, in step S9, load size curves of the battery current I _p and the battery temperature T _p are predicted as described above up to a predetermined prediction horizon (point in time in the future) using a usage pattern model 15. The load size curves are modeled into curves of the predicted battery voltage U _p and the predicted state of charge SOC _p using the additional battery dynamics model 14.

The load variable curves and the curves of the battery voltage U _p and the state of charge SOC _p determined therefrom are now evaluated in step S10 as predicted operating variable curves in a further hybrid aging state model 16 in order to obtain a curve of a predicted aging state SOH _p and at the same time the predicted internal battery states as latent states. The further hybrid aging state model 16 can correspond to the aging state model 9 that has been parameterized in the central unit 2 based on the large number of vehicles. The aging state model 9 is evaluated based on a latent state of the differential equation system at a current point in time, which results from the adjusted internal battery states Z' of the individual vehicle 4.

Together with the predicted aging state, predicted courses of the internal battery states result. Thus, by evaluating the further hybrid aging state model 16, a course of a battery state, such as an SEI layer thickness, can be predicted, as is the case, for example, in 4 is shown.

In 4 one can see the battery condition determined by evaluating the real operating variables (black) and the battery condition determined using the predicted operating variables and the battery condition determined from the execution of the further hybrid aging state model 16. Due to the use of a probabilistic data-based model as a correction model of the further hybrid aging state model 16, a confidence value can also be determined for the respective battery condition, which is 4 represented by the hatched area around the predicted course of the battery state.

Furthermore, the prediction by the purely electrochemical aging model can optionally be corrected by the data-driven, probabilistic correction model, which is advantageously implemented as a Gaussian process regressor. It is also conceivable that only the physical prediction is used as a point estimator and the probabilistic model acts only as a provider of uncertainty quantification (in this case without correction of the point estimator).

By executing the further hybrid aging state model 16, internal battery states are obtained for the period of the assumed prediction horizon. These internal battery states can be evaluated in step S11 with regard to their confidence, i.e. their prediction uncertainty, using the data-based correction model, which can be designed as a probabilistic model as part of the hybrid aging state model. In particular, anomaly detection can be carried out based on the predicted internal battery states and their confidences.

For this purpose, the anomaly detection model 17 is used, which can be carried out centrally in the vehicle 4 or by transmitting the predicted internal battery states and their competencies to the central unit 2. The anomaly detection model can be rule-based or data-based in order to evaluate the battery states and their confidences at a specific point in time within the prediction horizon in order to detect a fault or proper operation of the vehicle battery 41.

Advantageously, the anomaly prediction is carried out on the basis of the uncertainty-based prediction of at least one electrochemical internal battery state. For example, a quantile, e.g. the 98% quantile, can be used as an evaluation basis and compared against a predetermined threshold value. This threshold value can be based on a quantile of historical field data.

It is conceivable that the specified threshold value will be further refined and corrected on the basis of fleet data, with the combination of electrochemical parameters in particular having predictive power.

Claims (11)

Method for determining at least one predicted internal battery state of a specific device battery (41) of a technical device (4) and for detecting an anomaly of the specific device battery (41) based on the at least one predicted internal battery state, with the following steps: - providing (S3) the at least one internal battery state (Z) of the specific device battery (41) by evaluating an electrochemical battery model (9), in particular an aging model, based on operating variable profiles (F) of the specific device battery (41), wherein the electrochemical battery model (9) is parameterized and/or trained based on operating variable profiles of a plurality of device batteries (41) and can be evaluated via time series integration; - Providing a battery dynamics model parameterized using the at least one internal battery state (Z), in particular in the specific technical device (4), wherein the battery dynamics model (12) maps at least one battery current (I _mess ) to a battery voltage (U _mod ); - Adjusting or correcting (S4, S5, S6, S7) the at least one internal battery state (Z) using an observer model (13) based on a voltage difference between a battery voltage (U _mess ) of the specific device battery (41) measured in the specific device (4) and a battery voltage (U _mod ) of the specific device battery (41) modeled using the battery dynamics model at one or more points in time; - Predicting (S9) a predicted operating variable profile using a usage pattern model (15); - determining (S10) a predicted course of the at least one internal battery state depending on the predicted operating variable course with the aid of one or another electrochemical battery model (9, 16) which carries out the time series integration based on the at least one corrected internal battery state (Z') at a current point in time. Procedure according to Claim 1 , wherein the observer model (13) comprises a Luenberger observer or an extended Kalman filter or a particle filter. Procedure according to Claim 1 or 2 , wherein the predicted corrected internal battery state is signaled in the technical device (4). Procedure according to one of the Claims 1 until 3 , wherein an aging state model (9) is provided as an electrochemical battery model in a central unit (2) remote from the device, which comprises a physical aging model (5) which is based on a differential equation system and is designed to calculate an aging state by means of a time integration method, wherein the physical aging model (5) is fitted based on temporal operating variable profiles (F) of a multiplicity of device batteries (41) of the same type in order to assign an aging state to a temporal operating variable profile of the specific device battery (41). Procedure according to Claim 4 , wherein a battery performance model is provided which determines model parameters in the form of equilibrium parameters and/or kinetic parameters from time periods of operating variable curves of the plurality of device batteries and indicates these as internal battery states, wherein the internal battery states of the electrochemical battery model and the battery performance model are fused or statistically combined. Procedure according to one of the Claims 4 until 5 , wherein model parameters of the aging state model are transmitted to the technical device, so that the determination of the predicted course of the at least one internal battery state is carried out depending on the predicted operating variable course starting from the corrected internal battery states as a latent state of the physical aging model with the aid of a further aging state model executed in the respective device, which is executed with the transmitted model parameters. Procedure according to one of the Claims 1 until 6 , wherein a fault of the device battery is determined based on the predicted course of the at least one internal battery state, wherein in particular the fault of the device battery is determined based on a threshold value comparison of a lower limit and/or upper limit of a confidence range of the at least one internal battery state at a future point in time with a predetermined threshold value. Procedure according to one of the Claims 1 until 7 , wherein the confidence interval of the at least one corrected internal battery state is determined by evaluating the operating variable curves of the plurality of device batteries for the corresponding at least one internal battery state, in particular according to the law of large numbers. Device for carrying out one of the methods according to one of the Claims 1 until 8 . Computer program product comprising instructions which, when the program is executed by at least one data processing device, cause the device to carry out the steps of the method according to one of the Claims 1 until 8 to execute. Machine-readable storage medium comprising instructions which, when executed by at least one data processing device, cause the device to carry out the steps of the method according to one of the Claims 1 until 8 to execute.

Images