Method and device for providing a supportable charging distribution by a device battery of a device operated for the battery

Abstract

A method for determining an optimized charging distribution used when charging a battery pack of similar equipment comprises the following steps of: distributing different test charging distributions for the battery pack of the equipment, wherein the different test charging distributions respectively specify a maximum allowable charging current related to a system state; selecting, corresponding to the Pareto curve, a test charge profile of a combination of a modeled aging state change in which a specified cost is minimized and a modeled average / aggregate charge time resulting from the application of a related test charge profile; operating the device battery pack for a predetermined observation duration, and charging the device battery pack a plurality of times on the basis of the respectively allocated test charge profiles; then determining average or aggregated charging time information and aging state change for each equipment battery pack; training a data-based probabilistic charging distribution model with a training dataset that takes into account the average charging time and the assigned aging state change for each device battery pack; determining optimized charging distribution according to a Pareto curve and a charging distribution model; it is provided to control the charging process of all device battery packs.

Description

Technical Field

[0001] The invention relates to an equipment battery, in particular a vehicle battery of an electrically driven vehicle, and in particular to adapting or selecting a charge profile for such an equipment battery. Background Art

[0002] Charging and discharging a device battery causes degradation. The degradation measure depends on the charging and discharging parameters. While the degradation caused by the discharging process is essentially determined by the use of the device battery, which cannot usually be influenced, the stress applied to the device battery during the automated charging process is determined by the parameters that specify the type and manner of the charging process. The parameters of the charging process thus determine the aging behavior of the device battery.

[0003] The charging of a device battery is usually carried out on the basis of a predefined charging profile. The charging profile is usually specified by a characteristic curve which determines the dependency of the charging current or the maximum charging current on the state of charge. The state of charge specifies the ratio of the stored charge to the maximum charge in %. The charging profile is set in such a way that the charging time, the maximum charging current and the accepted aging behavior can be optimized, so that a compromise can be achieved between the load on the device battery and the required charging time.

[0004] The charge profile of the device battery is usually determined before production begins based on large-scale laboratory measurements, wherein the charge profile is created in such a way that the dependency of the degradation on the charge current, battery temperature and charge state of the device battery can be taken into account.

[0005] Since each user behavior loads the device battery differently, the aging state is determined not only by calendar aging, but also by cyclic loading due to the charge and discharge current profile, the charge state, and the battery temperature. However, since the charging profiles are currently only predefined for normal and fast charging for specific device battery types, the user can only influence the impact of the charging process on the aging behavior of the device battery to a limited extent. Summary of the invention

[0006] According to the invention, a method for determining an optimized charging profile for a battery pack of similar devices according to claim 1 and a corresponding device according to the independent parallel claim are provided.

[0007] Further refinements are specified in the dependent claims.

[0008] According to a first aspect, a method is provided for determining an optimized charging profile for use when charging a battery of a device of the same type as a battery-operated device, in particular an at least partially computer-implemented method, the method comprising the following steps:

[0009] - assigning a plurality of different test charge profiles to the device battery pack, which test charge profiles each specify a maximum permissible charging current as a function of time or a system state (e.g., charge state) or a state (e.g., temperature), wherein the test charge profiles are selected in accordance with a Pareto curve, wherein the test charge profiles specify those combinations of modeled average charging times and modeled aging state changes that result when the respective test charge profiles are applied, in which the costs according to a predefined cost function are minimized;

[0010] - operating the device battery for a predetermined monitoring period, wherein the device battery is respectively charged several times based on a respectively assigned test charge profile and, if necessary, discharged during operation;

[0011] - after the observation period, determining average or aggregated charging time information and aging state changes for each device battery pack therein;

[0012] - training a data-based probabilistic charging distribution model using a training dataset that takes into account average / aggregate charging time information and respectively assigned aging state changes for each device battery group;

[0013] -Determine an optimized charging profile based on the Pareto curve and the charging profile model;

[0014] - Providing the determined optimized charging profile to control the charging process for all device batteries.

[0015] The aging of device batteries depends largely on the load caused by their type of use, such as the magnitude of the charging and discharging currents, high temperatures, the frequency of rapid charging cycles, etc. Therefore, compliance with the guarantee conditions depends on the type of use of the device batteries, so that the goal is to keep the load on the device batteries as low as possible during use.

[0016] In the above-described method, the load on the device battery is reduced by adapting a predefined charging profile. The charging profile is usually determined on the basis of tests by the battery manufacturer and represents an average load on the device battery, by which the guarantee conditions can be observed in addition to other operating conditions. One or more charging profiles are respectively fixedly predefined for the device battery, which are selected depending on the requirements, for example normal charging or fast charging. The selected charging profile is usually applied to all device batteries of the same type during the charging process.

[0017] The usual charging profile can be determined based on a predetermined limit value of the anode overpotential (Anodenüberpotenzial), and specifies a maximum charging current related to the system or charging state, i.e., the filling or charging time of the device battery pack relative to the total storage capacity or the battery pack temperature. The variation of the maximum charging current or the target charging current as a function represents the charging curve. Such a charging profile limits the maximum charging current and thus avoids excessive stress loads on the identification battery pack by avoiding the anode overpotential below the predetermined limit value. The anode overpotential is the difference between the actual potential of the anode and its thermodynamic equilibrium potential. It is equivalent to the energy that must be applied to initiate an electrochemical reaction at the anode. The charging profile can be created by simulation under respectively predetermined charging states and predetermined safety values ​​or limit values ​​of the anode overpotential by means of modeling based on a predetermined electrochemical battery pack model, see also EP4046228A1.

[0018] By adapting the predefined charging profiles of the individual device batteries, the load on the device batteries is reduced, in this way it is possible to directly influence the further course of the aging of the device batteries concerned, for example by avoiding critical states. However, by predefining the charging profile, the charging time that the user must wait for to reach a predetermined charge state or a predetermined charging cycle is predetermined. In the case of a charging profile with a low maximum charging current (which leads to low aging), the charging time is longer, while in the case of a charging profile with a higher charging current (which leads to high aging), shorter charging times can be achieved.

[0019] The above method enables the selection of an optimized charging profile taking into account the evaluation of the charging profile of the device battery pack, the model points of the aging state change and the charging time information, with the help of which the conflict of objectives between reducing the charging time and maximizing the service life (reducing aging due to the charging process) can be determined. For this purpose, different charging strategies / charging profiles are evaluated across the devices with regard to aging effects and charging duration, and the optimal charging profile is selected model-based according to the charging profile model. Thus, the optimization can be carried out based on the real battery pack behavior via swarm intelligence, i.e. a large number of test charging profiles tested in field operation.

[0020] The background is that the actual situation / costs regarding the relationship between aging behavior and charging time may differ from the Pareto optimal relationship between aging behavior and charging time, for example due to series deviations, model limitations and unforeseen influences in the field. Depending on the charging profile for the actual device battery pack, better cost values ​​may result in terms of costs than the ideal Pareto optimal cost curve, or conversely, worse cost values.

[0021] It may be provided that a plurality of different test charge profiles are provided as a function of a parameterized electrochemical battery model of the device battery, in particular as a function of one or more electrochemical or physical parameters and one or more dimensioning parameters, such as a limiting value for the anode overpotential.

[0022] The charging curve of the test charging profile can be determined by simulation with the aid of an electrochemical battery model based on electrochemical and / or physical variables, such as the anode potential or the anode overpotential, and selected based on a Pareto optimal relationship between average / aggregate charging time information and aging state changes. The Pareto optimal relationship corresponds to the following Pareto curve, which can be determined by simulation based on a predetermined cost function, wherein the predetermined cost function takes into account the charging time and the aging state changes. The charging curve / charging profile is generated in a manner known per se so that the anode overpotential does not fall below the anode overpotential limit value specific to the battery type.

[0023] This can be done as a function of the aging state of the device battery, in particular as a function of one or more electrochemical or physical parameters, such as equilibrium or kinetic parameters or electrochemical states, such as cyclizable lithium, such as equilibrium diffusion coefficients, etc., for example by adapting the limiting value of the anode overpotential to an optimized value of the anode overpotential, wherein dimensioning can be done by taking into account the conflicting objectives between minimized aging state changes and minimized average or aggregated charging time information. In the process, a plurality of Pareto optimal charging profiles or configurations of such charging profiles can be derived by predefining different limiting values ​​for the anode overpotential with the aid of control methods (e.g. simulation or control to limiting values).

[0024] For example, the determination of an appropriate charge profile based on a battery pack model is known from the prior art, for example from US 9,153,991 B2, US2017 / 033866 A1, DE 102020206272 A1, DE 102020210132A1, DE102019209165A1, DE 102020212579A1, US10,447,054B2, US 9,153,991 B2, US10,312,699B2, DE 102019216015A1 and "Doyle, M, Fuller, TF and Newman, J. "Modeling of galvanostatic charge and discharge of the lithium / polymer / insertion cell", USA, https: / / doi.org / 10.1149 / 1.2221597".

[0025] According to the method, firstly different test charge profiles from the predetermined test charge profiles are assigned to a plurality of device batteries. When one or more test charge profiles are respectively assigned to the device batteries, the test charge profiles can be assigned to a plurality of the device batteries. This achieves that different usage behaviors of the device batteries, which lead to different aging behaviors, i.e., changes in the aging state, during the observation period have a smaller influence on the evaluation based on the different charge profiles due to the probabilistic method. The test charge profiles are selected from the Pareto optimal charge profiles for various predetermined limit values ​​of the anode overpotential.

[0026] Based on the correlation between the average / aggregate charging time information and the average aging state change (per unit time), the charging distribution is selected in a Pareto optimal manner, that is, selected from the following charging distribution set, for which a reduction in the average / aggregate charging time information while the aging behavior remains the same or a deterioration of the aging state while the average / aggregate charging time information remains the same is not possible.

[0027] For example, a number of 30 charging profiles may be selected in a discrete manner, wherein a test charging profile is respectively assigned to a device.

[0028] In ongoing operation, the charge profile is now applied to the respective device battery during the charging process. In this case, it is evaluated how the applied test charge profile influences the aging behavior of the device battery in question over the observation period.

[0029] It can be provided that the aging state change can be determined as the difference between the aging state at the beginning and at the end of the set observation period. Alternatively, instead of the entire aging state change deltaSOH within the observation period, only the aging state change accumulated during the charging process (preferably over a predetermined, constant number of charging processes) can also be determined, i.e., it can be determined as the sum of the reduction in the capacity-related aging state SOH-C during the charging process alone (for example by forming the difference in the aging state before and after each charging process).

[0030] Furthermore, the average / aggregate charging time information is determined as the average value of the charging time of the following charging processes, during which a change in the state of charge is greater than a predetermined threshold value. It is also possible to determine the change in the aging state by eliminating or eliminating or subtracting in a model-based manner aging effects or changes in the aging state that are not caused by the charging process from the overall change in the aging state determined within the observation duration. Cumulative or aggregated charging time information can also be used. For example, the historical charging time to date can be accumulated. In addition, it can be provided that for the cumulative charging time, only a representative filtered portion (Ausschnitt) is taken into account, that is, for example, the charging time between SOC_min=20% and SOC_max=80%. Filtering can therefore be performed to increase comparability or representativeness / relevance (Relevanz).

[0031] Thus, a change in the state of aging deltaSOH of the device battery can be determined after a predetermined calendar period after the start of the observation. It is assumed here that all device batteries have the same SoH value at the beginning of the duration, i.e. the same state of aging, for example a SoH value of 100%. In addition, average / aggregate charging time information is determined depending on the charging profile. The charging time can be taken into account in the case of each charging process or only in the case of charging processes in which charging is carried out via a specific charging stroke, for example a specific charging stroke with a delta SOH>80%.

[0032] Furthermore, average / aggregate charging time information may be specified herein in an absolute manner or in a manner relative to a change in state of charge at a particular state of charge level.

[0033] The current aging state can be determined using a conventional and known model for determining the aging state. In particular, an electrochemical aging state model can be used, which is substantially based on an electrochemical battery model. This electrochemical battery model may include a set of differential equations, which are based on differential equations parameterized via model parameters, and model the internal battery state, in particular the equilibrium state and possible kinetic state, by means of a time integration method, and provide a relationship between the battery current, battery voltage, battery temperature and charge state of the device battery and the internal battery state. This electrochemical battery model is known, for example, from the documents US2016 / 023,566, US2016 / 023,567 and US 2020 / 150,185. The aging state can be derived from these internal battery states.

[0034] For the above determination of the aging state change and the charging profile, an initial parameterized aging state model and an initial parameterized battery pack model are used, which can then be gradually improved

[0035] Then, a cross-device evaluation can be performed in a central unit remote from the devices with the goal of evaluating the average / aggregate charging time information and the real state point of aging state changes for the device battery pack by testing the charging profile in view of its original design aspects. The background is the assumption that not all aging effects can be accounted for by the charging profile used by the simulation to create the charging profile.

[0036] For this purpose, a charging profile model is first predetermined, which allocates the aging state change and the average / aggregate charging time information to each other. For this purpose, a cost function can be provided, which calculates a cost value based on the charging profile and the aging state change, such as a weighted sum of the inverse of the aging state change and the charging time.

[0037] For each device battery pack, there is now a state point consisting of the aging state change and average / aggregate charge time information, making it possible to model a charge distribution model designed as a probabilistic regression model on the average / aggregate charge time information (perhaps per percent charge state change) and the aging state change. This enables the solution space to be estimated The cost of using a specific charging profile for a specific device battery pack in the probabilistic charging distribution model. For example, a Gaussian process model with, for example, an exponential square kernel can be used, which can map aging state changes to average / aggregate charging time information. In addition, other characteristics of the device battery pack, such as electrochemical design characteristics, in particular known battery pack state variables, can be used or considered in the Gaussian process model. Therefore, the probabilistic charging distribution model at least maps aging state changes to average / aggregate charging time information and thus to corresponding cost values.

[0038] Alternatively, if the Pareto optimal state point consisting of aging state change and average / aggregate charging time information at the minimum cost value is known for the device battery type according to the Pareto curve, the probabilistic charging distribution model can be trained as a correction model.

[0039] With the aid of the probabilistic charge distribution model, in addition to the cost value, a confidence factor can also be estimated which specifies the reliability of the probabilistic charge distribution model at the evaluation point.

[0040] A cost-optimized selection of one or more Pareto optimal charging profiles from the test charging profiles can now be performed by evaluating the probabilistic charging profile model.

[0041] The probabilistic charging distribution model can be analyzed below by varying the aging state variation and determining the charging time for which the cost minimum of the resulting cost values ​​is obtained. The charging profile can then be assigned as an optimized charging profile, such that it is derived from the charging distribution model according to the modeled state point found for the cost minimum.

[0042] For example, the optimized charging profile can be determined by determining the nearest Pareto state point on the Pareto curve obtained by simulation (making the vertical line fall on the Pareto curve) from the modeled state point for which the minimum cost is obtained according to the cost function according to the charging distribution model. The Pareto state point is determined by the simulated aging state change and the simulated charging time, and the Pareto state point enables the value of the anode overpotential that best corresponds to the modeled state point or the limit value of the anode overpotential to be calculated by interpolation based on the anode overpotential of the Pareto curve or the adjacent and known limit values ​​of the anode overpotential of the Pareto curve. In particular, the anode overpotential can be determined by approximation. For example, the approximation can be performed iteratively. Then, the calculated anode overpotential can be used to subsequently determine the charging profile according to the above method. The charging profile found in this way can be assumed to be an optimized charging profile and allocated to a portion or all of the device battery packs.

[0043] In particular, it can be provided that the optimized charging profile is determined by determining a limit value for the anode overpotential for the modeled state point found for the cost minimum, wherein the limit value is obtained by interpolation at the charging profile closest to the state point found for the cost minimum on the Pareto curve.

[0044] With the help of a predefined charging profile model, an optimized charging profile can be determined. Together with the cost function, the charging profile model reflects that the higher the aging behavior or the greater the variation in the aging state and the greater the average / aggregate charging time information, the higher the cost assessed for the user. The greater the decline in the aging state means: the greater the decline in the residual value of the vehicle, while the longer the charging time means that the user has to wait longer during the charging process and thus loses comfort. With the help of the charging profile model, the actual cost can be assessed and provided for each device according to the cost function.

[0045] The advantage of probabilistic modeling, in addition to the quantification of uncertainty and the associated quantile modeling, can also be the evaluation of a range of charging profile models that have not been evaluated or observed so far. Cost estimates can be explicitly assigned to each possible charging profile through a cost function that includes a confidence assessment.

[0046] According to the specific cost value for the corresponding modeled state point consisting of the aging state change and the average / aggregate charging time information obtained by the charging distribution model, one or more Pareto state points are selected for which the charging distribution can be determined in the manner specified above. The selection can be based on minimizing the cost value, taking into account the relevant uncertainty, i.e. the confidence level, if necessary.

[0047] Provision may be made for determining average / aggregate charging time information and changes in the aging state during the observation duration, determining the cost value, selecting one of the test charging profiles and providing the selected test charging profile or the created new charging profile to be performed in a central unit remote from the device.

[0048] The charging profile model can be improved stepwise at predetermined or regular time intervals in that an aging state model for determining the aging state and an electrochemical battery model for determining the charging profile based on device batteries for which the uncertainty resulting in the charging profile model is too great and / or the cost values ​​are too high or deviate too greatly from the ideal course of the Pareto curve. Too high cost values ​​correspond to cost values ​​that are greater than the cost values ​​on which the predetermined ideal Pareto curve is based.

[0049] For this purpose, accelerated aging tests and performance tests can be carried out on a test bench with the relevant device batteries in order to use the measurement results (measurement of the operating points of the battery voltage, battery current, battery temperature and the state of charge in static and dynamic operating situations or determination of the aging state, for example by coulomb counting) to improve the aging state model and the electrochemical battery model for determining the aging state in a manner known per se by training or parameterization, wherein for the relevant device batteries, the values ​​obtained in the charge profile model that are too high or deviate too greatly from the Pareto curve are

[0050] Cost values ​​and / or confidence intervals that are too wide. In particular, the electrochemical battery model can be updated or reparameterized by fitting. This can correspond to an active learning algorithm.

[0051] Based on the aging state model, the charging time, aging state change and assigned test charging distribution for each device battery group are re-determined using the previously recorded operating variable change process of the device battery group with the assigned test charging distribution, and an updated probabilistic charging distribution model is created, which assigns average / aggregate charging time information to the corresponding aging state change. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The embodiments are explained in more detail below with reference to the accompanying drawings, wherein:

[0053] Figure 1 A schematic diagram of a system for providing driver and vehicle specific operating variables so that the effects of different charging profiles of a vehicle battery pack can be observed separately outside the vehicle in a central unit;

[0054] Figure 2 A diagram showing an exemplary charging profile for performing a vehicle battery pack charging process;

[0055] Figure 3 A flow chart illustrating a method for determining an optimal charging profile for vehicles in a fleet is shown;

[0056] Figure 4 A diagram showing the evolution of the correlation between the average charging time and the aging state change for different charging profiles;

[0057] Figure 5 A diagram showing measurement results of the charging time of a vehicle battery for which different charging profiles are predefined, and a corresponding change in the aging state; and

[0058] Figure 6 A diagram of a modeled cost function as a basis for selecting an optimized charging profile is shown. DETAILED DESCRIPTION

[0059] The method according to the invention is described below based on a vehicle battery as a device battery in a large number of motor vehicles as similar devices. The method can be partially executed in a central unit and is used to select a suitable charging profile for the vehicle battery.

[0060] The above examples represent a large number of fixed or mobile devices with an energy supply independent of the power grid, such as vehicles (electric vehicles, scooters, etc.), facilities, machine tools, household appliances, IoT devices, etc., which are connected to a central unit (cloud) outside the device via corresponding communication connections (e.g. LAN, Internet).

[0061] Figure 1 A system 1 for collecting fleet data in a central unit 2 for selecting a favourable charging profile is shown. Figure 1 A fleet 3 with a plurality of motor vehicles 4 is shown.

[0062] Figure 1 4 shows one of the motor vehicles 4 in more detail. These motor vehicles 4 each have a vehicle battery pack 41 as a rechargeable electrical energy storage, 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 the central unit 2 (the so-called cloud).

[0063] The motor vehicle 4 monitors the operation of the vehicle battery 41 by means of a control unit 43. In this case, the aging state change and the average / aggregate charging time information of the charging process are determined using the respectively assigned charging profile and the aging state change and the average / aggregate charging time information are determined to be transmitted to the central unit 2 at a predetermined time.

[0064] The motor vehicle 4 further transmits operating variables F to the central unit 2, which specify at least variables that influence the aging state of the vehicle battery 41. In the case of a vehicle battery, the operating variables F can be time series of the battery current, battery voltage, battery temperature and state of charge (SOC: State of Charge), both at the pack level, module level and / or at the cell level. The operating variables F are recorded in a fast time grid of 0.1 Hz to 100 Hz and can be regularly transmitted to the central unit 2 in uncompressed and / or compressed form.

[0065] The central unit 2 has a data processing unit 21 , in which the method described below can be executed, and a database 22 for storing data points, model parameters, states, and the like.

[0066] An algorithm may be implemented in the central unit 2, which algorithm selects one or more suitable charging profiles from a set of test charging profiles for use by the vehicle 4. The algorithm may be executed regularly to select a suitable charging profile.

[0067] The charging profile of the vehicle battery 41 predetermines a maximum charging current in each case as a function of the state of charge. A typical charging profile has a maximum charging current that decreases as the state of charge increases, such as Figure 2 , and each specifies a maximum charging current depending on the state of charge, ie the stored available charge of the vehicle battery 41 or during the charging time. The charging profile is predetermined and determined in such a way that the anode overpotential does not fall below a predetermined limit value for the anode overpotential.

[0068] Figure 3 A method for determining a charge profile to be distributed to the vehicle battery 41 is explained using a flow chart. The method can be performed in conjunction with the central unit 2.

[0069] The method for determining an optimized charging profile for a large number of vehicle batteries 41 first provides in step S1 that an electrochemical battery model is parameterized based on the received operating variable curve F of the vehicle battery 41, in particular by fitting. Such an electrochemical battery model can include a set of differential equations, which model internal battery states, in particular equilibrium states and possible kinetic states, based on differential equations parameterized by model parameters by means of a time integration method, and provide relationships between the operating variable curves of the device battery, i.e., the battery current, battery voltage, battery temperature and state of charge of the device battery and the internal battery state. Such electrochemical battery models are known, for example, from the documents US 2016 / 023,566, US 2016 / 023,567 and US 2020 / 150,185.

[0070] Then, in step S2, the electrochemical battery model enables, based on electrochemical and / or physical parameters and one or more dimension parameters, such as the limit value of the anode overpotential, to derive at least one test charge profile in the target conflict between the influence of aging and the required average / aggregate charge time information as a Pareto-optimized state point (Pareto state point) consisting of average / aggregate charge time information and aging state change, and to provide it as a charging curve cluster (Ladekurvenschar) or as a charging profile configuration. The determination of this curve cluster from possible test charge profiles can be carried out in a manner known per se, taking into account the anode overpotential resulting from the battery chemistry in the central unit 2 and the predetermined dimension parameters, while the average / aggregate charge time information is predetermined. The Pareto curve is obtained by minimizing a cost function, wherein the cost function provides a cost value that depends on the aging state change and the average / aggregate charge time information.

[0071] In practice, the average / aggregate charging time information may generally correspond to the average value of the charging time until a predetermined minimum charging stroke, for example delta SOC=80%, is reached, and may be normalized for sufficiently high state of charge strokes and, for example, may be relative to a fully charged equivalent (Full-Charge- ) is designated.

[0072] The change in the state of aging caused by the application of a corresponding charging profile to the vehicle battery 41 can now be determined by simulation during a set observation period. A correlation is derived between the average / aggregate charging time information (or the average value of the charging time per change in the state of charge) and the change in the state of aging after a predetermined simulated operating time, for example between 1 and 3 years (observation period).

[0073] The aging state change can be determined as the difference between the aging state at the start of the simulation of the vehicle battery pack 41 with the test charge profile and at the end of the simulation of the vehicle battery pack 41 with the test charge profile. Alternatively, the aging state change can be determined as the sum of the differences between the aging state at the start of the charging process with the corresponding test charge profile and at the end of the charging process during the predetermined observation duration. In an ideal case, a correlation is derived between the average / aggregate charging time information T and the aging state SOH or the aging state change, for example, Figure 4 As shown in . By optimizing according to a predetermined cost function, a Pareto curve can be determined. Different effects of the charging profile on the aging of the vehicle battery 41 can be identified.

[0074] In the subsequent step S3, a plurality of test charge profiles are selected from the test charge profiles and are respectively assigned to one or more electric vehicles 4 in the fleet 3 as an operating strategy for charging. This assignment can be performed by uploading the charge profiles to the control unit 43 of the relevant vehicle 4 and applying them there to each charging process. For example, 30 test charge profiles can be selected, which are respectively applied to 70 vehicles to obtain a sufficient data pool. A test charge profile for normal charging and a test charge profile for fast charging can be provided to each vehicle 3.

[0075] In step S4, each vehicle 4 or each vehicle battery 41 is operated on site and monitored over an observation period of, for example, between 6 months and 3 years, and it is determined how the application of the charging profile influences the aging behavior of the respective vehicle battery 41. For this purpose, the course of the operating variable changes is continuously recorded and the aging state is determined with the aid of an aging state model which is evaluated based on the time series of the operating variables. The influence is evaluated by the change in the aging state during the observation period and by the assigned average / aggregate charging time information caused by the test charging profile assigned to the relevant vehicle battery 41. For this purpose, the change in the aging state (in general or only during the charging process) is determined over a predefined observation period based on the aging state at commissioning, for example SoH=100%. The average / aggregate charging time information and the change in the aging state are transmitted to the central unit 2 after the observation period.

[0076] Based on the different charging profiles assigned to the vehicle, the course of the average / aggregate charging time information T about the aging state SOH after the observation duration is derived, for example: Figure 5 It can be identified that there are systematic ranges or patterns in which the relationship between the average / aggregate charging time information and the aging state SOH after the observation duration is over- or under-performing (Over-bzw. Underperformance) relative to the theoretical optimum of the above Pareto curve P.

[0077] Figure 5 Each individual real state point F in the graph of represents the charge state change and average / aggregate charging time information of a specific vehicle battery pack 41 at the current point in time or at the end of the observation duration.

[0078] With the aid of a predefined cost function, the efficiency of the respectively assigned test charging profile can be evaluated for each vehicle. The cost function specifies the cost of using the relevant test charging profile as a function of the aging state or the change in the aging state and the average / aggregate charging time information T, in particular such that the greater the change in the aging state and the greater the average / aggregate charging time information during the observation duration, the greater the cost value for the test charging profile for the relevant vehicle battery pack 41. For example, such a cost function can be given by Figure 6 The cost function therefore takes into account the user costs due to the long charging or waiting time at the charging station and the cost of the residual value loss (Restwertverlust) of the vehicle battery pack 41 due to the increase in the aging state change.

[0079] The average / aggregate charging time information related to the aging state change for all vehicle battery packs can now be modeled in step S5 with the aid of a data-based probabilistic charging distribution model, which is designed as a probabilistic regression model. To this end, the average / aggregate charging time information T is assigned to the aging state change dSOH as a label to obtain a training data set and a validation data set. The training data set and the validation data set are now used to train the probabilistic charging distribution model. This enables a global formula to obtain a true continuous assignment of the aging state change dSOH and the average / aggregate charging time information.

[0080] The probabilistic charge distribution model can be designed as a Gaussian process model, for example. For example, the Gaussian process model can be modeled with the aid of an exponential quadratischen kernel.

[0081] By evaluating the trained Gaussian process model in conjunction with the cost function, a minimum cost value and a confidence level can now be determined for the course of the charge profile model in step S6. The confidence level can specify how informative a charge profile previously selected from the family of test charge profiles is with respect to the underlying cost function.

[0082] Based on the probabilistic charging distribution model specified by the modeling state points of the curve L in its own variation process, the modeling state point with the minimum cost can be determined in this way by evaluating the cost value of the modeling state point for the variation process of the charging distribution model.

[0083] according to Figure 5This is possible according to the schematic diagram in which average / aggregate charging time information and aging state change can be assigned to each charging profile. Thus, a cost including an uncertainty measure for the respective cost estimate can be explicitly assigned to each charging profile with the aid of the charging profile model. It is thus possible to select a charging profile that can be characterized via aging state change and average / aggregate charging time information. The selection is made according to the evaluation of the minimum cost value and the uncertainty measure that can be assigned to the cost estimate.

[0084] The selection can be made based on a cost mean value and / or can be supplemented by determining a predetermined quantile of the cost value. For example, the 1% quantile or the 50% quantile can be used depending on the risk tendency.

[0085] The determination of the optimized charging profile suitable for the vehicle battery pack can be performed in step S7 by projecting a specific modeled state point from the charging profile model onto the nearest Pareto state point of the Pareto curve, for example by making a vertical line of the modeled state point fall on ( des Lots) Pareto curve, thereby generating a corresponding Pareto state point on the Pareto curve. The Pareto state point can then be determined with the assigned anode overpotential by corresponding interpolation between the tested charge profiles, thereby determining an optimized limit value for the anode overpotential, which can be assigned to a parameterization or control technology simulation of an electrochemical battery model with the following optimized limit value for the anode overpotential, wherein in the above-mentioned manner, a charge profile can be derived as an optimized charge profile based on the optimized limit value.

[0086] As an alternative or in addition to the limit value for the anode overpotential, further electrochemical parameters or states which can be assigned to the operating strategy, such as temperature or derating limits, can be used to derive the optimized charging profile.

[0087] For example, a new optimized charging profile can be created by changing one or more dimensional parameters, such as different predetermined limit values ​​for the anode overpotential, to simulate aging state changes and average / aggregate charging time information, wherein the aging state changes and average / aggregate charging time information define a Pareto state point that is closest to a specific modeled state point in the charging distribution model.

[0088] In a subsequent step S8 , the charging profile optimized in this way is transmitted to the vehicles of the fleet 3 , so that it can thus be used to carry out a charging process.

[0089] In step S9 , a check can be performed after a predetermined time period. To this end, the charging profile can be improved stepwise at predetermined or regular time intervals by retraining the aging state model for determining the aging state and the electrochemical battery model for determining the charging profile based on the operating variable data of the specific vehicle battery 41 .

[0090] Specific vehicle battery packs 41 for further training are identified in the following manner: one or more modeling state points with excessively high uncertainty of the charging distribution model, i.e., uncertainty expressed, for example, via a confidence interval, are identified in the charging distribution model via a predetermined threshold value and the following vehicle battery packs are correspondingly determined whose actual state points are closest to the one / multiple identified modeling state points.

[0091] Alternatively or additionally, modeling state points may be determined, the degree to which the cost values ​​of the cost function deviate too much from the cost values ​​of the Pareto curve under the same state of charge change or the same average / aggregate charging time information, that is, the following modeling state points are selected, the difference in number of modeling state points between the cost value of the relevant modeling state point and the cost value of the Pareto state point under the same state of charge change or the same average / aggregate charging time information ( This represents an active learning process, which makes it possible to identify in a model-based manner those vehicle batteries that should be fed to a precise measurement, for example in a laboratory, test bench or workshop.

[0092] Subsequently, in step S10, the identified operating variable curves of the vehicle battery can be used to train the aging state model and the electrochemical battery model. In addition, accelerated aging tests and performance tests can be performed on the test bench for one or more of the vehicle batteries determined in step S9 for which an excessively high cost value and / or an excessively high confidence interval is obtained in the charge profile model, so as to improve the aging state model for determining the aging state and the electrochemical battery model by training or parameterization in a manner known per se using these measurements (measurement of the battery voltage, battery current, battery temperature and operating point of the state of charge in static and dynamic operation or determination of the aging state, for example, by coulomb counting). In particular, the electrochemical battery model can be updated or reparameterized by fitting.

[0093] Based on the aging state model, in step S11, the aging state or aging state change for each vehicle battery pack and the assigned test charging distribution are re-determined using the previously recorded operating variable curves of the vehicle battery packs with the test charging distribution, and an updated probabilistic charging distribution model is created in the above manner, which includes the average / aggregate charging time information and the corresponding aging state or aging state change.

[0094] The method can be carried out cyclically in each case after a predefined period of time.

[0095] In this way, a plausibility check can be performed in a "back-test", i.e. the Pareto-optimal charging profile is simulated again using the improved aging state model and the electrochemical battery pack model. Furthermore, with the model adapted in this way, an optimized charging profile can be determined according to the method specified above. This can be used to determine and predict the aging state of all vehicle batteries. This makes it possible to compare reality with simulation in order to validate the new model.

Claims

1. A method for determining an optimized charging profile for use when charging a battery pack (41) of a device (4) operated with a battery pack, in particular an at least partially computer-implemented method, comprising the following steps: - assigning (S3) a plurality of different test charge profiles to the device battery pack (41), the test charge profiles respectively specifying a maximum permissible charging current associated with the system state, wherein the test charge profiles are selected in accordance with a Pareto curve, wherein the test charge profiles specify a combination of the modeled average / aggregate charging time and the modeled aging state changes resulting when the relevant test charge profiles are applied, in which combination the cost according to a predetermined cost function for each modeled aging state change is minimized; - operating (S4) the device battery (41) for a predetermined observation duration, wherein: charging the device battery packs (41) a plurality of times based on the respectively assigned test charging profiles; - determining average or aggregated charging time information and aging state change for each device battery pack (41) therein after the observation duration; - training (S5) a data-based probabilistic charging distribution model using a training data set, wherein the training data set takes into account, for each device battery pack (41), an average charging time and a respectively assigned aging state change; - determining (S7) an optimized charging profile based on the Pareto curve and the charging profile model; - providing (S8) the determined optimized charging profile to control a charging process for said device battery (41).

2. The method according to claim 1, wherein: With the aid of an electrochemical battery model of the device battery (41) which is predetermined for similar device batteries (41), a plurality of different test charge profiles are provided in the following manner: different limit values ​​of the anode overpotential are adopted and the following charge profiles are determined in each case, which charge current is determined with respect to the charge state by simulation in such a way that the resulting anode overpotential always corresponds to or does not fall below the limit value of the anode overpotential.

3. The method according to claim 1 or 2, wherein the test charge profiles are assigned to a plurality of the device battery packs (41) when one or more test charge profiles are assigned to the device battery packs (41), respectively.

4. The method according to claim 1 , wherein the change in aging state is determined as the difference in aging state at the beginning and at the end of the observation period, or wherein the change in aging state is determined as the sum of the differences in aging state at the beginning and at the end of the respective charging process.

5. The method according to any one of claims 1 to 4, wherein: The average charging time is determined as a mean value of the charging times of charging processes in which a change in the state of charge greater than a predefined threshold value occurs.

6. The method according to any one of claims 1 to 5, wherein: The determination of the optimized charging profile is performed based on the Pareto curve and the charging profile model, the method having the following steps: - determining a modeled state point of average charging time and aging state change based on the probabilistic charging distribution model, wherein the modeled state point has a minimum cost value according to the cost function, - Determining a charging profile based on the modeled state point with the minimum cost value and the Pareto curve.

7. The method according to claim 6, wherein: The optimized charging distribution is determined in the following manner: in particular, a specific modeled state point from the charging distribution model is projected onto the nearest Pareto state point of the Pareto curve by making a vertical line of the modeled state point fall on the Pareto curve, thereby generating a corresponding Pareto state point on the Pareto curve, wherein the Pareto state point is determined by the average charging time and aging state change for the minimum cost of the simulation, wherein the Pareto state point is assigned to the anode overpotential or a limit value for the anode overpotential by correspondingly applying the electrochemical battery model, wherein the optimized charging distribution is derived based on the limit value.

8. A method according to any one of claims 1 to 7, wherein within the scope of active learning, one or more device battery packs (41) are selected, the actual state point of which has the following cost value, and under the same aging state change or under the same average / aggregate charging time information, the deviation of the cost value according to the cost function from the minimum cost value of the Pareto curve exceeds a predetermined threshold.

9. The method according to any one of claims 1 to 8, wherein: One or more specific device batteries (41) for further training the aging state model and / or for re-parameterizing the electrochemical battery model are identified by: - determining in the charging profile model one or more modeled state points with uncertainty of the charging profile model via a predetermined threshold value, and correspondingly determining the device battery pack (41) whose actual state point is closest to the one / multiple determined modeled state points, and / or - determining a modeling state point whose cost value according to the cost function deviates from the cost value of the Pareto curve for the same aging state change or for the same average / aggregate charging time information by more than a predefined threshold value.

10. An apparatus for performing the method according to any one of claims 1 to 9.

11. A computer program product, the program product comprising instructions which, when executed by at least one data processing device, cause the data processing device to perform the steps of the method according to any one of claims 1 to 9.

12. A machine-readable storage medium comprising instructions which, when executed by at least one data processing device, cause the data processing device to perform the steps of the method according to any one of claims 1 to 9.

Images