DE102021203729A1 - Method and device for operating a system for providing predicted aging states of electrical energy stores using reconstructed time series signals using machine learning methods - Google Patents

Abstract

The invention relates to a computer-implemented method for modeling a current state of health (SOH) or predicting a state of health (SOH) of an energy store (41) of a technical device, with the following steps: - providing or receiving (S11) a time profile of at least one operating variable (F) of at least one energy store (41);- providing a data-based aging state model (9), which is trained in order to assign a modeled aging state (SOH) of the energy store (41) to the at least one operating variable (F) depending on the course over time,- checking whether there is a time data gap in the time profile of the at least one operating variable (F) of the energy store (41);- when determining the time data gap (S12), generating a time profile of the at least one operating variable (F) for the time duration of the data gap using a usage pattern model (13), where the Usage pattern model (13) is designed to generate a time profile of the at least one operating variable (F) for the temporal data gap and/or at least one load variable (L), from which at least one Operating variable (F) can be derived, for the duration of the data gap; - Determining (S13, S14) the current state of health (SOH) or predicting the state of health depending on the course of the at least one operating variable (F) over time, which is generated by the temporal Course of the at least one operating variable (F) is supplemented in the temporal data gap, and using the aging model (9).

Description

technical field

The invention relates to electrical devices operated independently of the mains with electrical energy stores, in particular electrically drivable motor vehicles, in particular electric vehicles or hybrid vehicles, and also to measures for determining an aging state (SOH: State of Health) of the electrical energy stores. Furthermore, the invention relates to mobile and stationary electrical energy storage systems.

Technical background

The power supply of mains-independent electrical devices and machines such. B. electrically driven motor vehicles, using electrical energy storage devices, usually device batteries or vehicle batteries. Energy storage supplies electrical energy to operate the devices. Energy converters with an energy supply, such as fuel cells with a hydrogen tank, are also referred to herein as electrical energy stores.

The aging state of an energy storage device decreases over the course of its service life, which results in a decreasing maximum storage capacity. A measure of the aging of the energy storage device depends on an individual load on the energy storage device, i.e. in the case of vehicle batteries in motor vehicles on the usage behavior of a driver, external environmental conditions and the type of vehicle battery.

Although the current aging state of the energy store can be determined using a physical aging state model based on historical operating state profiles, this model is imprecise in certain situations. This inaccuracy of the conventional aging model makes it difficult to predict the course of the aging state. However, the prediction of the progression of the state of aging of the energy store is an important technical variable, since it allows an economic evaluation of a residual value of the energy store. Furthermore, the prediction of the course of the aging state of the energy storage device can be used for optimized operation of the energy storage device, e.g. in order to be able to depict a desired aging profile.

Disclosure of Invention

According to the invention, a method for predicting an aging state of an electrical energy store according to claim 1 and a device in an electrically operable device according to the independent claim are provided.

Further developments are specified in the dependent claims.

• - Empfangen eines zeitlichen Verlaufs mindestens einer Betriebsgröße von mindestens einem Energiespeicher;
• - Bereitstellen eines insbesondere datenbasierten Alterungszustandsmodells, das trainiert ist, um abhängig von dem zeitlichen Verlauf der mindestens einen Betriebsgröße einen modellierten Alterungszustand des Energiespeichers zuzuordnen,
• - Überprüfen, ob eine zeitliche Datenlücke in dem zeitlichen Verlauf der mindestens einen Betriebsgröße des Energiespeichers vorliegt;
• - bei Feststellen der zeitlichen Datenlücke, Generieren eines zeitlichen Verlaufs der mindestens einen Betriebsgröße für die zeitliche Dauer der Datenlücke mithilfe eines Nutzungsmustermodells, wobei das Nutzungsmustermodells ausgebildet ist, um abhängig von einem Nutzungsmuster und einer kalendarischen Zeitangabe einen zeitlichen Verlauf der mindestens einen Betriebsgröße für die zeitliche Datenlücke und/oder mindestens eine Belastungsgröße, aus der mindestens eine Betriebsgröße ableitbar ist, für die zeitliche Dauer der Datenlücke bereitzustellen;
• - Bestimmen des aktuellen Alterungszustands oder Prädizieren des Alterungszustands abhängig von dem zeitlichen Verlauf der mindestens einen Betriebsgröße, der durch den generierten zeitlichen Verlauf der mindestens einen Betriebsgröße in der zeitlichen Datenlücke ergänzt ist, und mithilfe des Alterungszustandsmodells.

According to a first aspect, a computer-implemented method for modeling a current state of health or predicting a state of health (SOH) of an energy store of a technical device is provided, with the following steps:

• - Receiving a time profile of at least one operating variable of at least one energy store;
• - Providing a data-based aging state model in particular, which is trained to assign a modeled aging state of the energy store to the at least one operating variable depending on the course over time,
• - checking whether there is a time data gap in the time profile of the at least one operating variable of the energy store;
• - when determining the temporal data gap, generating a time profile of the at least one operating variable for the duration of the data gap using a usage pattern model, wherein the usage pattern model is designed to generate a time profile of the at least one operating variable for the temporal provide data gaps and/or at least one load variable from which at least one operating variable can be derived for the duration of the data gap;
• - Determining the current aging status or predicting the aging status depending on the chronological progression of the at least one operating variable, which is supplemented by the generated chronological progression of the at least one operating variable in the temporal data gap, and using the aging status model.

For the purposes of this description, energy storage devices include portable batteries, energy converter systems with an electrochemical energy converter with a supply of energy carriers, such as fuel cell systems with a fuel cell and a supply of energy carriers.

The aging condition of an electrical energy store, in particular a device battery, is usually not measured directly. This would include a number of sensors near the Require energy storage, which would make the production of such an energy storage cost-intensive and complex and would increase the space requirement. In addition, measurement methods suitable for everyday use to determine the state of aging in the energy storage devices are not yet available on the market. Therefore, the current aging status is usually determined using a physical aging model in the devices. This physical state of health model is inaccurate in certain situations and typically has model deviations of up to more than 5%.

Due to the inaccuracy of the physical aging model, this can also only indicate the current aging state of the energy store. A prediction of the aging condition, which depends in particular on the mode of operation of the energy storage device, such as e.g. B. on the level and quantity of the charge inflow and charge outflow in a portable battery, and thus on usage behavior and usage parameters, would lead to very imprecise predictions and is currently not planned at the location of the energy storage device, e.g. for vehicle batteries in the vehicle.

The state of health (SOH: State of Health) is the key variable for specifying a remaining battery capacity or remaining battery charge for device batteries as electrical energy storage devices. The state of aging generally represents a measure of the aging of the electrical energy store, which indicates the performance of the energy store. In the case of a portable battery or a battery module or a battery cell, the state of health can be specified as a capacity retention rate (SOH-C) or as an increase in internal resistance (SOH-R). The capacity retention rate SOH-C is given as a 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.

Approaches that provide user-specific and use-specific modeling and prediction of an aging state of the electrical energy storage device based on a aging state model that uses the progression of operating variables from the time of commissioning to adapt the aging state time step by time step based on the aging state at the time of commissioning are promising. This aging model can be implemented purely data-based but also as a hybrid data-based aging model. Such an aging state model can e.g. B. be implemented in a central unit (cloud) and be parameterized or trained using operating variables of a variety of energy storage devices of different devices that are in communication with the central unit.

Aging models for determining aging states for electrical energy storage devices can be provided in the form of a hybrid aging state model, ie a combination of a physical aging model with a data-based model. In a 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 the electrochemical states of a nonlinear differential equation system, continuously calculated and mapped to the physical aging state for output, as SOH-C and/or as SOH-R. The calculations can typically be carried out in the cloud, e.g. once a week. The calculation of the physical aging model is based on a time integration method to solve the differential equation system.

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

A prediction of the aging state is helpful, for example, if the remaining service life of the energy storage device is to be determined and this is to be evaluated against guarantee conditions or CO2 fleet specifications, for example. For this purpose, the data-based state of health model can be continuously queried in connection with a predefined usage pattern that specifies the type of usage and operation of the electrical energy store. For this purpose, starting from the current point in time, ongoing generation of future time profiles of artificial operating variables is necessary, which is derived from the physical cal aging model due to the time integration method for solving the differential equation are needed to model a predicted aging state. In this way, in particular, the predicted course of the aging state is determined based on the current point in time. For this purpose, the operating variables are generated either directly as a function of the recognized usage pattern or based on curves of load variables derived from the usage pattern, from which the curves of the required operating variables are generated.

The usage pattern model can thus provide the time profile of the at least one load variable, with an operating variable model being designed to generate a time profile of at least one load variable from the time profile of the at least one load variable, which is required on the input side for the state of health model.

This possibility of prediction uses the trained aging state model and usage pattern in an advantageous manner, so that a more precise prediction of the aging state is possible than with purely extrapolating methods.

Since the equations for the electrochemical physical aging states are solved in the differential equation system using time integration methods, it is necessary for the inputs of the physical aging model to be available as time series or time curves. In order to model an aging state of an energy storage device using a physical or electrochemical aging state model and an optional specification using a data-based correction model (hybrid aging state model), it is necessary to provide the time profiles (time series) of operating variables at a relatively high frequency. These time curves of the operating variables must continue to be provided as completely as possible for the necessary accuracy requirements, i. H. In order to determine the aging status of a device battery at a current point in time, it is necessary to provide the time profiles of the battery temperature, the battery current, the battery voltage and the state of charge, in particular at cell level.

The calculation of the electrochemical model together with the correction model preferably takes place external to the device, since this is very computationally expensive and the required processing power in or hardware-related to the battery-operated devices is often insufficient or should not be kept available for cost reasons. Furthermore, the electrochemical physical aging model can be retrained from time to time (e.g. every 6 months) based on newly generated labels in a device-external central unit (cloud) using data from a large number of energy storage devices.

Since the equations for the physical aging states are solved in the differential equation system using time integration methods, it is necessary for the inputs of the electrochemical aging model to be available as a time series. These can include, for example, temperature, current and voltage. For capacity reasons, the time curves of the operating variables are transmitted to a device-external central unit and the aging condition is determined there according to the physical aging model (electrochemical model) and, if necessary, the correction model.

However, due to various reasons, data gaps can arise in the time series of company sizes. For example, a longer parking phase and/or sporadic use of the device can result in longer-lasting signal gaps that are not easy to fill. Due to connectivity problems such as a disruption in network communication, information on the development of the operating variables can be lost, which leads to incomplete time series. In this regard, it is necessary to fill or reconstruct the time course of the operating variables in signal gaps that occur in order to be able to operate the aging model used to determine a current or future aging status. As a result, the robustness and accuracy of the modeled and/or predicted aging state can be significantly increased. The above method makes it possible to complete the time profile of the at least one operating variable required for determining the state of health, in particular the battery temperature, the battery current, the battery voltage and the state of charge of a device battery as an energy store.

Reconstruction is performed using a self-learning usage pattern model that progressively increases the accuracy of reproduction as the amount of data increases. By using plausible assumptions, the usage pattern model can fill gaps in the time curves of the operating variables and thereby significantly increase the robustness of the aging model by increasing the input quality of the aging model. In combination with a separate observational basic aging model for a different determination of an aging state, which is used for verification, the subsequently reconstructed load pattern can be evaluated and validated based on the model.

The usage pattern model can be device-specific based on the provided or emp be formed or trained over the course of the at least one operating variable.

The above method provides for the use of a usage pattern model that generates the time curves of the at least one operating variable from one or more usage parameters of a usage pattern or contributes to its generation. The learned usage pattern is characterized by the parameterization or the trained model parameters of the usage pattern model. The usage pattern is used to simulate a time series of the at least one operating variable based on the usage pattern model. Thus, with the usage pattern model, time series predictions of the learned load patterns are possible in the form of chronological progressions of at least one operating variable and/or at least one load variable, such as current or temperature.

The course of the at least one company variable over time is generated using a usage pattern model. The usage pattern model is designed to output continuous curves of at least one variable, depending on one or more usage parameters that the usage pattern explicitly or implicitly, i.e. with the help of a data-based model, which can correspond to at least one load variable that can correspond to the at least one operating variable or from which the at least one operating variable can be generated based on a model. For a device battery as an energy store, the at least one load variable can correspond, for example, to the time profile of a battery temperature and/or a battery current. This makes it possible to convert usage behavior parameterized by the usage pattern into a time profile of the at least one operating variable. The usage pattern can thus indicate types of load on the energy store in the form of the time profile of the at least one load variable and/or the at least one operating variable.

Provision can be made for the usage patterns, in particular specified by time series of the at least one load variable, to be created on the basis of data-based usage pattern models with the aid of historical usage behavior, with the usage patterns created being determined in particular to predict the aging status.

The usage pattern model is preferably designed as a data-based model that characterizes the type of operation of the energy store. For this purpose, based on usage data, which is reflected in the course of at least one operating variable in relation to its calendar time information (date, time, weekday, season, etc.), artificial courses of operating variables are generated, which lead to a comparable load on the energy storage device and thus cause a similar aging behavior that is representative for the user.

Alternatively, the usage pattern model can be used as a hybrid usage pattern model with a non-data-based model that, depending on an operating characteristic, provides a chronological progression of the at least one operating variable for the temporal data gap and/or the at least one load variable for the duration of time depending on the calendar time specification, and a data-based usage correction model for correcting the course over time of the at least one operating variable for the temporal data gap and/or the at least one load variable for the duration depending on the calendar time specification.

The usage pattern results from the model parameters of a usage pattern model, which can be data-based, in particular as a recurrent neural network, such as a Bayesian LSTM (LSTM: Long Short Term Memory) or other models that use supervised learning methods, such as Gaussian processes or alternatively use attention layer. This is trained by assigning the calendar time specification to the at least one time profile of the at least one load variable and/or the at least one operating variable. The trained usage pattern model then assigns a corresponding course of the at least one load variable or the at least one operating variable to a predefined calendar time specification or generates this with a corresponding default of the calendar time specification. The calendar time specification can specify the calendar date and a time and information to be derived therefrom, such as day of the week, season or month and the like.

If data gaps in the time profiles of the operating variables are now determined, artificial time profiles of the operating variables for the data gap are generated based on the usage pattern model previously trained with the aid of previously recorded temporal profiles of the at least one load variable and/or the at least one operating variable.

Historical load profiles in the form of the time profile of the at least one load variable and/or the at least one operating variable are used as training data, ie for a device battery, for example, the current profile as a function of the calendar time or battery temperature as a function of the calendar time.

If a time profile of at least one load variable is provided by the usage pattern model, an operating variable model can be used that corrects the at least one load variable and provides the corrected load variable as an operating variable. The operating variable model makes it possible to characterize the usage behavior with regard to the load on the energy store in a parameterizable manner and to fit the course of the at least one operating variable during operation. The operating variable model can include at least one domain model or a data-based model in order to learn about non-linearities in usage behavior, to describe them and to be able to calculate these non-linearities for future predictions. For example, efficiency effects or changed usage behavior based on permanently changed or seasonal capacity restrictions, e.g. more frequent charging by the user in winter to complete the same mileage.

With the help of the usage pattern model and the company size model, all available time series are used for training on a device-specific basis. With the help of the usage pattern model, all device-specific usage patterns can be implicitly learned, such as periodic patterns, which enable a distinction to be made between weekdays and weekends, for example. In the identified data gap, the missing temporal progression of the at least one operating variable can be reconstructed, e.g. the operating variable model is used to determine the progression of the state of charge of the battery current, the battery voltage and the battery temperature from the artificially generated progression of the load variable during a disruption or a longer standstill phase reconstructed. These can be used as an input for the aging model, in order to use a time profile of the at least one operating variable, which consists of the time periods in which the time profile of the at least one operating variable was recorded, and the time periods in which the time profile of the at least one company size was artificially generated.

Provision can be made for a data gap to be determined if the device is inactive for at least a predetermined period of time and/or there is no information on the time profiles for the operating variables due to connectivity problems. For example, a data gap in the time history of the operating variables can be determined if operating variables have not been recorded for a predetermined period of time, such as two weeks, as can happen, for example, when the technical device is not used, or if there is a communication disruption between the technical device and the central unit. This can be recognized, for example, by providing the operating variables with time stamps or other operation-dependent variables such as a mileage for a vehicle, so that discontinuities in the time profiles of the operating variables can be determined. Even in the event of a standstill, during which no operating variables are recorded, the chronological progression of the at least one operating variable and/or the chronological progression of the at least one load variable are reconstructed by the useful model. For example, the temperature profile, which also contributes to calendar aging (in combination with the SOC status), is reconstructed in order to be able to operate the physical aging model with time series data.

Furthermore, a further state of health model can be provided in a central unit which is connected to the plurality of devices. The further aging state model can be designed to be observational, based on a time profile of the at least one operating variable using empirical methods such as Coulomb counting or measurement of a change in internal resistance, trajectory points that each assign a further aging state to an aging point in time, with an aging state trajectory for the plurality of energy storage devices a confidence band for the trajectory points that indicates the accuracy of the estimate of each trajectory point. Furthermore, the usage pattern model can be adjusted if the determined aging status of the further aging status model for a current or considered aging time is outside the confidence band for the current or considered aging time.

The further observational aging status model can be used to check and correct the chronological progression of the at least one load variable and/or the at least one operating variable if the aging status of the aging status model for a current or considered predicted evaluation period is outside the confidence band for the current or considered predicted evaluation period, i.e. if the reconstructed time profile of the at least one operating variable and/or the time profile of the at least one stress variable does not match the independent measurements of the further aging status model being observed.

In particular, using the further aging state model based on electri Technical operating parameters of the energy store or the device batteries, such as an aging state being determined during a charging process, in order to determine an aging state trajectory in the central unit based on a large number of data points determined in this way. The base model uses observer-based algorithms in the energy storage management system and can fuse multiple historical observations of the state of health into a trajectory of the state of health with statistical confidence. The further aging state model can also be designed in such a way that it can calibrate itself to aging and temperature-dependent OCV characteristics (open circuit voltage characteristics) in idle phases in order to improve the calculation of the aging state.

The physical aging model and, if applicable, the correction model and the usage pattern model, which is used to provide time profiles of operating variables during data gaps, a deviation between the reconstructed profile of the at least one operating variable and/or the time profile of the at least one load variable or their simulated effects on the Aging of the energy store and the real measured aging are evaluated and used to iteratively improve the artificially generated time curves of the at least one load variable and / or the at least one operating variable.

In particular, if the deviation of the modeled state of health trajectory for a specific device battery and the state of health trajectory determined generally for the device battery type and generated based on basic models causes a deviation above a specified tolerance threshold value, a new time profile of the at least one load variable and/or the at least one can be created for the data gap Operating size are generated. The generation of the time curve of the at least one load variable and/or the at least one operating variable can be determined by sampling from a confidence band of the reconstructed time curves of the operating variables. In the case of a data gap with a duration that is less than a maximum duration, this can be done by interpolation and thus reconstruction of the time profile of the at least one operating variable, such as the battery temperature, the battery current, the state of charge and the battery voltage in the case of a device battery. For example, with a maximum duration of e.g. less than 30 s, the time curves can be interpolated purely based on data, because the temperature cannot change significantly during this time.

If the aging determined using the hybrid aging state model is determined to be too low during a data gap, the operating variables can be sampled from a part of the confidence band that causes a higher aging effect, and vice versa. A numerical optimization method such as a Bayesian optimization method can preferably be used for this purpose in order to be able to precisely reconstruct the reconstructed time curves with inherent stress factors. Alternatively, numerical optimization methods such as Regula Falsi (Pegasus method) can also be used.

Alternatively or additionally, the/another aging state model can be provided as an observation, which is designed to determine trajectory points that assign a further aging state to an aging time based on a time profile of the at least one operating variable by empirical methods such as Coulomb counting or measuring a change in internal resistance to model an aging state trajectory for the plurality of energy stores, the modeled aging state being determined by merging the aging state of the aging state model for a current or observed aging point in time depending on the determined aging state of the aging state model and a further aging state of the further aging state model. The merging can take place taking into account the confidences of the aging models used. In the hybrid aging model, the confidences correspond exactly to the uncertainty of the probabilistic or data-based correction model or the uncertainty of the Gaussian process model. The confidences of the further observed aging model are estimated empirically from the combination of the individual observations and taking into account the error propagation from the sensor (e.g. the battery temperature sensor) to the calculated model value (e.g. observed SOHC).

Furthermore, the energy store or the device battery can be operated depending on the course of the predicted modeled state of health (SOH), with a remaining service life of the energy store depending on the course of the predicted modeled state of health (SOH) being signaled in particular. Furthermore, the operating strategy of the energy store can be modified based on the predicted remaining service life in order to positively influence the remaining service life. Possible measures include adjusting operating intervention limits, such as derating limits with regard to permissible battery temperatures.

Furthermore, the energy store can be used to operate a device such as a motor vehicle, a pedelec, an aircraft, in particular a drone, a machine tool, a consumer electronics device such as a mobile phone, an autonomous robot and/or a household appliance.

Furthermore, the current and future usage behavior can be taken into account by updating the usage pattern model based on updating the parameters of the operating variable model when the parameters of the operating variable model are updated depending on the calculated aging of a further aging state model.

• - Bereitstellen oder Empfangen eines zeitlichen Verlaufs mindestens einer Betriebsgröße von mindestens einem Energiespeicher;
• - Bereitstellen eines insbesondere datenbasierten Alterungszustandsmodells, das trainiert ist, um abhängig von dem zeitlichen Verlauf der mindestens einen Betriebsgröße einen modellierten Alterungszustand des Energiespeichers zuzuordnen,
• - Überprüfen, ob eine zeitliche Datenlücke in dem zeitlichen Verlauf der mindestens einen Betriebsgröße des Energiespeichers vorliegt;
• - bei Feststellen der zeitlichen Datenlücke, Generieren eines zeitlichen Verlaufs der mindestens einen Betriebsgröße für die zeitliche Dauer der Datenlücke mithilfe eines Nutzungsmustermodells, wobei das Nutzungsmustermodells ausgebildet ist, um abhängig von einem Nutzungsmuster und einer kalendarischen Zeitangabe einen zeitlichen Verlauf der mindestens einen Betriebsgröße für die zeitliche Datenlücke und/oder mindestens eine Belastungsgröße, aus der mindestens eine Betriebsgröße ableitbar ist, für die zeitliche Dauer der Datenlücke bereitzustellen;
• - Bestimmen des aktuellen Alterungszustands oder Prädizieren des Alterungszustands abhängig von dem zeitlichen Verlauf der mindestens einen Betriebsgröße, der durch den generierten zeitlichen Verlauf der mindestens einen Betriebsgröße in der zeitlichen Datenlücke ergänzt ist, und mithilfe des Alterungszustandsmodells.

According to a further aspect, a device for modeling a current state of health or predicting a state of health (SOH) of an electrical energy store of a technical device is provided, the device being designed for:

• - Providing or receiving a time profile of at least one operating variable of at least one energy store;
• - Providing a data-based aging state model in particular, which is trained to assign a modeled aging state of the energy store to the at least one operating variable depending on the course over time,
• - checking whether there is a time data gap in the time profile of the at least one operating variable of the energy store;
• - when determining the temporal data gap, generating a time profile of the at least one operating variable for the duration of the data gap using a usage pattern model, wherein the usage pattern model is designed to generate a time profile of the at least one operating variable for the temporal provide data gaps and/or at least one load variable from which at least one operating variable can be derived for the duration of the data gap;
• - Determining the current aging status or predicting the aging status depending on the chronological progression of the at least one operating variable, which is supplemented by the generated chronological progression of the at least one operating variable in the temporal data gap, and using the aging status model.

character list

• 1 eine schematische Darstellung eines Systems zur Bereitstellung von fahrer- und fahrzeugindividuellen Betriebsgrößen zur Bestimmung eines Alterungszustands einer Fahrzeugbatterie in einer Zentraleinheit;
• 2 eine schematische Darstellung eines funktionalen Aufbaus eines hybriden Alterungszustandsmodells;
• 3 ein Flussdiagramm zur Darstellung eines Verfahrens zum Trainieren eines datenbasierten Alterungszustandsmodells; und
• 4 eine schematische Darstellung eines funktionalen Aufbaus eines hybriden Alterungszustandsmodells mit nutzungsabhängiger Prädiktion des Alterungszustands;
• 5 ein Flussdiagramm zur Veranschaulichung eines Verfahrens zum Ermitteln eines Alterungszustands bei Datenlücken in den zeitlichen Verläufen der Betriebsgrößen; und
• 6 ein Verlauf einer Alterungszustandstrajektorie über einem Alterungszeitpunkt von Fahrzeugbatterien einer Vielzahl von Fahrzeugen nebst einem Konfidenzband.

Embodiments are explained in more detail below with reference to the accompanying drawings. Show it:

• 1 a schematic representation of a system for providing driver and vehicle-specific operating variables for determining an aging condition of a vehicle battery in a central unit;
• 2 a schematic representation of a functional structure of a hybrid aging model;
• 3 a flow chart for representing a method for training a data-based state of health model; and
• 4 a schematic representation of a functional structure of a hybrid state of health model with use-dependent prediction of the state of health;
• 5 a flowchart to illustrate a method for determining an aging state in the event of data gaps in the time profiles of the operating variables; and
• 6 a course of an aging state trajectory over an aging point in time of vehicle batteries of a large number of vehicles, together with a confidence band.

Description of Embodiments

The method according to the invention is described below using vehicle batteries as electrical energy stores in a large number of motor vehicles as devices of the same type. A data-based state of health model for the respective vehicle battery can be implemented in a control unit in motor vehicles. The state of health model can be continuously updated or retrained in a central unit based on operating variables of the vehicle batteries from the vehicle fleet. The aging state model is operated in the central unit and used for aging calculation and aging prediction.

The above example is representative of a large number of stationary or mobile devices with a network-independent energy supply, such as vehicles (electric vehicles, pedelecs, etc.), systems, machine tools, household appliances, IOT devices and the like, which have a corresponding communication connection (e.g. LAN, Internet, LTE/5G) are connected to a central unit (cloud).

1 shows a system 1 for collecting fleet data in a central unit 2 for the creation and operation as well as for the evaluation of an aging model. The aging state model is used to determine an aging state of an electrical energy store such. B. a vehicle battery or a fuel cell in a motor vehicle. 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 as a rechargeable electrical energy store, an electric drive motor 42 and a control unit 43 . The control unit 43 is connected to a communication module 44 which is suitable for transmitting data between the respective motor vehicle 4 and a central unit 2 (a so-called cloud).

The motor vehicles 4 send the operating variables F to the central unit 2, which specify at least variables on which the aging state of the vehicle battery depends or via which it can be determined. In the case of a vehicle battery 41, 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 the pack, module and/or cell level. The operating variables F are recorded in a fast time frame from 2 Hz to 100 Hz and 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.

Operating characteristics M, which relate to an evaluation period, can be generated from the operating variables F in the central unit 2 or, in other embodiments, already in the respective motor vehicles 4 . The evaluation period for determining the aging state can be a few hours (e.g. 6 hours) to several weeks (e.g. one month). A usual value for the evaluation period is one week, i.e. once a week a new evaluation of the aging status of each vehicle battery takes place.

The operating characteristics can include, for example, characteristics relating to the evaluation period and/or accumulated characteristics and/or statistical values ascertained over the entire service life to date. In particular, the operational characteristics may include, for example: electrochemical conditions such as SEI layer thickness, change in cyclizable lithium due to anode/cathode side reactions, rapid uptake of electrolyte solvent, slow uptake of electrolyte solvent, lithium deposition, loss of anode active material and loss of cathode active material, the internal resistances, histogram features, such as temperature versus state of charge, charging current versus temperature and discharging current versus temperature, in particular multi-dimensional histogram data regarding the battery temperature distribution versus state of charge, the charging current distribution versus temperature and/or the discharging current distribution versus temperature, the current throughput in ampere hours, the accumulated total charge (Ah), an average capacity increase over a charge (particularly for charges where the charge increase is above a threshold proportion (e.g. 20%) of the total battery capacity), the charging capacity and an extreme value (maximum) of the differential capacity dQ/dU or the accumulated mileage. These variables are preferably converted in such a way that they characterize the real usage behavior in the best possible way. For example, the accumulated charge throughput (Ah) is normalized with the SOHR so that the poorer battery efficiency for covering the same route (in km) is correctly mapped. The operating features M can be used in whole or in part for the method described below.

Further information can be derived from the operating characteristics M and the operating variables F: a load pattern over time such as charging and driving cycles, determined by usage pattern N (such as fast charging at high current intensities or strong acceleration or braking processes with recuperation), a usage time of the vehicle battery , a charge accumulated over the term and a discharge accumulated over the term, a maximum charging current, a maximum discharging current, a charging frequency, an average charging current, an average discharging current, a power throughput during charging and discharging, a (particularly average) charging temperature, a ( in particular average) spread of the state of charge and the like. This temporal load pattern characterizes the typical temporal usage behavior and can be used for prediction.

The state of health (SOH: State of Health) is the key variable for indicating a remaining battery capacity or remaining battery charge. The aging state is a measure of the aging of the vehicle battery or a battery module or a battery cell and can be specified as a capacity retention rate (SOH-C) or as an increase in internal resistance (SOH-R). The capacity conservation rate of charge SOH-C is given 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.

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 and the like.

A state of health model is implemented in the central unit 2, which is in particular entirely or partially data-based. The aging state model can be used regularly in order to determine the current aging state of the vehicle battery 41 based on the operating characteristics and/or the operating variables. In other words, it is possible, based on the operating variables and/or the operating characteristics that result from the operating variable curves of one of the motor vehicles 4 of the fleet 3, to determine an aging state of the vehicle battery 41 in question or of the modules or cells associated with this energy store.

2 shows a schematic example of the functional structure of an embodiment of a hybrid data-based state of health model 9, which is constructed in a hybrid manner. The hybrid aging state model 9 includes a physical aging model 5 and a correction model 6. These receive operating variables F or operating characteristics M of a current evaluation period/aging time (age of the vehicle battery since the time of commissioning). The operating characteristics M of the current evaluation period/time of aging are generated in a characteristic extraction block 8 based on the time series of the operating variables F.

The operating variables F go directly into the physical aging state model 5 as time series data, including temperature and current, which is preferably designed as an electrochemical model and corresponding electrochemical states, such as layer thicknesses (e.g. SEI thickness), change in the cyclable lithium due to anode/cathode -side reactions, rapid consumption of electrolytes, slow consumption of electrolytes, loss of active material in anode, loss of active material in cathode, etc...) using non-linear differential equations.

The physical aging model 5 can correspond to an electrochemical model of the battery cell and the cell chemistry. Depending on the operating variables F, this model determines internal physical battery states in order to create a physically based aging state SOHph of at least one dimension in the form of the electrochemical states mentioned above, which are linearly or non-linearly related to a capacity maintenance rate (SOH-C) and/or an internal resistance increase rate (SOH -R) to provide these (SOH-C and SOH-R). The physical aging model 5 can be parameterized using training data sets from the vehicles 4 of the vehicle fleet 3 and/or using laboratory data.

However, the model values for the physical aging state SOHph provided by the electrochemical model are imprecise in certain situations, and it is therefore provided to correct them with a correction variable k. The correction variable k is provided by the data-based correction model 6, which is trained using training data records from the vehicles 4 of the vehicle fleet 3 and/or using laboratory data.

To determine a corrected aging state SOH to be output, the outputs SOHph, k of the physical aging model 5 and of the correction model 6, which is preferably designed as a Gaussian process model, are applied to one another. In particular, these can be added or multiplied (not shown) in a summation block 7 in order to obtain the modeled aging state SOH to be output for a current evaluation period or aging time. In addition, 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 correction model 6 receives operating features M on the input side, which are determined from the curves 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 aging 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 features 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 features M also include the internal states from the state vector of the electrochemical physical aging model 5 and advantageously the physical aging state (SOHph).

Other configurations of the data-based aging model are also possible, for example the data-based aging model can be non-hybrid, purely data-based th model based on a probabilistic regression model or a regression model based on artificial intelligence, in particular a Gaussian process model, or a Bayesian neural network. This is trained to provide a modeled aging state SOH from an operating feature point that is determined by current operating features M of the current evaluation period/ageing time, with the operating features M being determined in a feature extraction block 8 based on the time series of the operating variables F.

For scaling and dimensionality reduction of the operational features, a PCA (Principal Components Analysis) can be used to appropriately reduce redundant linear-dependent information in the feature space before training the correction model (unsupervised). Alternatively, a kernel PCA can also be used in order to be able to map non-linear effects in the complexity reduction of the data. A normalization of the entire operating feature space (or the main component space) takes place both before the dimension reduction and especially afterwards, e.g. B. with min/max scaling or the Z transformation.

The calculation of the aging state and the prediction of the aging state are therefore possible for energy storage devices with at least one electrochemical unit, e.g. B. a battery cell. The method can also be applied to the overall system of the energy store through rule-based and/or data-based mapping. Using the battery as an example, the aging prediction can be applied not only at the cell level but also directly at the module and pack level.

Training data sets are defined to train the hybrid aging model. These training data records can be collected in the central unit 2 from a large number of vehicles. While the physical aging state model can be parameterized in a manner known per se by measuring vehicle batteries or their battery cells, the correction model 6 can be trained for a deviation of an actual aging state and an aging state SOHph modeled by the physical aging state model 5 . The training data points therefore map the operating characteristics of an operating characteristic point to the corresponding deviation between the actual aging state and the aging state SOHph modeled by the physical aging state model.

3 shows a flow chart to illustrate a further method for training the hybrid aging model in the central unit 2. For this purpose, the training data sets are divided into a training set and a test set. The training set is used to train the hybrid aging model, while the test set is used to validate the hybrid aging model using new, unknown data.

The determination of an aging state as a label can be done in a manner known per se by evaluating the performance variable curves with an additional model in the vehicle or in the central unit 2 under defined load and environmental conditions of a label generation, such as in a workshop, on a test bench or a diagnosis 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. For this purpose, other models can be used to determine the aging status, e.g. based on the analysis of a recognized charging and/or discharging phase of battery use. An SOH-C estimate is preferably made by Coulomb counting or by forming a current integral over time during the charging process, which is divided by the difference in the state of charge between the beginning and end of the relevant charging and/or discharging phase. In this case, the calibration is advantageously carried out on the no-load voltage characteristic in idle phases in order to also calculate the course of the state of charge in the central unit. A sufficiently reliable indication of the aging state for use as a label can be obtained, for example, if the vehicle battery is brought from a fully discharged state of charge to a fully charged state during a charging process from a defined relaxed state under reproducible load and environmental conditions. The maximum charge detected in this way can be related to an initial maximum charge capacity of the vehicle battery. Resistance-related aging states (SOH-R values) can also be calculated from voltage changes related to a current change. These are usually related to a defined time interval.

A training data set for a vehicle battery is thus obtained from the state of health label determined at a particular point in time and the performance variable curves for the relevant vehicle battery up to this point in time. A number of training data sets can be determined for a vehicle battery at different points in time, with the points in time preferably being specified relative to the start of the service life. the train ning data sets are collected and made available for the large number of vehicles.

The aging model can be trained in a conventional manner using the training data sets. I.e. while retaining the physical aging model 5, the training data sets are evaluated by the hybrid aging state model 9 and an error measure, e.g. the RMSE (relative mean squared error) (loss function) between the output value of the modeled aging state SOH of the training data set under consideration and the associated label Adaptation and training of the correction model 6 used in a manner known per se. It is provided here that the training takes place on the residue of the physical model, so that the correction model can make data-driven corrections exactly where the data situation allows it with sufficient confidence. The training data sets then represent the training set.

Alternatively, the training can be done by dividing the training data sets into a training set and a test set. The training set is used to train the hybrid aging model, while the test set is used to validate the hybrid aging model using new test set unknown data not used for training. A third data set, the validation data set, is preferably used in order to optimize the hyperparameters of the correction model. Finally, the hybrid aging state model 9 is always tested on new data, with the performance of the hybrid aging state model 9 being verified on this independent data set before it is deployed in the central unit 2 .

In step S1, the physical aging model 5 is parameterized as a function of a first part of the training set, in particular by parameter optimization using the least squares method or the like. The physical aging state SOHph as the output of the physical aging model 5 is assumed to be the aging state of the respective training data set.

In step S2, the physical aging model is applied to the entire training set of the hybrid model, i. H. a number of training data sets that is at least equal to, or even exceeds, the set of training data sets with which the physical aging model has been parameterized. The error of the physical aging model 5 is correspondingly evaluated in a total error with respect to the residue as a histogram of the model deviation. In combination with the operating characteristics M or the operating variables F, this residual contains all relevant information regarding the systematic weaknesses of the physical aging model 5. Information is also obtained as to how the physical aging model 5 was not used for the parameterization of the physical aging model 5 with regard to new ones training datasets behaves.

In a next step S3, the data-based correction model 6 is trained on the complete training set of the hybrid model. This training set of the hybrid model includes at least the training set of the physical model according to step S1. For the training of the correction model 6, both the operating characteristics M are extracted from the operating variables F and the internal states of the physical aging model 5 are used as a subset of M in order to check all operating characteristics for an error between the model prediction (physical aging state) of the physical aging model and the map the labeled aging state according to the training data set. As a result, the correction model 6 can learn the weaknesses of the physical aging model 5 in order to be able to correct the physical aging state in the correction block.

The training of the data-based correction model 6 can be performed with cross-validation and sequential bagging (bootstrap aggregating) to improve robustness and accuracy. When the correction model is trained, the trained hybrid state of health model can be validated using the test set in step S4, so that the overall performance for the state of health calculation can be validated.

The trained hybrid aging state model can now be used to determine the aging state based on operating variables F.

The training of the hybrid aging model can be triggered whenever new labeled data is available, especially if it contains new and relevant information. When operating in a central unit based on fleet data, continuous retraining of the hybrid aging state model to determine the aging state and to predict the aging state is thus possible.

To predict an aging condition depending on usage data, such as e.g. B. usage pattern N of a driver of a motor vehicle, a model can be used as described in 4 is shown. The trained hybrid aging model according to 2 can therefore also have a flag dimension extraction block and a dimension reduction block (e.g. with principal component analysis: PCA).

4 shows an overall structure of a system for determining an aging condition in the central unit 2, which is based on the hybrid aging condition model 2 is based, in which case an operating variable model 10 is additionally used in order to artificially generate time profiles of operating variables, such as the battery voltage U, the state of charge SOC, the battery current I and the battery temperature T, if the provision of operating variables from the vehicles has 4 data gaps . The provision of the operating variable model 10 is necessary because the aging state model 9 requires time series or time profiles of all operating variables F. The determination of the operating variable curves of all operating variables is ensured by the operating variable model 10 and takes place as a function of the time curves of load variables and the state of health SOH of the vehicle battery 41. The operating variable model can have at least one domain model for determining at least some of the operating variables from one or more of the load variables or a data-based model, in order to learn non-linearities in usage behavior, to describe them and to be able to calculate these non-linearities for future predictions. For example, efficiency effects or changed usage behavior based on permanently changed or seasonal capacity restrictions, eg more frequent charging by the user in winter to complete the same mileage, can be taken into account.

Operating variable model 10 is generally designed to model a battery current and a battery temperature by taking into account electrochemical equilibrium states, a battery voltage and a state of charge. Operating variable model 10 uses an artificially generated profile of one or more load variables L, which characterize a load on vehicle battery 41 as a result of its operation over a longer period of time.

The artificially generated profile of one or more load variables L, here the battery current and the battery temperature, are specified by a usage pattern model 13 . In order to generate the curves of the load variables L for determining or predicting the aging state, the usage pattern model 13 converts a predefined usage pattern N into a time curve of the load variables L as a function of a calendar time specification that reflects the load on the vehicle battery 41, which it uses during the is subjected to the use and operation specified by the usage pattern. The usage pattern N thus leads to the output of a time profile of a battery current I and a battery temperature T as load variables L by the usage pattern model 13, with which the set of operating variables F with the profiles of the battery voltage U and the state of charge SOC is completed with the aid of the operating variable model 10.

The usage patterns N can be checked for plausibility by the operating variable model and, if necessary, iteratively improved in combination with the usage pattern model 13 in order to also be able to map non-linearities in the predicted usage behavior, e.g. changed usage behavior due to a changed efficiency or specifically lower capacity.

The usage patterns N are defined by usage parameters N, which are learned vehicle-specifically by the usage pattern model 13, preferably with the aid of data-based methods, and are used to simulate the usage behavior of a user or a drive train with regard to the relevant vehicle battery 41.

The usage pattern model 13 can be designed as a recurrent neural network, such as an LSTM (Long Short Term Memory) or GRU, in particular as a Bayesian LSTM network, and based on the progression of load variables or operating variables that are a type indicate the use of the energy store, be trained. The curves of load variables or operating variables to be taken into account should be based on a period of the same type of use and the same operating mode of the battery.

Alternatively, the usage pattern model 13 can be in the form of a hybrid model. For this purpose, a non-data-based model can be specified, which, depending on one or more cumulative operating characteristics, such as the Ah throughput, and the calendar time specification, provides a chronological progression of the at least one load variable and/or the at least one operating variable, which relevant operating feature which is corrected accordingly by a data-based usage correction model. The data-based usage correction model can be designed as a recurrent neural network, such as an LSTM or GRU, in particular as a Bayesian LSTM network, and based on curves of load variables or operating variables that indicate a type of use of the energy store, be trained to correct the time course of the at least one load variable and/or the at least one operating variable from the non-data-based model.

The usage parameters that indicate the usage pattern then correspond to the model parameters of the usage pattern model, ie in the case of a neural network the weights and bias values of the individual neurons. Furthermore, prior and posterior distributions as well as probabilities conditioned on observations according to Bayes' theorem can be regarded as relevant parameters.

The usage patterns result from training the usage pattern model 13 based on known profiles of the load variables and/or the operating variables with regard to their calendar reference. I.e. the usage pattern model is trained on the input side with a calendar time specification and on the output side with the load variables assigned to the corresponding time specification (current, temperature preferably as a time series) and/or the operating variables in a manner known per se. By specifying a date and a time using the trained usage pattern model, an artificial course of the load variables and/or the operating variables can be generated. The calendar time specification can also contain the day of the week, the month and information about public holidays and, in particular, take account of seasonalities through feature engineering.

As long as no driver change has been detected, the usage pattern model 13 is regularly trained, for example once a month, taking into account new data. Typically, in the middle of this section, a period of time is extracted from the training data set that is not used for training but to validate the usage pattern model. Typically, a Bayesian LSTM network is used for the driver-specific characterization of the current flow. Furthermore, a Bayesian LSTM network is used for the driver-specific characterization of the temperature profile. Alternatives to the LSTM approach are ARIMAX models or Gaussian process models, especially for modeling the temperature curve.

The usage pattern model can thus be formed directly from raw data of the curves of the load variables L and/or the operating variables F. For the vehicle battery 41, typical patterns of the current profile, e.g. due to recurring commuter routes and typical standing and rest times and loads for temperature ranges, are recognized and made reproducible.

The usage patterns N can thus implicitly indicate types of loading of the energy store, in particular periodic loads.

The usage pattern N can also indicate, in particular, environmental conditions and a periodic load profile. The environmental conditions can, for example, be derived from a climate table, indicate a course of the battery temperature within a day-night rhythm, for the seasons and the like, preferably with the help of GPS-dependent weather data from the central unit (cloud). For this purpose, the usage pattern model can be trained and used in addition to the calendar time specification with a temperature progression specification. Predictions of GPS-dependent temperature profiles can preferably be incorporated into the prediction of the useful pattern.

The temperature history indication may result from an average temperature over a recent period, such as a month, which may be predicted using seasonal variations found from a climate table. The climate table can be derived from a location (geoposition) of the vehicle (vehicle location: location of the most frequently determined vehicle position). The usage pattern model thus provides a mapping of the calendar time information and the temperature profile information to the profiles of the load variables and/or the operating variables as an input variable and is also trained accordingly.

In addition, the usage pattern model 13 can be operated depending on the modeled state of health SOH. In this way, for example, in the case of a vehicle operated with the vehicle battery, it can be taken into account that a driver with an aged battery tends to have to charge it three times a week instead of only twice, as initially, in order to cover the desired route.

The usage patterns N are trained and specified for each vehicle individually and characterize the type of use and operation or the usage and operating behavior of the respective vehicle batteries 41.

The one or more usage parameters of the usage pattern N can thus implicitly indicate types of load on the vehicle battery 41 specific to the vehicle, in particular periodic loads depending on the calendar time specification. The training of the usage pattern model 13 can result in time series of load variables L for the usage parameters of the usage pattern N, which contain both weekly and seasonal periodicity effects.

Due to the possibility of predicting the modeled aging state SOH, a driver-specific aging state trajectory for a usage pattern N can be created. The usage pattern N can be derived from stress factors and/or can be learned on the basis of historical fleet data in a driver-specific and data-based manner. Preferably, autoregressive models or, alternatively, deep learning methods for pattern recognition are used for this purpose.

The following is based on the flow chart of the 5 and the functional block diagram of 4 the method for determining an aging state of a vehicle battery using a vehicle-external central unit 2 is explained in more detail. The method is preferably carried out entirely in the central unit 2 and is implemented there in the form of software and/or hardware. The procedure is described below with reference to the previous explanations.

In step S11, the operating variables F are transmitted to the central processing unit 2 at regular time intervals. The operating variables F are transmitted from a large number of vehicles 4 to the central unit 2 and temporarily stored there in a vehicle-specific manner.

In step S12 it is checked whether the time series to be analyzed have data gaps. A gap detection model 11 is used to check whether the time series are continuous or whether historical data for the operating variables F are missing for one or more specific time periods. The detection of a data gap by the gap detection model 11 in the time curves of the operating variables F can be done, for example, by monitoring the time series of a running operating variable, such as a time stamp or a consumption variable, such as a distance driven, for discontinuities or by detecting sudden changes in a state of charge or something else Operating characteristics are recognized. If a temporal data gap is determined (alternative: yes), the method continues with step S13. Otherwise (alternative: no), the method continues with step S11.

In step S13, the data gap is supplemented or filled by an artificially generated time profile of operating variables F', which result from the operating variable model 10, with the aid of a coupling unit 12. For this purpose, the operating variable model 10 receives a driver-specific time profile of load variables N (battery current and battery temperature), which are generated in the usage pattern model 13 .

As described above, the usage parameters N are determined by training the usage pattern model 13 with time profiles of operating variables in the past that are recorded using measurement technology and are each associated with a calendar time specification. The time curves of the load variables L for a vehicle battery then result from the learned usage pattern and the calendar time specification for the evaluation period/point in time for the assigned usage characteristics of the vehicle battery.

The operating variables are now supplied to a hybrid aging state model in step S14. As described above, the hybrid aging state model includes a physical aging model 5, a feature model 8, which is designed to determine operating features M from the curves of the operating variables F provided. The operating variables F provided include the operating variables F, F′, with data gaps being filled in as described above.

The modeled aging state SOH from the physical aging model 5 is corrected using a correction variable k from the data-based correction model 6 . The correction model 6 corresponds to the data-based model, which corrects the correction variable k for the aging state SOHph determined by the physical aging model in the correction block 7 depending on operating characteristics M that result from the time curves of the operating variables provided. The correction quantity k can be applied to the modeled aging state SOHph after the correction quantity k has been filtered. In particular, the correction variable can be low-pass filtered or smoothed, in particular with a PT1 or PT2 element.

A further observing state of health model 18 can be provided for verification of the usage pattern model 13, which evaluates the actually measured time curves of the operating variables F provided in order to determine a further state of health SOH′ based on the operating variables F. The further aging state model 18 can determine the further aging state as a reference or observer model, for example based on methods such as Coulomb counting and/or methods for determining a change in internal resistance, if corresponding operating states of the vehicle batteries 41 are identified using the time profiles of the operating variables F. Thus, for the plurality of vehicle batteries 41, data points each having an aging state value and an associated aging time of the relevant vehicle batteries 41, in particular all vehicle batteries 41 of the vehicle fleet, can be collected. After any Smoothing and cleaning up the data points, a aging state trajectory can be modeled using conventional regression methods over an aging point in time together with confidence intervals σ (standard deviations), which indicates a course of the further aging state SOH' for the large number of comparable vehicle batteries in the fleet. A course of such an aging state trajectory is shown for the aging state dependent on the change in internal resistance 6 shown. The confidence in the observed aging state model 18 is calculated empirically from the combination of the individual observations and taking into account the error propagation from the sensor (eg battery temperature sensor) to the calculated model value (eg observed SOHC).

The further aging status model 18 determined and made available in the central unit is evaluated in step S15 in order to obtain a further aging status SOH′ at the current aging time (age of the vehicle battery since commissioning).

By using the further aging state model 18, the aging state determined by the hybrid aging state model can be checked for plausibility.

In step S16, an aging time SOH can be provided for a vehicle battery considered in the hybrid aging state model, which is compared in a comparison block 19 with the trajectory point of the aging state trajectory at the current aging time under consideration. If it is determined in comparison block 16 that the aging state SOH modeled by the hybrid aging state model and the further aging state determined by the further aging state model 15 differ from one another by more than the confidence interval specified for the further aging state (alternative: yes), then in step S17 the usage pattern model 13 is adjusted accordingly, so that the reconstructed curves of the at least one load variable L match the independent observation of the further aging status model 18 in the best possible way. and returns to step S14. The hybrid state of health model can thus be used again to determine the state of health SOH based on new time profiles of the operating variables F provided, in which the data gaps have been determined based on the adapted usage pattern model 13 .

If, for example, the further aging state SOH′ indicates a lower aging state than the modeled aging state SOH, the one or more n usage parameters N can be changed in the direction of a lower load in order to generate the operating variables for filling the data gaps. This process can be carried out iteratively, in which the usage pattern parameters are changed incrementally in each iteration until the aging state SOH modeled by the hybrid aging state model and the further aging state determined by the further aging state model 15 differ from one another by no more than the confidence interval specified corresponding to the further aging state differ. A numerical optimization method such as the Bayesian optimization method can preferably be used for this purpose in order to be able to precisely reconstruct the reconstructed time curves with inherent stress factors. Alternatively, numerical optimization methods such as Regula Falsi (Pegasus method) can also be used.

If it is determined in step S16 that the modeled aging state SOH and the further aging state SOH 'do not differ by more than the amount specified by the confidence interval (alternative: no), then in step S18 in a merging block 20 the modeled aging state SOH and the further aging state SOH' merged with each other. This can be done, for example, by averaging, weighted averaging, or the like to provide a fused state of health SOH _Fus .

In particular, control-technical fusion methods such as particle filters or Kalman filters, preferably a dual extended Kalman filter, can be used in the fusion block 20 . In the merging block 20, two aging state trajectories, namely the trajectories of the modeled aging state and the further aging state since the device battery in question was put into operation, can be combined with the respective temporal confidence intervals, resulting in a fused trajectory of the aging state that is as reliable as possible with minimized uncertainty.

The method described above can be carried out in the central unit 2 in whole or in part.

Based on the merged trajectory of the aging state, the optimized operation of the energy storage device is possible, e.g. in order to be able to map a desired aging process.

Claims (15)

Computer-implemented method for modeling a current state of aging (SOH) or predicting an aging state (SOH) of an energy store (41) of a technical device, with the following steps: - providing or receiving (S11) a time profile of at least one operating variable (F) of at least one energy store (41); - Providing a data-based aging state model (9), in particular, which is trained to assign a modeled aging state (SOH) of the energy store (41) to the at least one operating variable (F) depending on the course over time, - Checking whether a temporal data gap in the temporal course of the at least one operating variable (F) of the energy store (41); - when determining the temporal data gap (S12), generating a time profile of the at least one operating variable (F) for the duration of the data gap using a usage pattern model (13), wherein the usage pattern model (13) is designed to depend on a usage pattern (N ) and a calendar time indication (Z) to provide a time profile of the at least one operating variable (F) for the temporal data gap and/or at least one load variable (L), from which at least one operating variable (F) can be derived, for the duration of the data gap ; - Determining (S13, S14) the current aging status (SOH) or predicting the aging status depending on the time profile of the at least one operating variable (F), which is supplemented by the generated time profile of the at least one operating variable (F) in the temporal data gap, and using the aging state model (9). procedure after claim 1 , wherein a data gap is determined when the device is inactive for at least a predetermined period of time and/or due to connectivity problems there is no information about the time profile for the at least one operating variable (F). procedure after claim 1 or 2 , wherein the usage pattern model (13) is designed as a data-based usage pattern model, which is designed as a recurrent neural network, in particular as a Bayesian LSTM, as a Gaussian process or as an attention layer, or as a hybrid usage pattern model with a non-data-based model, which, depending on an operating characteristic (M), provides a time profile of the at least one operating variable for the temporal data gap and/or the at least one load variable (L) for the duration depending on the calendar time specification (Z), and a data-based usage correction model (13) for correcting the course over time of the at least one operating variable (F) for the temporal data gap and/or the at least one load variable (L) for the duration as a function of the calendar time specification (Z). Procedure according to one of Claims 1 until 3 , wherein the usage pattern model (13) device-specifically based on the provided or received time profile of the at least one operating variable (F) is formed or trained. Procedure according to one of Claims 1 until 4 , wherein the usage pattern model (13) provides the time profile of the at least one load variable (L), with an operating variable model (10) being designed to generate a time profile of at least one load variable (L) from the time profile of the at least one load variable (L). , which is required on the input side for the aging state model (13). Procedure according to one of Claims 1 until 5 , wherein a further aging state model (18) is provided, which is designed to assign trajectory points, based on the time profile of the operating variables (F) by empirical methods such as Coulomb counting or measurement of an internal resistance change, which assign a further aging state (SOH') to an aging time , to be determined, wherein a state of health trajectory for the plurality of energy stores (41) is modeled together with a confidence band for the trajectory points, which indicates the accuracy of the estimation of each trajectory point, wherein the usage pattern model (13) is adapted if the determined state of health (SOH ') of the further aging state model (18) for a current or considered aging time is outside the confidence band for the current or considered aging time. Procedure according to one of Claims 1 until 6 , wherein a further aging state model (18) is provided, which is designed to assign data points, which assign a further aging state (SOH') to an aging time, based on time profiles of the operating variables (F) by empirical methods such as Coulomb counting or measurement of a change in internal resistance, to determine in order to model an aging state trajectory for the plurality of energy stores (41), the modeled aging state by merging the aging state (SOH) of the aging state model (9) for a current or considered predicted aging point in time depending on the determined aging state of the aging state model and another Aging state (SOH ') of the further aging state model (18) is determined. Procedure according to one of Claims 1 until 7 , where the operating variables (F) a current (I), a Voltage (U), a temperature (T) and a state of charge (SOC) correspond to a device battery as an energy store (41). Procedure according to one of Claims 1 until 8th , wherein the data-based aging state model (9) is designed as a hybrid model and a physical aging model (5), which is based on electrochemical model equations and is designed to output a physical aging state (SOHph), and a trainable data-based correction model (6), in particular in Form of a regression model, preferably a Gaussian process, wherein the correction model (6) is trained to correct the physical state of health (SOHph) and to provide the corrected physical state of health as the modeled state of health (SOH), in particular with quantified uncertainty. procedure after claim 9 , the data-based aging state model being trained based on training data sets, the training data sets being divided into a training set and an extended, total training set, the aging model being parameterized with the training set, and the correction model (6) being trained based on the total training set, wherein the data-based state of health model (9) is tested based on the entire training set in order to determine the validity of the data-based state of health model. Procedure according to one of Claims 1 until 10 , wherein the device battery (41) is operated depending on the course of the predicted modeled aging state (SOH), in particular a remaining service life of the energy store (41) depending on the course of the predicted modeled aging state (SOH) is signaled. Procedure according to one of Claims 1 until 11 , wherein the energy store (41) is used to operate a device such as a motor vehicle, a pedelec, an aircraft, in particular a drone, a machine tool, a consumer electronics device such as a mobile phone, an autonomous robot and/or a household appliance. Device for modeling a current state of health or predicting a state of health (SOH) of an energy store of a technical device, the device being designed for: - Receiving a time profile of at least one operating variable of at least one energy store; - Providing a data-based aging state model in particular, which is trained to assign a modeled aging state of the energy store to the at least one operating variable depending on the course over time, - checking whether there is a time data gap in the time profile of the at least one operating variable of the energy store; - when determining the temporal data gap, generating a time profile of the at least one operating variable for the duration of the data gap using a usage pattern model, wherein the usage pattern model is designed to generate a time profile of the at least one operating variable for the temporal data gap and/or depending on a usage pattern provide at least one load variable, from which at least one operating variable can be derived, for the temporal data gap; - Determining the current aging status or predicting the aging status depending on the chronological progression of the at least one operating variable, which is supplemented by the generated chronological progression of the at least one operating variable in the temporal data gap, and using the aging status model. Computer program product comprising instructions which, when the program is executed, cause at least one data processing device to carry out the steps of the method according to one of the Claims 1 until 12 to execute. Machine-readable storage medium, comprising instructions which, when executed by at least one data processing device, cause the latter to carry out the steps of the method according to one of Claims 1 until 12 to execute.

Images