DETERMINATION OF BATTERIES' HEALTH STATUS USING A DYNAMIC DIAGNOSTIC LOAD PROFILE

Description

TECHNICAL AREA

[0001] Several examples of the revelation concern techniques for determining the health of rechargeable batteries. BACKGROUND

[0002] Electric vehicles use rechargeable traction batteries. As they age, the batteries' health can deteriorate, leading to a decrease in their capacity and affecting the reliability of the electric vehicle's operation.

[0003] Therefore, determining the health status of traction batteries can be desirable. Various techniques for determining this health status are known. These techniques acquire measurement data and, based on this data, determine the battery's health status. Many of these techniques require a comparatively long time to complete a diagnosis.

[0004] For example, US2019170827A1 discloses a method that, in practice, requires at least two complete charging cycles. For vehicles, a longer drive is necessary between two charging cycles, so that, due to the battery size, determining the battery's health status takes a long time.

[0005] For example, Stroe, Daniel I., et al., "Diagnosis of lithium-ion batteries state-of-health based on electrochemical impedance spectroscopy technique." 2014 IEEE Energy Conversion Congress and Exposition (ECCE). IEEE, 2014, describes a laboratory technique that requires the use of a special instrument for impedance spectroscopy.

[0006] For example, US2019170827A1 describes a method for determining the state of health (SOH) by repeatedly measuring the voltage during a charging process and using a Kalman filter. This method requires a large amount of measurement data and is therefore slow.

[0007] CN201911402190.9A describes a method intended to determine the SOH particularly quickly. However, it has been observed that the accuracy of the method is reduced. The same applies to US 2007 / 0236225 A1.

[0008] FR 3 044 424 A1 discloses a portable diagnostic device for a traction battery with measuring instruments for voltage, current, and temperature. The diagnostic device includes a control unit that uses a charging profile to determine the battery's state of health and its remaining service life. BRIEF SUMMARY

[0009] Therefore, there is a need for improved techniques to determine the health of rechargeable traction batteries in electric vehicles as quickly and accurately as possible. There is a need for techniques that allow the health of an entire battery pack in electric vehicles to be determined without complex measuring equipment, such as that used in impedance spectroscopy, for example, using a communication interface readily available in workshops.

[0010] This problem is solved by the features of independent patent claims.

[0011] The features of the dependent claims define embodiments. Using the techniques described herein, it is possible to determine the state of health of an electric vehicle's traction battery by means of a partial charging or discharging process. For this purpose, for example, a test bench equipped with a charger can be used. A diagnostic load profile with variable parameters can be employed. This profile is used to charge or discharge the traction battery. Measurement data acquired during the application of the diagnostic load profile, for example by the battery's battery management system (BMS), can be acquired and recorded via a suitable vehicle interface (e.g., OBD2 interface) using a user terminal (e.g., a computer). The measurement data can then be transmitted to a server in the cloud.The server can then use an estimation algorithm to estimate the state of health based on the measurement data.

[0012] A computer-implemented method involves obtaining type and state data of an electric vehicle's traction battery. The traction battery is connected to a test bench. The method further includes determining a diagnostic load profile based on the type and state data. This diagnostic load profile comprises several operating blocks, each associated with different load scenarios. The computer-implemented method also includes controlling a test bench to apply the diagnostic load profile. This generates measurement data for a voltage and current profile of the traction battery during a load process. Finally, the method includes determining an estimate of the traction battery's state of health based on these measurement data.

[0013] A computer program, a computer program product, or a computer-readable storage medium comprises program code. The program code can be loaded and executed by a processor. This causes the processor to execute a procedure. The procedure involves obtaining type data and state data of an electric vehicle's traction battery. The traction battery is connected to a test bench. The procedure further involves determining a diagnostic load profile based on the type data and the state data. The diagnostic load profile comprises several operating blocks. These operating blocks are associated with different load scenarios. The computer-implemented procedure also includes controlling a test bench to apply the diagnostic load profile. In this way, measurement data for a voltage profile and a current profile of a load process on the traction battery are obtained.Furthermore, the procedure also includes determining an estimate of the traction battery's health status based on the measurement data.

[0014] A device comprises a processor and memory. The processor can load and execute program code from memory. This causes the processor to execute a procedure. The procedure involves obtaining type data and state data of an electric vehicle's traction battery. The traction battery is connected to a test bench. The procedure further includes determining a diagnostic load profile based on the type data and the state data. The diagnostic load profile comprises several operating blocks. These operating blocks are associated with different load scenarios. The computer-implemented procedure also includes controlling a test bench to apply the diagnostic load profile. In this way, measurement data for a voltage profile and a current profile of a load process on the traction battery are obtained.Furthermore, the procedure also includes determining an estimate of the traction battery's health status based on the measurement data.

[0015] The features set out above and those described below can be used not only in the corresponding explicitly set out combinations, but also in further combinations or in isolation, without leaving the scope of protection defined by the claims of the present invention. BRIEF DESCRIPTION OF THE FIGURES

[0016] FIG. 1 schematically illustrates a device for data processing according to various examples. FIG. 2 This is a flowchart of an exemplary procedure. FIG. 3 This is a flowchart of an exemplary procedure. FIG. 4 shows an exemplary reference dependency of a parameter of a diagnostic load profile on the states of a traction battery. FIG. 5is a signal flow diagram that illustrates the communication between a test bench, a user terminal, a server and a database according to various examples. FIG. 6 schematically shows a diagnostic stress profile according to various examples. DETAILED DESCRIPTION

[0017] The properties, features and advantages of this invention described above, as well as the manner in which they are achieved, will become clearer and more easily understood in connection with the following description of the exemplary embodiments, which are explained in more detail in conjunction with the drawings.

[0018] The present invention is explained in more detail below with reference to preferred embodiments and the drawings. In the figures, identical reference numerals denote identical or similar elements. The figures are schematic representations of various embodiments of the invention. Elements depicted in the figures are not necessarily shown to scale. Rather, the various elements depicted in the figures are represented in such a way that their function and general purpose are understandable to a person skilled in the art. Connections and couplings between functional units and elements shown in the figures can also be implemented as indirect connections or couplings. A connection or coupling can be implemented as a wired or wireless connection. Functional units can be implemented as hardware, software, or a combination of hardware and software.

[0019] The following describes techniques for determining the state of health of a battery. The state of health can correspond to, for example, the battery's capacity. It could also correspond to the battery's impedance. Typically, as a battery ages, its capacity decreases and its impedance increases.

[0020] The techniques described herein enable a particularly quick yet reliable determination of the battery's health status. These techniques do not require any complex measuring equipment.

[0021] Therefore, the techniques described herein can be used particularly in connection with traction batteries of electric vehicles connected to a test bench. For example, the battery's state of health could be checked during a technical suitability test of the electric vehicle on a suitable test bench. In such applications, it is necessary that the diagnostic time (i.e., the time required to acquire the measurement data used to estimate the battery's state of health) is not too long in order to reduce the time required per vehicle.

[0022] The test bench can charge the batteries or, in some cases, discharge them. A corresponding diagnostic load profile can be used to apply a specific load to the battery during charging and / or discharging. Relevant measurement data—such as voltage and current values—can then be recorded, and an assessment of the battery's health can be made based on this data.

[0023] The measurement data can therefore include a voltage profile and a current profile, that is, voltage values ​​as a function of time and current values ​​as a function of time, each while the diagnostic load profile is applied.

[0024] As a general rule, it would be conceivable to provide the current profile and / or the voltage profile averaged across the various battery cells. Alternatively or additionally, it would also be possible to provide the current profile and / or the voltage profile resolved for individual battery cells, meaning that voltage and current are indexed in different battery cells. In this way, it could be possible to detect imbalances between the different battery cells and to indicate or correct the battery's health accordingly.

[0025] The measurement data can include other information, either as an alternative or in addition to the voltage or current profile. For example, the measurement data could indicate the battery's state of charge and / or its operating temperature.

[0026] The diagnostic load profiles used can have multiple operating blocks, each associated with different battery load scenarios.

[0027] Several examples are based on the understanding that different electrochemical processes within the battery cells can be distinguished by using different load scenarios. The voltage and current profiles can therefore contain characteristic features of these electrochemical processes. This makes it possible to determine the battery's health status with greater accuracy, taking into account the various relevant electrochemical processes.

[0028] Some examples of operational blocks are listed below in Table 1. Table 1: Various examples of operating blocks that together can form a load profile. Parameters are also given that characterize a corresponding operating block and, in some examples, can be determined in a type-specific and state-specific manner (details will be explained later). Operating block Explan parameter Resting phase During the rest phase, no charging or discharging takes place. This means that no external current is applied to the battery. Duration of the rest phase A resting phase can be used to prepare for a resting voltage measurement. This means that the battery can be in a relaxed state, and then, for example, a current pulse can be used to determine a value for the resting voltage. electrical pulse One or more current pulses can be used here, for example for rapid charging of the battery. - Time interval between electrical pulses - Duration / Number of electrical pulses A sudden increase in current flow to the load could also occur. - Plateau duration of the current pulses Overvoltages can be determined by using current pulses. This allows the time response of the system to be estimated, i.e., the time lag between the current pulse and the voltage change. - Amplitude of the current pulses Constant current flow In an operating block with constant current flow, a finite charging or discharging current with constant amplitude can be used over an extended period of time. - Amplitude - Duration The charging and discharging curves of lithium batteries differ (hysteresis). By using a constant current flow, the hysteresis effect can be compensated for, so that the battery voltage corresponds to the value on the charging or discharging curve. Furthermore, charging with a constant current allows for the creation of a controlled difference in charge states. This enables the setting of a specific charge state by charging or discharging the battery. Variable current flow In such an operating block, charging or discharging can be performed continuously with a current flow that is either greater or less than zero, while varying the amplitude of the current flow. For example, amplitude modulation could be achieved with a sinusoidal, rectangular, triangular, or sawtooth amplitude modulation pattern. Combinations of different patterns, even at different frequencies, would also be possible. - Amplitude modulation patterns (shape, frequency, etc.) - Duration - Amplitude

[0029] Table 2 below shows an example of a diagnostic stress profile. TABLE 2: An exemplary diagnostic load profile, composed of a specific sequence of operating blocks according to TABLE 1. phase Operating block Explan I Resting phase Measurement begins with a resting phase to determine the resting voltage of the battery system and the individual cells / modules II Constant current flow Charging with constant current to compensate for the hysteresis effect III electrical pulse Multiple pulses with intervening rest periods to determine occurring overvoltages, to evaluate the time constant of the impulse response IV Resting phase Determination of the resting voltage V Constant current flow SOC difference to Phase III VI electrical pulse Multiple pulses with intervening rest periods to determine occurring overvoltages, to evaluate the time constant of the impulse response VII Resting phase Resting phase to determine the resting voltage of the battery system and the individual cells / modules

[0030] Several examples are further based on the understanding that electrochemical processes (e.g., the hysteresis effect, reaching a resting state, etc.) can occur on different timescales and / or with varying intensity depending on the battery type and its condition. Electrochemical processes behave differently for different battery types. Even for batteries of the same type, different timescales and / or intensities of electrochemical processes can be observed, depending on the battery's condition, for example, temperature or state of charge.

[0031] To determine a diagnostic load profile based on these findings, minimizing diagnostic time while simultaneously capturing sufficient measurement data to reliably determine the battery's health status, the appropriate diagnostic load profile is determined based on type data and condition data of an electric vehicle's traction battery connected to a test bench. This means that the diagnostic load profile can be determined in a type-specific and condition-specific manner.

[0032] The type data is indicative of a battery type. For example, the type data could specify a manufacturer, a manufacturer's number, a serial number, etc. The type data could be indicative of the battery's cell chemistry, e.g., specifying an electrode material. The type data could, for example, be indicative of a cell geometry. The type data could indicate whether the battery has an active cooling system. In any case, the type data can be unchanging for the battery's lifetime; this means that the type data can be independent of the battery's current condition.

[0033] The battery's condition changes over time. The current state of the battery can be indicated by its condition data. For example, the condition data could be indicative of the battery's current state of charge (SOC), i.e., the relative charge of the battery in relation to its nominal capacity. The condition data could also indicate the battery's operating temperature.

[0034] By selecting the diagnostic stress profile based on the type and condition data, a suitable diagnostic stress profile can be chosen. This means selecting the shortest possible diagnostic stress profile, i.e., with the shortest possible diagnostic time. Nevertheless, the appropriate diagnostic stress profile can still be used to determine the battery's health status as accurately as possible.

[0035] Based on the diagnostic stress profile, the test bench can then be activated to apply the diagnostic stress profile. In this way, the measurement data can be obtained. Based on this measurement data, it is then possible to determine an estimate of the patient's health status.

[0036] As a general rule, different techniques can be used to determine the health status estimate. This means that different types of estimation algorithms can be employed. For example, an estimation algorithm could include a machine-learned model. Examples of machine-learned models include artificial neural networks, regression models, etc. See, for example: Roman, Darius, et al. "Machine learning pipeline for battery state-of-health estimation." Nature Machine Intelligence 3.5 (2021): 447-456. An example of an artificial neural network would be an LSTM network (see, for example, Qu, Jiantao, et al. "A neural-network-based method for RUL prediction and SOH monitoring of lithium-ion battery." IEEE Access 7 (2019): 87178-87191). However, the estimation algorithm does not necessarily have to include a machine-learned model.For example, conventionally parameterized models could also be used. Filtering techniques could be employed, such as a Kalman filter or a method for fitting predefined dependencies. Exemplary techniques are described, for example, in: Andre, Dave, et al. "Advanced mathematical methods of SOC and SOH estimation for lithium-ion batteries." Journal of power sources 224 (2013): 20-27. In various examples, a single-stage estimation algorithm can be used that directly estimates the state of health based on the measurement data. Alternatively, it would also be conceivable to estimate one or more values ​​for one or more battery parameters in a first stage (for example, to parameterize an equivalent circuit model). Then, in a second stage, the state of health can be estimated based on these one or more values.In such a scenario, the battery can be modeled using a physical model, such as an equivalent circuit model or a thermal dissipation model. See, for example, Chang, Fengqi, et al. "Modelling and evaluation of battery packs with different numbers of paralleled cells." World Electric Vehicle Journal 9.1 (2018): 8. As is evident from these explanations, the specific estimation algorithm is not crucial for the techniques described herein. In principle, any well-known estimation algorithm can be used.

[0037] FIG. 1 Figure 501 schematically illustrates a device 501 that can be configured to perform techniques related to the estimation of a health condition, as described herein. For example, the device 501 could be implemented by a computer. The device 501 could be implemented by a server.

[0038] The device 501 comprises a processor 504 and a memory 505. The device 501 also includes a communication interface 506, through which the processor 504 can communicate with other units. For example, the processor 504 could use the communication interface 506 to control a test bench to which an electric vehicle is connected, so that the test bench applies a diagnostic load profile. The processor 504 could receive measurement data via the communication interface 506. Examples of such measurement data would be, in particular, a voltage profile or a current profile. Other measurement data could also be received, for example, a profile concerning the state of charge, that is, the change in the state of charge as a function of time; and / or a profile concerning the temperature, that is, the operating temperature of the battery as a function of time. The processor 504 can load and execute program code from the memory 505.This causes the processor to execute techniques related to estimating battery health, as described herein. For example, the processor can determine a diagnostic load profile based on battery type and state data. The processor can also estimate the battery's health based on measurement data, for example, by executing one or more estimation algorithms.

[0039] FIG. 2 illustrates a flowchart according to various examples. The flowchart from FIG. 2 shows different phases related to the preparation and use of an estimation algorithm used to estimate a health condition. Fig. 2 It also shows various aspects related to the preparation for determining the diagnostic load profile depending on the condition and type of traction batteries.

[0040] First, one or more reference measurements are performed in Box 6000. These reference measurements can be used to characterize rechargeable traction batteries for electric vehicles. For example, the reference measurements can quantify certain physicochemical properties of the traction batteries. Reference dependencies between parameters of a diagnostic load profile and different states of traction batteries could be determined for different types of traction batteries, for instance, based on such a quantification of physicochemical properties.

[0041] Examples of physicochemical properties that can be quantified within the scope of reference measurements include: hysteresis voltage; battery impulse response; and open-circuit voltage. Such physicochemical properties can be determined for different battery states, such as different charge levels or operating temperatures, etc.

[0042] It is possible that the reference measurements also include the application of a reference diagnostic load profile. This allows training measurement data to be obtained. The reference diagnostic load profile can then serve as a template for a diagnostic load profile, which is used during the inference phase to capture corresponding measurement data.

[0043] Reference measurements can establish a basic understanding of the traction battery's health. This means that reference measurements can be used to determine, for example, the capacity or impedance. These reference measurements may differ from those taken when applying the diagnostic profile; for instance, laboratory procedures may be used to determine the battery's health in conjunction with the reference measurements.

[0044] InBox 6005 describes the parameterization of an estimation algorithm that can be used to determine the health status of traction batteries. This can be based on reference measurements. Based on these reference measurements, training measurement data can be obtained, which is then used to parameterize the estimation algorithm in Box 6005. Ground truths for the health status can also be determined based on the reference measurements. For example, the health status of a battery could be determined using the reference measurements and then mapped to corresponding training measurement data. These reference measurements could, for example, be performed in a laboratory. Typically, such laboratory measurements can determine the health status of batteries with high accuracy, for example, using electrical impedance spectroscopy, etc.

[0045] Box 6005 can, for example, include training for machine-learned algorithms. For instance, backward propagation techniques (gradient descent methods) could be used to determine the weights of an artificial neural network.

[0046] Manual parameterization is also possible.

[0047] Regardless of the specific mode of parameterization, it can be checked in Box 6005 that the output of the estimation algorithm (i.e., the estimation of the health status) corresponds as closely as possible to the basic truth.

[0048] Subsequently, a diagnosis can be made in Box 6010. This means that the assessment of the health status can be carried out without a corresponding underlying data being available. Box 6010 is also referred to as the inference phase. For this, one or more estimation algorithms, parameterized in Box 6005, can be used.

[0049] In principle, it would be conceivable that both Box 6005 and Box 6010 could be operated by a device such as Device 501 (compare FIG. 1 ) are executed. However, it would also be conceivable that different devices are used for executing Box 6005 and Box 6010.

[0050] Next, details related to Box 6010 will be provided in FIG. 3 described. FIG. 3 This is a flowchart of an exemplary procedure. FIG. 3 is a flowchart of a procedure for determining an estimate of the health status of a battery. FIG. 3This concerns the inference phase; that is, rules for determining the diagnostic burden profile depending on type and condition data are already known, for example, based on the reference measurements from Box 6000. Furthermore, an estimation algorithm is available that is parameterized and can be used to determine the health status estimate based on measurement data.

[0051] For example, the procedure could consist of FIG. 3 from device 501 FIG. 1 be carried out. In particular, the procedure could be carried out by FIG. 3 be executed from a server.

[0052] Box 6105 receives the type data of an electric vehicle's traction battery. The electric vehicle is connected to a test bench. The test bench establishes a data connection with the traction battery (for example, with the battery's management system) to read the type data or information indicative of the type data. The test bench can then transmit the type data to the server. The type data can identify the traction battery type. Certain type characteristics of the traction battery that are not directly dependent on operation can be indicated by the type data, such as serial number, manufacturer identification, cell geometry, number of cells, etc.

[0053] Box 6110 will receive status data from the electric vehicle's traction battery. For example, this status data can be obtained from the traction battery's management system via the data link between the test bench and the battery. The status data can describe the current state of the traction battery. For instance, the status data could describe the state of charge (SOC) of the traction battery. The status data could also indicate the operating temperature of the traction battery.

[0054] In Box 6115, a diagnostic load profile is then determined based on the type data and also on the condition data. The diagnostic load profile comprises several operating blocks, with the operating blocks being associated with different load scenarios. Exemplary operating blocks were described above in connection with Table 1. An exemplary diagnostic load profile is explained in Table 2.

[0055] This means that the diagnostic load profile can be specifically adapted to the respective type of traction battery and also to its current condition. This allows for the determination of a particularly short diagnostic load profile – meaning that the diagnostic load profile can include as few operating blocks as possible and / or that the operating blocks are particularly short. At the same time, the individual adjustment of the diagnostic load profile to the battery type and condition allows for a particularly precise determination of the battery's health status.

[0056] The diagnostic stress profile can be determined based on reference dependencies between parameters of the diagnostic stress profile and the condition data and / or the type data. Such reference dependencies can be based on the reference measurements (compare Fig. 2 : Box 6000).

[0057] The diagnostic workload profile can be determined in such a way as to minimize the diagnostic time. More generally, the diagnostic workload profile can be determined by considering the duration of the diagnostic workload.

[0058] The diagnostic burden profile can be determined in such a way as to maximize the accuracy of the health status assessment. More generally, the diagnostic burden profile can be determined taking into account the accuracy of the health status assessment.

[0059] Box 6120 then controls the test bench to apply the diagnostic load profile. The corresponding information is received from Box 6115. For example, specific parameters can be transmitted to the test bench, such as a sequence of operating blocks and corresponding parameter values ​​for the properties of the operating blocks, like duration, current amplitudes, etc. If different candidate diagnostic load profiles are already stored on the test bench, an index of the candidate diagnostic load profile to be activated can also be transmitted to the test bench.

[0060] Then, in connection with the activation of the test bench, corresponding measurement data is also obtained, including the voltage and current profiles of the traction battery's load process, which are associated with the diagnostic load profile. Alternatively or additionally, an operating temperature profile and / or a state-of-charge profile could also be obtained. The measurement data can be generated by the battery's management system and transmitted to the test bench.

[0061] In some examples, the measurement data can be continuously acquired, that is, during the application of the diagnostic stress profile. In other examples, it would also be conceivable that the measurement data are only acquired after the diagnostic stress profile has been completed. If the measurement data are acquired continuously, it would be conceivable to adjust the diagnostic stress profile before the end of the diagnostic period; that is, it would be possible to react directly to specific signatures in the measurement data and, for example, shorten or lengthen the diagnostic stress profile. Corresponding aspects will be discussed later in connection with Fig. 5 Box 5035 explained.

[0062] Based on such measurement data, the health status of the traction battery can then be estimated in Box 6125. One or more previously parameterized estimation algorithms can be used for this purpose (see [reference]). FIG. 2 : Box 6005).

[0063] As explained above, a predefined set of rules can be used to determine the diagnostic stress profile. In a simplified version, this predefined set of rules could be determined by expert knowledge. Two further examples are shown in Table 3. TABLE 3: Two exemplary strategies for determining the diagnostic stress profile. How is the set of rules for determining the diagnostic stress profile determined? Explan According to the parameterization of the estimation algorithm The rule set can be used in connection with the parameterization of the estimation algorithm, see below. FIG. 2 Box 6005. Training measurement data, based on a reference diagnostic stress profile, can be used to parameterize the estimation algorithm. A diagnostic stress profile can then be used that is as similar as possible to the reference diagnostic stress profile. This means that the reference measurement is replicated as accurately as possible during the inference phase. This means that a dependency between parameters of the diagnostic stress profile and the condition data and the type data can be chosen such that the diagnostic stress profile matches as closely as possible with the reference diagnostic stress profile that was used to capture the training measurement data with which the estimation algorithm was parameterized. Such techniques are based on the understanding that the parameterization of the estimation algorithm is adapted to a specific measurement scenario and that reproducing this scenario typically yields particularly accurate results. This is especially true for machine-learned models, such as artificial neural networks. For machine-learned models, the accuracy of the prediction is particularly high when the input data during inference is comparable to the input data during training. Based on the chemical and physical properties of the battery The rule set can be determined based on the physicochemical properties of the battery, whereby these physicochemical properties are determined, for example, using reference measurements or based on literature knowledge. For example, reference measurements can be used to determine fundamental dependencies between the physicochemical states of the battery. This approach has the advantage that statements can also be made about areas of the condition data or the type data for which no reference measurements have been recorded. This means, for example, that interpolation can be used to determine a suitable diagnostic stress profile for a specific value of the condition data. For example, reference measurements might reveal that a particular battery type exhibits a particularly high hysteresis voltage. In this case, it may be necessary to adjust the diagnostic load profile accordingly to compensate for the high hysteresis voltage. This could result in a longer diagnostic time for that battery type. However, another battery type might not have a similarly high hysteresis voltage, and the diagnostic time for such a battery type could then be reduced. The reference measurements can also be used to parameterize the estimation algorithm (see...). FIG. 2 : Box 6005).

[0064] Specific implementations of the strategies from Table 3 are described below.

[0065] For example, it would be conceivable that the operating blocks of the diagnostic load profile could be selected from a large number of predefined operating blocks. This means that, in a modular principle, different, predefined operating blocks could be combined to form the diagnostic load profile.

[0066] This has the advantage that individual parameters of the various operating blocks can be predefined, and only the predefined operating blocks need to be chained together. In particular, predefined operating blocks can be used for which parameters were already defined during the parameterization of the estimation algorithm (see below). FIG. 2 (Box 6005) training measurement data were recorded. In other words, comparable load scenarios can be applied to the actual diagnosis in Box 6010 and the parameterization in Box 6005. This allows for high accuracy in determining the state of health, while still taking into account different boundary conditions, such as batteries of different types or in different states.

[0067] In addition to selecting the various operating blocks, determining the diagnostic load profile according to Table 3: Scenario 1 can also include determining the sequence of the multiple operating blocks. In other words, the sequence and / or the number of different operating blocks used in the diagnostic load profile can be varied.

[0068] It may also be possible to parameterize one or more parameters of at least one of the multiple operating blocks. For example, in connection with Table 1 above, various parameters for the different operating blocks were presented. Such parameters may relate to, for example: the duration of the corresponding operating block; the discharge rate or charge rate (also referred to as C-rate) of a corresponding discharge or charge current; the amplitude of current pulses; etc.

[0069] Next, in connection with FIG. 4A predefined reference dependency is explained, which can be taken into account when determining the diagnostic load profile. For example, it would be possible to parameterize one or more parameters of at least one operating block based on such a dependency.

[0070] In FIG. 4 Figure 800 shows the relationship between the maximum possible hysteresis voltage and the state of charge of the battery for a specific battery type. This relationship characterizes a chemical-physical property of the respective battery type. This relationship is shown for several temperatures (solid and dashed lines).

[0071] The maximum possible hysteresis voltage affects the time required to reach the open-circuit voltage characteristic in the charging direction.

[0072] The maximum possible hysteresis voltage is therefore directly proportional to the duration of an operating block of the diagnostic load profile with constant current flow (see Table 1). Therefore, it shows FIG. 4 a dependency between the corresponding parameter of the diagnostic load profile and the corresponding state data (SOC, temperature) and type data (battery type "XYZ").

[0073] In FIG. 4 This shows that lower maximum hysteresis voltages are expected for higher temperatures, and therefore the time required to reach this maximum voltage can also be reduced. Furthermore, in FIG. 4It has been shown that lower maximum hysteresis voltages are expected for higher charge states, thus reducing the time required to reach the target voltage. This means that, depending on the charge state and temperature of a battery of a corresponding type – the charge state and temperature are provided via the state data – and taking into account the dependency of FIG. 4 , the duration of the charging phase with constant current flow can be adjusted.

[0074] In FIG. 4 This is just one example of a dependency between parameters of the diagnostic stress profile and condition or type data. Further examples of such dependencies are summarized below in Table 4. Table 4: Various examples of dependencies between parameters (more precisely: parameter values) of the diagnostic load profile and condition data; these dependencies differ for different batteries. As a general rule, such predefined dependencies can be determined based on reference measurements (see Box 6000 in Fig. 2). The different examples can be considered individually or in combination. dependence Details Hysteresis voltage / duration of charging phase - state of charge, battery type and / or temperature (see below). FIG. 4 ) A change in the charge state can result in a certain hysteresis voltage. This hysteresis voltage can distort subsequent measurements of the open-circuit voltage. Therefore, it may be necessary to provide compensation for the hysteresis voltage. This hysteresis voltage varies depending on the battery type. The hysteresis voltage changes depending on the battery's state of charge. Furthermore, the hysteresis voltage changes depending on the cell temperature. By taking this dependency into account, the diagnosis time can be chosen to be as long as necessary, but as short as possible. State of charge difference - Battery type Depending on the battery type, it may be necessary to provide a different offset between charge states. To obtain significant results, a relatively small offset, for example of 5 percentage points, may be sufficient for some battery types; while other battery types require a larger offset between charge states. Accordingly, the charging phase can then be shortened or lengthened. Dynamic response - battery type, state of charge and / or temperature As discussed above in connection with Table 2, measuring the battery's dynamic response (e.g., impulse response) to current pulses at different charge levels can be helpful in detecting overvoltages. The duration of the current pulses and the subsequent pause depend on the battery type, the current charge level, and the cell temperature.

[0075] As a general rule, within the framework of such dependencies (see Table 4), a respective expected value for the accuracy of the health status assessment could also be considered (i.e., a confidence value). This means that the diagnostic burden profile is determined taking into account the respective expected value for the accuracy of the health status assessment. For example, in FIG. 4A tolerance range is specified for a charge state of 20%. This tolerance range could, for example, be associated with a predefined interval for the expected accuracy of the health state estimation. For instance, by selecting the duration of the corresponding operating block within the tolerance range, an accuracy of at least 80% of optimal accuracy (indicated by the solid line) could be achieved. Thus, tolerances can be provided that can be exploited, for example, when a further reduction in diagnostic time is desired.

[0076] As a general rule, such an expected value for the accuracy of the health estimate could be obtained from reference measurements for corresponding battery types. For example, an estimation algorithm could be tested using measurement data from reference measurements, thus specifying the accuracy range / variance of the algorithms. There are also estimation methods (Kalman filters, neural networks) that inherently provide an accuracy value for the estimate.

[0077] In In such scenarios, it is therefore possible to determine the diagnostic stress profile taking into account both the duration and the accuracy of the diagnostic stress profile. For example, by exploiting tolerances, the duration can be specifically increased or the accuracy specifically reduced.

[0078] Based on Table 4, it is evident that scenarios can arise where conflicting objectives exist. For example, to achieve the highest possible accuracy in estimating health status, it may be desirable to include certain operational blocks or to parameterize them in a specific way; conversely, to minimize diagnosis time, it may be advantageous to omit these operational blocks or to parameterize them differently. As a general rule, the diagnostic workload profile could be determined using iterative numerical optimization. This optimization could have an objective function that penalizes both longer diagnosis times and lower accuracy in estimating health status.This means that the conflict of objectives can be resolved by using iterative numerical optimization, striving for both a particularly short diagnosis time and the highest possible accuracy in estimating the health status. Iterative numerical optimization could be implemented, for example, using a gradient descent method, a simplex optimization algorithm, or genetic algorithms. The objective function can be defined taking into account predefined dependencies, such as those mentioned above in connection with... FIG. 4 or as explained in Table 4, will be formulated.

[0079] Using the aforementioned techniques, it is therefore possible to achieve a shorter diagnostic time while simultaneously ensuring high accuracy in estimating the battery's aging state by determining the diagnostic load profile specific to the battery type and condition. Overall, it was found that high accuracy in estimating the aging state can also be achieved when the battery is partially charged or partially discharged.

[0080] For example, the techniques described herein may eliminate the need to fully discharge or fully charge the battery when applying the diagnostic load profile. Instead, a partial discharge or partial charge of the battery can be performed. This means that the state of charge (SOC) can change by a few percentage points, specifically by less than 100 percentage points. For instance, the state of charge could change by less than 15 percentage points, meaning, for example, a charge from 40% SOC to 55% SOC could be performed (this is referred to as a charge increment, ΔSOC). This can reduce the diagnostic time. Generally, the charge increment could be less than or equal to 50 percentage points, optionally less than or equal to 20 percentage points, and further optionally less than or equal to 15 percentage points.

[0081] Such a shortened diagnostic time enables specific application scenarios. For example, in the context of a technical suitability test of motor vehicles—where the time per vehicle is typically very limited to ensure a sufficient number of vehicles can be inspected—the health of the traction batteries can be estimated when the vehicles are connected to a test bench. A corresponding application scenario is described below in connection with FIG. 5 described. FIG. 5 This is a signal flow diagram illustrating the communication between a test bench 91, to which an electric vehicle 90 is connected, a user terminal 93 (e.g., a computer or laptop) operated by a user 92, a server 94, and a database 95. The communication according to FIG. 5 For example, the procedure can consist of FIG. 3 implement.

[0082] In the example shown, Server 94 implements the logic for determining the diagnostic load profile (Box 5025) and for determining the health status (Box 5055). Server 94 can provide a corresponding service for a large number of test benches.

[0083] For example, server 94 can be connected to user terminal 93 via the internet, and user terminal 93 can be connected to test bench 91 via a local network connection. Test bench 91 can be connected to a BMS of the traction battery of electric vehicle 90 via a diagnostic connection, for example via a vehicle bus system.

[0084] The use of test bench 91 makes it possible to specify a precisely defined diagnostic load profile. In real-world traffic, only indirect specifications would be possible (e.g., stopping at certain points, specifying target speeds).

[0085] First, setup 5005 (initialization) of a diagnostic connection between the test bench 91 and the user terminal 93 is performed when the vehicle 90 is connected to the test bench 91 and when a corresponding diagnostic protocol is initiated by the user 92 at the user terminal 93. For example, the test bench 91 can exchange data with a battery management system of the traction battery of the vehicle 90 via a vehicle bus system. As part of setup 5005, status data 81 and type data 82 can be transmitted to the user terminal 93.

[0086] The type data could also be entered, at least partially, by user 92 at user terminal 93. The user selects, for example, the vehicle model including battery type and size.

[0087] Then, in box 5006, one or more boundary conditions for the state data 81 can be determined, for example, based on the type data 82. It can be checked whether the state of the traction battery is compatible with the one or more boundary conditions.

[0088] For example, it could be checked whether a specific minimum or maximum charge level is reached or exceeded. Such a minimum or maximum charge level can vary depending on the battery type. It could also be checked whether the battery's operating temperature is within a specific predefined range. Finally, it could be checked whether a predefined minimum resting period has been observed.

[0089] Box 5006 can therefore be used to perform a preliminary check to determine whether the condition of the traction battery of the electric vehicle 90 is fundamentally suitable to allow a quick assessment of its health.

[0090] If the verification in box 5006 is successful, a request 5010 can then be sent to server 94. Request 5010 contains the status data 81 and the type data 82.

[0091] While in the illustrated example the FIG. 5 If the status data 81 and the type data 82 are transferred from the user terminal 93 to the server 94, it would generally also be possible for the status data 81 and the type data 82 to be transferred directly from the test bench 91 to the server 94.

[0092] Server 94 can then send a corresponding request 5015 to database 95. Optionally, this request 5015 can include the state data 81 and / or the type data 82.

[0093] Then, a message 5020 is transmitted from database 95 to server 94, which is indicative of one or more references. For example, predefined dependencies, such as those mentioned above in connection with FIG. 4 and TABs. 3 and 4 are discussed and transferred.

[0094] The diagnostic load profile can then be determined in box 5025. References according to message 5020 can be taken into account.

[0095] The diagnostic load profile can then be transmitted to test bench 91 with a message 5026, for example directly from server 94 or via user terminal 93.

[0096] Test bench 91 then applies the diagnostic load profile over a specific period of time, which in FIG. 5is marked with reference number 5028. It is possible that the test bench 91 transmits measurement data multiple times in corresponding messages 5030 to the server 94. The measurement data are indicative of a voltage profile and a current profile of the load process of the traction battery of vehicle 90, which is caused by the diagnostic load profile.

[0097] In FIG. 5This illustrates a scenario in which the measurement data is transmitted to server 94 even before the diagnostic load profile is completed. This allows for optional monitoring of the measurement data in box 5035 during the application of the diagnostic load profile. The diagnostic load profile can then be adjusted based on this monitoring using a corresponding message 5040. For example, the preliminary measurement data might reveal that additional operating blocks of the load profile are needed to obtain a reliable health assessment. The diagnostic load profile could then be extended ad hoc by sending message 5040. Monitoring the measurement data can specifically include monitoring for voltage profile deviations between different traction battery cells.Such techniques are based on the understanding that a deviation in the voltage profiles of different battery cells can indicate varying degrees of aging. In such scenarios, it is typically necessary to acquire more detailed measurement data that allows for a breakdown of the health status of the individual battery cells. For example, a corresponding operating block of the diagnostic load profile, enabling further differentiation between different battery cells, could be selectively activated when the measurement data indicates a corresponding deviation in voltage values ​​for different battery cells.

[0098] In various scenarios, it would be conceivable that the estimation algorithm for determining the health status of vehicle 90's traction battery is permanently stored in server 94. This means that the same estimation algorithm can be used regardless of the battery type and / or its condition.

[0099] In other examples, it would also be conceivable that, for example depending on the type of data and / or the state data, an estimation algorithm is selected from several candidate estimation algorithms. In the scenario of FIG. 5 Server 94 sends a corresponding request 5045 to database 95. Database 95 then sends a message 5050, which indicates the specific estimation algorithm. Request 5045 could, for example, contain the type data 82.

[0100] Such techniques are based on the realization that sometimes certain preferred types of estimation algorithms are not available for all types of batteries, e.g. because they have not yet been appropriately parameterized (cf. FIG. 2 (Box 6005). For example, a machine-learned model might only be available for some battery types because no corresponding training has yet taken place. If no specifically signed machine-learned model is available, a general-purpose estimation algorithm can be used, which may inherently provide lower accuracy.

[0101] Box 5055 then determines the health status using the appropriate estimation algorithm. Multiple estimation algorithms could also be combined. The results could then be compared, which can increase accuracy.

[0102] The result of the estimation can be transmitted to user terminal 93 using message 5060. The result of the diagnosis can be output to user 94.

[0103] FIG. 6 This illustrates aspects related to an exemplary diagnostic load profile 700. The diagnostic load profile 700 comprises several operating blocks 701-709 (see Table 1). During operating block 701, a constant current flow is used to charge the battery. This is followed by a rest phase in operating block 702. In operating block 703, a variable current flow is used. This is followed by another rest phase in block 704. Subsequently, current pulses are applied in operating block 705. Blocks 706-709 then correspond to a repetition of blocks 701-705.

[0104] In FIG. 6 Corresponding measurement data for voltage profile 751 and current profile 752 are also shown.

[0105] In summary, the preceding sections described techniques that enable the rapid acquisition of measurement data for assessing a person's health status. These techniques can be performed in workshops, driver fitness centers during repairs, or during fitness-to-drive assessments. Only a single partial charging process is required. Ensuring comparable testing conditions during data recording (which is not possible during driving or only achievable with considerable effort using a roller dynamometer) is possible.

[0106] Naturally, the features of the embodiments and aspects of the invention described above can be combined with one another. In particular, the features can be used not only in the combinations described, but also in other combinations or individually, without leaving the scope of the invention.

[0107] For example, techniques have been described above in which a diagnostic load profile is determined based on the condition and type data of a traction battery. Alternatively or additionally, other decision criteria could be considered when determining the diagnostic load profile. For instance, user input could be taken into account. The user could specify a desired voltage, power, battery charge, or battery temperature to serve as the basis for estimating the battery's health. The diagnostic load profile could then be determined such that measurement data is recorded at the corresponding target values. Another aspect concerns obtaining condition data from the test bench. For example, the test bench could signal the ambient temperature of the vehicle.The ambient temperature could be taken into account when determining the diagnostic stress profile.

[0108] Furthermore, techniques have been described above in which a load diagnostic profile is used to charge the traction battery. It would correspondingly be possible to use a load diagnostic profile to discharge the traction battery. Combined charging and discharging would also be conceivable.

[0109] Furthermore, aspects related to determining the health status of traction batteries in electric vehicles were described. In general, however, it would be possible to apply the corresponding techniques to other types of batteries or other application scenarios. For example, it would be conceivable to use the techniques described here for batteries in stationary energy storage systems. Even in such an application scenario, it may be desirable to interrupt normal operation only briefly, meaning that the diagnostic time should be kept as short as possible.

Claims

1. Computer-implemented method, the method comprising: - obtaining type data (82) and state data (81) of a traction battery of an electric vehicle (90) connected to a test bench (91), - based on the type data (82) and based on the state data (81), determining a diagnostic load profile (700) comprising multiple operational blocks (701-709) associated with different load scenarios, - actuating the test bench (91) to apply the diagnostic load profile (700) so as to obtain measurement data for a voltage profile (751) and a current profile (752) of a loading process of the traction battery, and - determining an estimate of a state of health of the traction battery based on the measurement data.

2. Computer-implemented method according to claim 1, wherein determining the diagnostic load profile (700) comprises at least one of the following: selecting the multiple operational blocks (701-709) from a plurality of multiple predefined operational blocks; determining a sequence of the multiple operational blocks (701-709); and / or parameterizing one or more parameters of at least one operational block of the multiple operational blocks (701-709), wherein the one or more parameters of the at least one operational block are selected from a group comprising: duration of the corresponding operational block; discharge or charge rate of a discharge or charge current; charging stroke; amplitude of a current pulse.

3. Computer-implemented method according to claim 1 or 2, wherein the diagnostic load profile is determined based on one or more predetermined dependencies (800) between parameters of the diagnostic load profile (700) and the state data (81) and / or the type data (82), wherein the one or more predetermined dependencies are determined based on reference measurements that detect physicochemical properties of the respective battery type.

4. Computer-implemented method according to claim 3, wherein the estimate is determined by means of an estimation algorithm, wherein the one or more predetermined dependencies are determined based on training measurement data used to parameterize the estimation algorithm.

5. Computer-implemented method according to claim 3 or 4, wherein the one or more predetermined dependencies (800) are associated with expected values for accuracy of the estimate of the state of health, wherein the diagnostic load profile is determined by taking into account the expected values for the accuracy of the estimate of the state of health.

6. Computer-implemented method according to any one of the preceding claims, wherein the diagnostic load profile is determined by taking into account a time period of the diagnostic load profile (700) and / or an accuracy of the estimate of the state of health, wherein the diagnostic load profile is determined using an iterative numerical optimization that takes into account a target function that penalizes longer time periods of the diagnostic load profile (700) and penalizes lower accuracies of estimate of the state of health.

7. Computer-implemented method according to any one of the preceding claims, wherein the method further comprises: - while applying the diagnostic load profile (700), monitoring the measurement data and optionally adjusting a remaining portion of the diagnostic load profile (700) based on monitoring the measurement data, wherein monitoring the measurement data comprises monitoring a deviation of voltage values of the voltage profile between different battery cells of the traction battery.

8. Computer-implemented method according to any one of the preceding claims, wherein the method further comprises: - based on the type data (82), selecting an estimation algorithm from a plurality of candidate estimation algorithms, wherein the estimate is determined using the selected estimation algorithm.

9. Computer-implemented method according to any one of the preceding claims, wherein determining the diagnostic load profile (700) and determining the estimate of state of health is performed by a central server (94) connected to a plurality of test benches (91), the plurality of test benches (91) comprising the test bench (91).

10. Computer-implemented method according to any one of the preceding claims, wherein the method further comprises: - determining one or more boundary conditions for the state data (81) and checking whether the state of the battery is compatible with the one or more boundary conditions, wherein the one or more boundary conditions are determined based on the type data (82).

11. Computer-implemented method according to any one of the preceding claims, wherein the diagnostic load profile defines a partial charge of the battery that increases a state of charge of the battery by no more than 15 percentage points.

12. Computer-implemented method according to any one of the preceding claims, wherein the method further comprises: - obtaining further state data (81) of the test bench (91), wherein the diagnostic load profile is further determined based on the further state data (81).

13. Computer-implemented method according to any one of the preceding claims, wherein determining the diagnostic load profile comprises: based on a state of charge of the traction battery indicated by the state data, determining a state-of-charge difference of the state of charge, wherein the state-of-charge difference between a beginning and an end of the diagnostic load profile is caused by applying the diagnostic load profile.

14. Computer-implemented method according to any one of the preceding claims, wherein determining the diagnostic load profile comprises: based on a state of charge and / or a temperature of the traction battery indicated by the state data, determining a duration of a charging phase of the diagnostic load profile.

15. Computer program comprising program code, wherein the execution of the program code by a processor causes the processor to perform the method comprising: - obtaining type data (82) and state data (81) of a traction battery of an electric vehicle (90) connected to a test bench (91), - based on the type data (82) and based on the state data (81), determining a diagnostic load profile (700) comprising multiple operational blocks (701-709) associated with different load scenarios, - actuating the test bench (91) to apply the diagnostic load profile (700) so as to obtain measurement data for a voltage profile (751) and a current profile (752) of a loading process of the traction battery, and - determining an estimate of a state of health of the traction battery based on the measurement data.