Method and device for the initial provision of an aging state model for energy storage devices based on active learning algorithms

Abstract

Method for initially providing an at least partially data-based aging state model (4) for an electrical energy storage device (3), comprising the following steps: - Providing (S1) a number of energy storage devices (3) on a test bench (1) for measurement depending on a respective load profile, wherein the load profiles are different and characterize a time course of at least one load-bearing operating parameter for the energy storage device (3); - Operating (S2) the number of energy storage devices (3) with their respective assigned load profiles and recording operating parameter profiles over time; - At each given evaluation time, determine (S4) an aging state of a subset of the energy storage devices (3) as a label and generate a training data set with the operating parameter profiles and the determined label for each energy storage device (3) of the subset of energy storage devices (3); - Selecting (S6) the subset of energy storage devices (3) with their respective assigned load profiles depending on an optimization procedure that considers total costs (C J ) a measurement of the energy storage (3) on the test bench (1) and of a total information measure (Info J ) the measurement of the energy storage (3) depends.

Description

Technical field

[0001] The invention relates to methods and devices for initially providing an at least partially data-based aging state model for electrical energy storage devices of the same type, and in particular methods for predicting the surveying costs and / or the surveying time. Technical background

[0002] The energy supply for the operation of off-grid electrical devices and machines, such as electrically powered motor vehicles, is usually provided by electrical energy storage devices, such as device batteries or vehicle batteries.

[0003] Electrical energy storage devices degrade over their lifespan, depending on their load and usage. This so-called aging leads to a continuously decreasing maximum power or storage capacity. The state of aging (SOH-C) is a measure used to indicate the aging of energy storage devices. For device batteries used as electrical energy storage, a new device battery can, by convention, have a SOH-C of 100%, which decreases progressively over its lifespan.

[0004] A measure of the aging of an electrical energy storage device (time-dependent change in the state of aging) depends on an individual load on the energy storage device, i.e., in the case of vehicle batteries of motor vehicles, on the usage behavior by a driver, external environmental conditions and the type of vehicle battery.

[0005] To monitor the aging states of electrical energy storage devices in a variety of devices, operating parameter data can be continuously recorded and transmitted as operating parameter profiles in blocks to a central unit external to the device.

[0006] To determine the aging state of an electrical energy storage device using model-based operating data, an initial aging state model is required. This involves conducting an initial measurement, for example in a laboratory or test bench, on a specific number of energy storage devices to generate training data for the model. The energy storage devices are operated in various ways, with energy being supplied or discharged depending on the type of device to simulate operating cycles. The energy expenditure required is considerable and scales with the number of energy storage devices to be initially measured. The initial measurement can also be time-consuming, especially if sufficient training data needs to be collected for aged energy storage devices.

[0007] Document DE 10 2020 206 592 A1 discloses a computer-implemented method for operating a motor vehicle, in particular an electrically powered motor vehicle, depending on a predicted aging state of an electrical energy storage device.

[0008] Document DE 10 2020 108 365 A1 discloses a device for determining lifetime information for an electrical energy storage device, wherein the device is configured to determine statistical usage data relating to a previous use of the energy storage device and, using a machine learning-based prediction unit and based on the statistical usage data, to determine lifetime information relating to an aging state and / or to a remaining lifetime of the energy storage device.

[0009] Document DE 10 2017 103 617 A1 discloses a method for estimating the aging state of a battery system, in particular in a vehicle. Disclosure of the invention

[0010] According to the invention, a method for initially providing an at least partially data-based aging state model for an energy storage type according to claim 1 and a corresponding device according to the dependent claim are provided.

[0011] Further details are specified in the dependent claims.

[0012] According to a first aspect, a procedure for the initial provision of an at least partially data-based aging state model for an electrical energy storage device is planned, comprising the following steps: - Providing a number of energy storage devices on a test bench for measurement depending on a respective load profile, wherein the load profiles are different and characterize a time course of at least one stressing operating parameter for the energy storage device; - Operating the number of energy storage systems with their respective assigned load profiles and recording operating parameter trends over time; - At each predetermined evaluation time, determine an aging state of a subset of the energy storage devices as a label and generate a training dataset with the operating parameter profiles and the determined label for each energy storage device in the subset of energy storage devices; - Selecting the subset of energy storage devices with their respective assigned load profiles depending on an optimization procedure that depends on the cost of measuring the energy storage devices on the test bench and on an information measure of the measurement of the energy storage devices.

[0013] The aging state of an electrical energy storage device is not usually measured directly. This would require a number of sensors inside the device, making its manufacture expensive and complex, and increasing its space requirements. Furthermore, practical measurement methods for directly determining the aging state of energy storage devices are not yet commercially available.

[0014] For capacity reasons, monitoring the energy storage systems of numerous devices is therefore performed in a central unit external to the devices. The devices can transmit operating parameter profiles of the energy storage systems to the central unit, which then determines the current electrochemical state and / or aging state. Depending on the model used, time series of operating parameters, such as battery current, battery temperature, state of charge, and / or battery voltage for a device battery, are continuously recorded and transmitted to the central unit in blocks and, if necessary, in compressed form. The operating parameter profiles are then evaluated so that, based on one or more aging state models, a device-specific aging state and, if necessary, other parameters can be calculated / determined.Furthermore, the operating parameters from the multitude of energy storage systems can be evaluated using statistical methods to improve the applied aging state models, so that the determination and prediction of the aging state of the energy storage systems can be progressively improved.

[0015] The state of health (SOH) is the key parameter for indicating remaining battery capacity or charge in device batteries used as energy storage devices. The state of health represents a measure of the device battery's aging. In the case of a device battery, battery module, or battery cell, the state of health can be expressed as the capacity retention rate (SOH-C). The capacity retention rate (SOH-C), i.e., the capacity-related state of health, is expressed as the ratio of the measured instantaneous capacity to the initial capacity of the fully charged battery and decreases with increasing age. Alternatively, the state of health can be expressed as the increase in internal resistance (SOH-R) relative to an internal resistance at the beginning of the device battery's life. The relative change in internal resistance (SOH-R) increases with increasing battery aging.

[0016] Due to the often physically difficult-to-describe electrochemical effects during the operation of an energy storage system, the use of a data-based model as or in conjunction with an aging state model has proven effective.

[0017] One possible aging state model can be a hybrid model, combining a physical aging model with a data-driven model. In a hybrid model, a physical aging state can be determined using a physical or electrochemical aging model and then treated with a correction value derived from a data-driven correction model, specifically through addition or multiplication. The physical aging model is based on electrochemical model equations that characterize the electrochemical states of a nonlinear system of differential equations with respect to aging reactions. These equations are continuously calculated using a time integration method and mapped to the physical aging state for output, expressed as SOH-C and / or SOH-R.The calculations can typically be performed in the central unit (cloud) at intervals of predefined evaluation periods, e.g. once a week.

[0018] Furthermore, the correction model of the hybrid data-based aging state model can be designed using a probabilistic or artificial intelligence-based probabilistic regression model, particularly a Gaussian process model, and can be trained to correct the aging state obtained from the physical aging model. This may include a data-based correction model for correcting the capacity-related aging state and, if necessary, another data-based correction model for correcting the resistance-change-related aging state. Possible alternatives to the Gaussian process include other supervised learning methods, such as those based on a random forest model, an AdaBoost model, a support vector machine, or a Bayesian neural network.

[0019] When commissioning a new type of energy storage system, it is necessary to determine its aging state using an aging state model. Since precise knowledge of the energy storage system's aging is typically unavailable at commissioning, it is necessary to initially define an aging state model that can at least approximately predict the aging state based on operational parameters. Therefore, initially defining the aging state model requires initial training of the data-driven model.

[0020] Training data for such initial training is typically generated in a laboratory or on a test bench and comprises a predefined number of energy storage devices, usually randomly selected, each operated with different load profiles. These load profiles include, or are converted into, cyclic current flows, temperature profiles, and the like. Particularly in the case of device batteries used as energy storage devices, these load profiles can include different charging and discharging current profiles at varying temperatures and can be defined and characterized in a condensed form, especially using histogram data. These load profiles are converted into time series of current inputs and outputs, and corresponding operating parameters, such as battery current, battery voltage, battery temperature, and state of charge, are recorded and stored as operating parameter profiles.

[0021] At predetermined time points, aging condition measurements are performed using suitable additional models or measurement methods to determine a label for the operational parameter profiles. From this, training datasets are created, which can be used to parameterize the aging condition model and / or to train the data-driven model in the case of a hybrid or purely data-driven aging condition model. Various aging condition models or methods are available for label determination.

[0022] A possible model or method for determining the state of aging could be a basic model in which a state-of-charge (SOH) capacitance (C) measurement is performed by coulomb counting or by calculating a current integral over time during the charging process. This integral is then divided by the state-of-charge variation between the beginning and end of the respective charging and / or discharging phase. Advantageously, calibration is performed using the open-circuit voltage characteristic during periods of inactivity to incorporate the state-of-charge profile into the calculations of the central processing unit. A sufficiently reliable indication of the state of aging can be obtained, for example, if the vehicle battery is brought from a fully discharged state to a fully charged state during a charging process under reproducible load and environmental conditions, starting from a defined relaxed state.The maximum charge detected in this way can be related to the initial maximum charging capacity of the vehicle battery. Resistance-related aging states (SOH-R values) can also be calculated from voltage changes in relation to current changes. These are typically based on a defined time interval, defined environmental conditions, and the energy flow direction of the system.

[0023] According to the above procedure, the aging states of a subset of energy storage devices can now be determined as labels, and a training dataset containing the operating characteristic curves and the determined label for each energy storage device in the subset can be generated. The selection of the subset of energy storage devices with their respective assigned load profiles is performed using an optimization procedure.

[0024] The optimization procedure for selecting the subset of energy storage devices can aim to minimize the costs of measurement on the test bench and to maximize the information gain for the creation of the initial aging state model, as determined by the information measure.

[0025] According to the above procedure, the number of energy storage devices measured or currently being measured is thus intended to be successively reduced, depending on the information gain that can be obtained for the further development of the aging state model through further measurement of the energy storage device in question.

[0026] According to one embodiment, in a Gaussian process model as a probabilistic model, the predictive covariance can be determined from the input vectors of one or more of the energy storage devices for the evaluation times of an entire measurement period, wherein the input vectors are determined from the load profiles.

[0027] A method is proposed in which the aging state model is successively trained, and those batteries that contribute little to the model's development are removed from the measurement or disregarded. For this purpose, when using a probabilistic regression model as the data-driven model, the predictive covariance is calculated at each evaluation time point. ∑(SOHj(t+t1),SOHj(t+t2),…,SOHj(t+tn)) The predictive covariance is determined by the data-based model, which results from the probabilistic regression model (Gaussian process model) provided in the aging state model. In Gaussian process models, as probabilistic regression models, the predictive covariance does not depend on previously unseen aging states, but only on the input variables x. J (t) = [x(t), m( ), z, Phys[x(t)]) of the Gaussian process model and can therefore be considered ∑(x¯j(t+t1),x¯j(t+t2),…,x¯j(t+tn)) are written where x(t) represents one or more operating parameter profiles, m(t) represents histogram data as a load profile, z(t) represents multidimensional electrochemical states of the physical aging model, such as the SEI thickness, the amount of cyclable lithium, the amount of active material, electrochemical concentrations, and the like, and Phys[x(t)] represents the physical (modeled) aging state.

[0028] The predictive covariance is a matrix to which information measures, such as the determinant or the maximum eigenvalue, can be applied. This yields an information measure for each of the measured energy storage devices. The information measure for multiple energy storage devices J = (j1 ... jm) can be expressed as InfoI=det(∑(x¯j1(t+t1),…,x¯j1(t+tn),x¯jm(t+t1),…,x¯jm(t+tn))) are specified where J corresponds to the number of selected energy storage devices and j1...jm to the index of a specific energy storage device.

[0029] Furthermore, a cost measure is applied and determined for each energy storage device to be measured. This cost measure can include the energy input or consumption during laboratory or test bench measurements and the total duration of the energy storage device measurements. Additionally, the cost measure can also consider the total test bench costs, the test bench occupancy time, and other material usage. Thus, a cost measure C can be predictively provided for each of the number J of energy storage devices to be measured, as follows: Cj=C(x¯j(t+t1),x¯j(t+t2),…,x¯j(t+tn))Cj=C1+..+Cm

[0030] With the set of cost information points for many of the batteries to be measured, the optimization problem can now be solved and a Pareto front determined. The Pareto front represents the cost measure over the achievable information level of the initial aging state model to be created.

[0031] In line with a variant of an active learning method, the selection of cost-information measure points can now involve selecting those energy storage systems for which these cost-information measure points are as close as possible to or on the Pareto front.

[0032] It may be planned that the optimization procedure - is based on an objective function and in particular is carried out using a greedy algorithm, where the objective function represents a weighted sum of the costs of measuring the number of energy storage devices and an overall measure of information by measuring the subset of energy storage devices, - by maximizing an overall measure of information subject to the constraint that the total cost of measuring the selected vehicle batteries is less than a predetermined maximum cost; - by minimizing total costs subject to the constraint that the total information measure is greater than a given minimum information measure; - by minimizing total costs subject to the constraint that the probability of total costs being less than a given maximum cost is greater than a given probability; or - by maximizing an overall information measure subject to the constraint that the probability that the information measure is greater than a given minimum information measure is greater than a given probability.

[0033] According to the above approaches for the optimization process, a weighting between costs and information gain can be specified by an objective function, for example Z. J = Info J + αC Jwith a user-defined weighting parameter α. Now the best energy storage devices can be determined, in particular by applying a greedy algorithm. Greedy algorithms first select the best energy storage device and then add the second-best, assuming that the best energy storage device has already been measured. Alternatively, instead of preselecting the above objective function, a constrained optimization problem can also be solved by limiting the maximum cost and by limiting the minimum information gain.

[0034] As an alternative to weighting the objective function with the weighting parameter, a constrained optimization problem can also be solved by maximizing the information measure Info. JThe problem must be solved to select a subset of energy storage devices. This is done under the constraint that the total cost of measuring the selected energy storage devices is less than a predefined user parameter that specifies the maximum permissible cost.

[0035] As an alternative to weighting the objective function with the weighting parameter, a constrained optimization problem can also be solved by minimizing costs to select a subset of energy storage devices. This is done under the constraint that the information measure Info J is greater than a predetermined minimum information gain.

[0036] When probabilistic regression models are used, energy storage systems can be selected for which the probability that the sum of the costs for measuring the selected energy storage systems with the highest information measures is less than a predefined parameter is greater than a predefined parameter. This allows for the provision of a discrete cost-information gain point on the Pareto front, which is linked to actionable recommendations based on the model and specifically identifies energy storage systems for further operation.

[0037] According to the procedure, specific measurements of the aging state are now taken for the selected energy storage devices, but not for the unselected ones. This, together with the associated operating parameter profiles, yields a training dataset.

[0038] The data-based aging state model can be designed with a probabilistic data-based model, wherein for one of the energy storage devices an input vector of the data-based model, comprising at least one operating parameter profile, operating characteristics from the operating parameter profiles, an internal state of the energy storage device and / or a physically modeled aging state, is mapped to the aging state to be modeled of the energy storage device in question or to a correction parameter for correcting a physically modeled aging state of the energy storage device in question, wherein the information measure for an energy storage device is determined as the determinant of the predictive covariance.

[0039] The predictive covariance can be determined from the input vectors of one or more of the energy storage devices for the evaluation times of an entire measurement period.

[0040] When sufficient new information is available, the data-driven aging state model is retrained. This training can be supplemented by automated hyperparameter tuning, for example, using gradient-based methods or black-box methods such as Bayesian optimization. The process terminates when a specific accuracy requirement is met, such as < 1.5 SOHC on a relevant, previously provided validation dataset, and other robustness requirements, such as those evaluated via cross-validation, are satisfied.

[0041] It may be provided that at each evaluation point only the energy storage units of the subset of energy storage units are measured as labels to determine an aging state, and the remaining energy storage units continue to be operated according to the load profile.

[0042] In an alternative embodiment, at each evaluation time only the energy storage devices of the subset of energy storage devices can be measured as labels to determine an aging state and continue to be operated according to the load profile, while the remaining energy storage devices are removed from the test bench.

[0043] Thus, depending on the selected energy storage devices, the remaining energy storage devices, which would result in insufficient information gain, can be excluded from further measurement and removed from the test bench. This allows the measurement to continue with only a reduced number of selected energy storage devices on the test bench. This procedure can be carried out successively to reduce the number of energy storage devices being measured and thereby lower the costs of using the test bench.

[0044] According to another aspect, a device for carrying out one of the above procedures is provided. Brief description of the drawings

[0045] The embodiments are explained in more detail below with reference to the accompanying drawings. These show: Fig. 1 a schematic representation of a test bench for measuring a large number of vehicle batteries to create an initial aging state model; Fig. 2 a schematic representation of a hybrid aging state model; Fig. 3. A flowchart illustrating a procedure for the optimized creation of an initial aging state model; Fig. 4. A representation of a Pareto front from cost-information measure points for the various vehicle batteries in the survey; and Fig. 5 A representation of the progression of aging states of various vehicle batteries under different load patterns. Description of embodiments

[0046] Fig. Figure 1 shows a schematic arrangement of a test bench with a test bench unit 2, which is connected to a plurality of vehicle batteries 3 as electrical energy storage devices. The test bench unit 2 controls the vehicle batteries 3 according to a predefined load pattern that characterizes various current and / or temperature profiles. The current and / or temperature profiles result in different loads and thus different cyclic aging of the vehicle batteries 3, for example, by specifying an ampere-hour throughput, specifications for charging profiles, in particular maximum charging currents depending on the state of charge during charging, maximum charging currents during discharging, maximum and average charging and discharging currents, and corresponding temperature conditions. The load patterns are specified for each of the vehicle batteries 3 and correspond to more or less demanding operating modes of the vehicle batteries 3.Operating parameters of the vehicle batteries 3 are continuously recorded and temporarily stored. These parameters include battery voltage, battery current, battery temperature, and state of charge.

[0047] Test bench 1 serves to acquire data on the aging of the vehicle batteries 3 in order to provide a corresponding initial aging state model 4, which makes it possible to determine the current aging state of the respective vehicle battery with a predefined minimum accuracy, based on operating parameter profiles recorded during real-world operation. The aging state model 4 to be created for this purpose can be at least partially data-based and may include a data-based probabilistic regression model.

[0048] The measurement of the numerous vehicle batteries 3 on test bench 1 involves determining the aging state of selected vehicle batteries 3 at regular intervals using a suitable method. This, in conjunction with the corresponding operating characteristic curves, which are defined by the load pattern assigned to each vehicle battery 3, yields training data sets that can be used to train the data-based / hybrid aging state model 4. The use of test bench 1 and the measurement of the numerous vehicle batteries 3 incur costs, resulting from energy consumption, test bench occupancy time, and other material usage. These costs are to be reduced during the measurement of the numerous vehicle batteries 3 without compromising the quality of the initially trained aging state model 4.

[0049] Fig. Figure 2 schematically illustrates the functional structure of an embodiment of a data-based aging state model 9, which is designed in a hybrid manner. The aging state model 9 comprises a physical aging model 5 and a data-based correction model 6.

[0050] The physical aging model 5 is a mathematical model based on nonlinear differential equations. Evaluating the physical aging model 5 of the aging state model 9 with operating parameter profiles, particularly since the beginning of the vehicle battery's service life, leads to the establishment of an internal state of the system of physical differential equations that corresponds to a physical internal state of the vehicle battery. Since the physical aging model is based on physical and electrochemical laws, its model parameters are quantities that specify physical properties.

[0051] The time series of the operating parameters x(t) of the vehicle battery to be evaluated are thus directly incorporated into the physical aging state model 5, which is preferably implemented as an electrochemical model and models corresponding internal electrochemical states z(t), such as layer thicknesses (e.g. SEI thickness), changes in the cyclable lithium due to anode / cathode side reactions, rapid consumption of electrolytes, slow consumption of electrolytes, loss of active material in the anode, loss of active material in the cathode, etc., using nonlinear differential equations in a multidimensional state vector.

[0052] The physical aging model 5 thus corresponds to an electrochemical model of the vehicle battery in question 3. Depending on the operating parameter profiles x(t), this model determines internal physical battery states z(t) in order to establish a physically based aging state SOHph = Phys[x(t)] of dimension at least one, depending on the electrochemical states z(t) mentioned above, which are mapped linearly or non-linearly to a capacity conservation rate (SOH-C) and / or an internal resistance increase rate (SOH-R) in order to provide these as an aging state (SOH-C and SOH-R).

[0053] However, the model values ​​for the physical aging state SOHph provided by the electrochemical model are inaccurate in certain situations, and it is therefore intended to correct these with a correction factor k. The correction factor k is provided by the data-based correction model 6, which is trained using training data sets obtained by test rig 1.

[0054] Preferably, the data-based correction model corresponds to a Gaussian process model. GP=GP[x(t),x(thist),z,Phys[x(t)]))∼N(μ,Σ), where µ is an mx 1-dimensional vector representing the predictive mean and Σ is the mxm dimensional predictive covariance matrix of the Gaussian process model. The formulas for the mean and variance of the Gaussian process are as follows: μ(x∗)=kTCN−1y Σ(x∗)=c−kTCN−1k where N corresponds to the number of labels, x* mxd is dimensional and specifies a set of m new points in the input space, where [x(t),x(t hist ),z,Phys[x(t)]) =x* and y each correspond to an N x 1 dimensional vector, where y represents a measured or labeled aging state. k corresponds to an N x m dimensional matrix of kernel evaluations that represent correlations encoded in the kernel of the Gaussian process between the m new points and the N measured points. C corresponds to an N x N dimensional matrix of kernel evaluations between the N measured points, and c to an m x m matrix of kernel evaluations between the m new points. T denotes the transpose. See also Bishop, “Pattern Recognition and Machine Learning”, 2006.

[0055] The correction model 6 receives operating characteristics m(t) as input, which are determined from the profiles of the operating variables x(t) and may also include one or more of the internal electrochemical states of the differential equation system of the physical model 5. Furthermore, the correction model 6 can receive the physical aging state SOHph, obtained from the physical aging model 5, as input. The operating characteristics m(t) of the current evaluation period are generated in a feature extraction block 8 based on the operating variable profiles x(t). Additionally, the internal states z(t) from the state vector of the electrochemical physical aging model 5, and advantageously the physical aging state SOHph, are supplied to the correction model 6. The feature vector m(t) is robust because it is independent of the model quality of the physical aging model or its state.Therefore, considering the feature vector m(t) is a useful addition to the internal states z(t) of the physical or electrochemical aging model.

[0056] From the operating parameters x(t), operating characteristics m(t) can be generated in the central unit 2 for each vehicle fleet 3, or in other embodiments, directly within the respective vehicles. These characteristics relate to an evaluation period. The evaluation period for determining the aging state can range from a few hours (e.g., 6 hours) to several weeks (e.g., one month). A typical value for the evaluation period is one week.

[0057] The operating characteristics m(t) can include, for example, characteristics related to the evaluation period and / or accumulated characteristics and / or statistical quantities determined over the entire lifetime to date. In particular, the operating characteristics can include, for example: electrochemical states, such as SEI layer thickness, changes in cyclable lithium due to anode / cathode side reactions, rapid uptake of electrolyte solvents, slow uptake of electrolyte solvents, lithium deposition, loss of active anode material and loss of active cathode material, information on impedances, etc.the internal resistances, histogram features such as temperature versus state of charge, charging current versus temperature and discharging current versus temperature, in particular multidimensional histogram data regarding the battery temperature distribution versus the 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 during a charging process (especially for charging processes where the charge increase is above a threshold fraction [e.g. 20% ΔSOC] of the total battery capacity), the charging capacity as well as an extreme value (e.g. maximum) of the differential capacity during a measured charging process with a sufficiently large state of charge range (smoothed curve of dQ / dU: charge change divided by change in battery voltage) or the accumulated driving performance.These quantities are preferably converted in such a way that they best characterize real-world usage behavior and are normalized in the feature space. The operational characteristics m(t) can be used in whole or in part for the procedure described below.

[0058] To determine a corrected output aging state SOH, the outputs SOHph, k of the physical aging model 5 and the data-based correction model 6, which is preferably implemented as a Gaussian process model, are applied to each other. In particular, these can be added in a summing block 7 or multiplied (not shown) to obtain the output modeled aging state SOH for a current evaluation period. In the case of addition, the confidence of the Gaussian process can also be used as the confidence of the output corrected aging value SOH of the hybrid model. The confidence or confidence value of the Gaussian process model thus characterizes the modeling uncertainty of mapping operating characteristic points to an aging state.

[0059] The initial training of the hybrid aging state model 9 takes place in the test bench 1. For this purpose, training data sets are created that assign operating parameter profiles of a vehicle battery operated depending on a load profile to an empirically or model-based determined aging state as a label.

[0060] For example, to determine an aging state as a label for training the hybrid or data-based aging state model, a base model can be provided according to which a state-of-health (SOH) capacitance (C) measurement is performed by coulomb counting or by calculating a current integral over time during the charging process, which is then divided by the state-of-state difference between the beginning and end of the respective charging and / or discharging phase. A sufficiently reliable indication of the aging state can be obtained, for example, if the vehicle battery is brought from a fully discharged state to a fully charged state during a charging process under reproducible load and environmental conditions, starting from a defined relaxed state. The maximum charge thus recorded can be related to the initial maximum charge capacity of the vehicle battery.Resistance-related aging states (SOH-R values) can also be calculated by changing the voltage in relation to changing the current. These are typically defined over a specific time interval, under specific environmental conditions, and according to the energy flow direction of the system.

[0061] Determining an aging state as a label can be achieved in a known manner by evaluating the operating parameter profiles with an additional aging model under defined load and environmental conditions for label generation, such as constant temperature, constant current, and the like. Other models can be used to determine the aging state. Training the data-driven correction model can be performed conventionally based on the training datasets and the residuals of the modeled aging state.

[0062] In Fig. Figure 3 shows a flowchart describing the procedure for measuring the multitude of vehicle batteries 3 on the test bench 1. The procedure is executed in the test bench control unit 2 and results in providing an initial data-based aging state model 4, which enables sufficient accuracy in determining the aging state based on the operating characteristic curves of vehicle batteries 3 in real-world operation.

[0063] In step S1, test bench 1 is equipped with a large number of brand-new (or reference-condition) vehicle batteries 3 of the same type, and each vehicle battery 3 is assigned a predefined load pattern. The load patterns are different and each represents a load ranging from low to high for the assigned vehicle battery 3. The load patterns specify parameters from which time-dependent battery current profiles in conjunction with temperature profiles can be derived, which stress the vehicle batteries 3 in different ways.

[0064] In step S2, the multitude of vehicle batteries 3 are operated according to the specified load pattern.

[0065] Step S3 checks whether an evaluation point has been reached. The evaluation point can be scheduled at regular intervals, such as between one week and two months.

[0066] If an evaluation point is reached (alternative: Yes), the procedure continues with step S4; otherwise, it jumps back to step S2.

[0067] In step S4, starting from the current training state of the aging state model, an information measure is determined for each of the multiple vehicle batteries 3. This measure indicates the information gain that can be obtained during further measurements of the respective vehicle battery 3. This information measure can be expressed as the predictive covariance. In particular, when using a Gaussian process model as a probabilistic regression model, the predictive covariance depends on ∑(SOHj(t+t1),SOHj(t+t2),…,SOHj(t+tn)) not from previously undetermined aging states, but only from input variables x J (t) = [x(t), m(t), z, Phys[x(t)]). Therefore, one can also write Σ(x j (t + t1), x j (t + t2), ..., xj (t + t n This represents a matrix that can be evaluated using information measures such as the determinant or the maximum eigenvalue. The information gained can thus be easily derived from the analysis of the operational parameter profiles resulting from the load patterns.

[0068] As a result, step S4 yields the information measures about the expected information gains from the individual vehicle batteries 3 of the multitude of vehicle batteries 3. The combination of information measures (total information measure) of several vehicle batteries 3 J = (j1 ... jm) can be described by the covariance matrix. InfoJ=det(∑(x¯j1(t+t1),…,x¯j1(t+tn),x¯jm(t+t1),…,x¯jm(t+tn))) be specified.

[0069] In step S5, a cost measure is calculated for each of the vehicle batteries 3. This cost measure can include the costs of using the test bench, including the anticipated energy consumption and test time, as well as other material usage. Thus, the cost measure can be predictively provided for each of the multiple vehicle batteries 3 if the costs at the discretized evaluation times up to the end t n The surveying time is added to this. Therefore, the cost for each vehicle battery is 3. Cj=C(x¯j(t+t1),x¯j(t+t2),…,x¯j(t+tn)) and as total costs for measuring a certain quantity of vehicle batteries 3 CJ=C1+..+Cm where 1...m indices of vehicle batteries correspond to 3 a subset of the number J.

[0070] Thus, for each of the vehicle batteries 3, one obtains an information measure regarding the expected information gain and the corresponding costs that a further measurement of the respective vehicle battery 3 on the test bench will incur. This results in a Pareto front, as is found, for example, in Fig. 4 is shown schematically.

[0071] By selecting cost-information measure points that correspond to the costs and information measures of each of the vehicle batteries 3, in step S6 those vehicle batteries 3 can be selected for which a label in the form of a measured aging state is to be determined at the current evaluation time.

[0072] The determination of the aging state can be carried out using a suitable aging state model or a suitable measurement procedure for determining the aging state.

[0073] A possible model or method for determining the state of aging is measurement by Coulomb counting or by calculating a current integral over time during the charging process. Here, the transferred charge is divided by the state of charge variation between the beginning and end of the respective charging and / or discharging phase. Advantageously, calibration is performed using the open-circuit voltage characteristic during periods of inactivity to incorporate the state of charge profile into the calculations of the central processing unit. A sufficiently reliable indication of the state of aging can be obtained, for example, if the vehicle battery is brought from a fully discharged state to a fully charged state during a charging process under reproducible load and environmental conditions, starting from a defined relaxed state.The maximum charge detected in this way can be related to an initial maximum charging capacity of the vehicle battery 3.

[0074] The selection of the vehicle batteries to be measured 3 can be determined using an optimization procedure according to an objective function that weighs costs against information gain. The objective function can have the form: Z J = Info J + αC J With a user-defined weighting parameter α, a predetermined number of vehicle batteries can be determined for which the objective function yields the maximum values. Greedy algorithms can be used to solve the optimization problem, which first select the best vehicle battery (3) (highest result of the objective function Z). J ) and then the next best one (with respect to Z) J ) is added, assuming that the previously identified vehicle batteries 3 have already been measured.

[0075] As an alternative to weighting the objective function with the weighting parameter, a constrained optimization problem can also be solved by maximizing the information measure Info. J This can be achieved by selecting a subset of vehicle batteries from the large number of available batteries. This is subject to the constraint that the total cost of measuring these selected batteries is less than a predefined user parameter that specifies the maximum permissible cost.

[0076] As an alternative to weighting the objective function with the weighting parameter, a constrained optimization problem can also be solved by minimizing costs through the selection of a subset of vehicle batteries from the multitude of available vehicle batteries. This is subject to the constraint that the total information measure Info Jis greater than a specified minimum information gain about all selected vehicle batteries.

[0077] If probabilistic models are used for the aging state model, the constraint can also be specified as a probability greater than a given probability that the costs are less than the given parameter. Alternatively, the constraint can also be specified as a probability greater than a given probability that the total information measure Info J is greater than a specified minimum information level.

[0078] In step S7, the selected vehicle batteries 3 are used to perform a highly accurate measurement of their state of aging, which is then labeled. This results in aging state profiles for the numerous vehicle batteries 3, as seen, for example, in Fig.Figure 5 illustrates this. In conjunction with the operational parameter profiles derived from the load pattern and the defined aging state as a label, training datasets are available with which the aging state model 4 can be trained. The newly determined training datasets are used in step S8 for further training of the data-driven probabilistic regression model. Additionally, automated hyperparameter tuning can be performed, e.g., using a gradient-based method or a black-box method such as Bayesian optimization.

[0079] In a subsequent step S9, it is checked whether the trained aging state model for a provided validation dataset exceeds a sufficient accuracy, e.g., a maximum error of 1.5% SOHC. If this is the case (alternative: Yes), the procedure continues with step S2; otherwise (alternative: No), the measurement of the vehicle batteries 3 is terminated.

[0080] In an alternative embodiment, the selected vehicle batteries can be measured as the sole vehicle batteries, while the remaining, unselected vehicle batteries 3 are removed from the measurement or from the test bench to free up test bench slots. This gradually reduces the number of vehicle batteries remaining in the measurement, thus significantly reducing the overall measurement costs, which scale considerably with the number of vehicle batteries 3.

Claims

[1] Method for initially providing an at least partially data-based aging state model (4) for an electrical energy storage device (3), comprising the following steps: - Providing (S1) a number of energy storage devices (3) on a test bench (1) for measurement depending on a respective load profile, wherein the load profiles are different and characterize a time course of at least one load-bearing operating parameter for the energy storage device (3); - Operating (S2) the number of energy storage devices (3) with their respective assigned load profiles and recording operating parameter profiles over time; - At each given evaluation time, determine (S4) an aging state of a subset of the energy storage devices (3) as a label and generate a training data set with the operating parameter profiles and the determined label for each energy storage device (3) of the subset of energy storage devices (3); - Selecting (S6) the subset of energy storage devices (3) with their respective assigned load profiles depending on an optimization procedure that considers total costs (C J ) a measurement of the energy storage (3) on the test bench (1) and of a total information measure (Info J ) the measurement of the energy storage (3) depends. [2] The method of claim 1, wherein the optimization method aims to reduce the total costs (C J ) the measurement on the test bench (1) to minimize and to achieve a result through the information measure (Info J) to maximize the information gain for the creation of the initial aging state model (4), whereby the costs depend in particular on an energy input or requirement during the measurement on the test bench (1) and / or a duration of the entire measurement of the energy storage devices (3) to be measured and / or on total test bench costs, which take into account at least an occupancy time of the test bench (1) and / or a material input. [3] Method according to claim 2, wherein the optimization method - is based on an objective function and in particular is carried out using a greedy algorithm, where the objective function is a weighted sum of the total costs (C J ) for measuring the number of energy storage devices (3), and an overall information measure (Info J ) by measuring the subset of energy storage (3), - by maximizing an overall measure of information (Info J) subject to the constraint that the total costs (C J ) of measuring the energy storage devices selected in this way (3) are smaller than the specified maximum costs; - by minimizing total costs (C J ) subject to the constraint that the total information measure (Info J ) is greater than a given minimum information measure; - by minimizing total costs (C J ) subject to the constraint that there is a probability that the total costs (C J ) are smaller than a given maximum cost, is larger than a given probability; or - by maximizing an overall measure of information (Info J ) subject to the constraint that there is a probability that the total information measure (Info J ) is greater than a given minimum information measure, is greater than a given probability. [4] Method according to one of claims 1 to 3, wherein the data-based aging state model (4) is designed with a probabilistic data-based model, wherein for one of the energy storage devices (3) an input vector of the data-based model, comprising at least one operating parameter profile (x(t)) and / or at least one operating characteristic (m(t)) from the at least one operating parameter profile, an internal state of the energy storage device (3) and / or a physically modeled aging state, is mapped to the aging state to be modeled of the energy storage device (3) in question or to a correction parameter for correcting a physically modeled aging state of the energy storage device (3) in question, wherein the information measure () for an energy storage device (3) is determined as the determinant of the predictive covariance. [5] Method according to claim 4, wherein the predictive covariance is determined from the input vectors of one or more of the energy storage devices (3) for the evaluation times of an entire measurement period. [6] Method according to claim 5, wherein, in a Gaussian process model as a probabilistic model, the predictive covariance is determined from the input vectors of one or more of the energy storage devices (3) for the evaluation times of an entire measurement period, wherein the input vectors are determined from the load profiles. [7] Method according to any one of claims 1 to 6, wherein at each evaluation time only the energy storage devices (3) of the subset of energy storage devices (3) are measured as labels to determine an aging state and the remaining energy storage devices (3) continue to be operated according to the load profile. [8] Method according to any one of claims 1 to 6, wherein at each evaluation time only the energy storage devices (3) of the subset of energy storage devices (3) are measured as labels to determine an aging state and continue to be operated according to the load profile, while the remaining energy storage devices (3) are removed from the test bench. [9] Method according to any one of claims 1 to 8, wherein the aging state model (4) is trained with the determined training data sets. [10] Device for carrying out one of the methods according to any one of claims 1 to 9. [11] Computer program product comprising instructions which, when the program is executed by at least one data processing device, cause it to perform the steps of the method according to any one of claims 1 to 9. [12] Machine-readable storage medium comprising instructions which, when executed by at least one data processing device, cause it to perform the steps of the method according to any one of claims 1 to 9.

Images