Method and device for operating a battery system in a motor vehicle

Abstract

The invention relates to a method for operating a vehicle battery (3), in particular to perform a balancing cycle, wherein in the balancing cycle at least one battery cell of the vehicle battery having a higher battery cell charge than at least one of the remaining battery cells is discharged by a specified degree, the method having the following steps: - confirming (S3) the start of an idle phase of the vehicle battery (3), during which the vehicle battery (3) is in a currentless state; - determining (S6) the expected duration of the idle phase by means of a data-based operating behavior model from the current calendar time reading; - performing (S8) the balancing cycle if the expected duration of the idle phase exceeds a specified minimum balancing duration.

Description

Technical Field

[0001] This invention relates to electric motor vehicles, and more particularly to battery pack systems for electric motor vehicles. The invention also relates to a method for optimizing the passive charge balance of battery cells used in vehicle battery packs through a balancing cycle. Background Technology

[0002] Vehicle battery packs used to operate electrically powered motor vehicles typically consist of multiple battery cells. These battery cells are usually connected in series and / or parallel. During operation, slight variations in discharge behavior due to component variations cause deviations in the charge or state of charge of each battery cell over time. However, to ensure optimal operation of such a vehicle battery pack, large differences in battery cell charge need to be avoided. For this purpose, so-called balancing cycles are performed at regular intervals to equalize the charge levels of the battery cells.

[0003] The balancing cycle can either follow an active strategy in which charges move between the cells in these battery packs in order to achieve charge balance.

[0004] However, a more common approach provides a passive balancing cycle, in which the charge differences between the cells in these battery packs are compensated for by converting the charge of the battery pack with the higher charge into waste heat through a balancing resistor. This reduces the charge of the battery pack with the higher charge.

[0005] Passive balancing cycles place a high load on the components of a vehicle battery pack. In particular, the heavy load on the battery cells, balancing resistors, and switching elements (semiconductor devices) used to connect or disconnect these balancing resistors to the battery cells during activation and deactivation significantly accelerates battery pack aging. Therefore, it is necessary to reduce the number of passive balancing cycles and maximize the duration of balancing operation. Summary of the Invention

[0006] According to the present invention, a method for operating a vehicle battery pack in an electrically driven motor vehicle as described in claim 1 is provided, as well as an apparatus and battery pack system as described in the parallel claims.

[0007] Other design options are described in the dependent claims.

[0008] According to a first aspect, a method is provided for operating a vehicle battery pack, particularly for performing a balancing cycle, wherein in the balancing cycle, at least one battery pack having a higher battery charge than at least one battery pack in the remaining battery packs is discharged to a specified degree, the method comprising the following steps: - Confirm the start of the idle phase of the vehicle battery pack, during which the vehicle battery pack is in a no-current state; - Based on the current calendar time readings, the expected duration of the idle phase is determined using a data-driven operational behavior model; - If the expected duration of the idle phase exceeds the specified minimum balancing duration, a balancing loop is executed.

[0009] The following issues exist: on the one hand, the charge levels of the battery cells in the vehicle battery pack should not deviate too much from each other; on the other hand, the number of balancing cycles used to protect the components of the vehicle battery pack should be kept as low as possible. Balancing cycles are typically used to compensate for the differences in charge levels between the battery cells.

[0010] In these balancing cycles, charge is extracted from battery cells with higher charge levels to balance their charge levels with the remaining battery cells. This charge extraction is achieved by connecting a balancing resistor, which is designed as a power resistor and converts the flowing discharge current into heat loss. These balancing cycles are performed during the vehicle's battery pack idle periods—that is, when the battery pack is in a no-current state—because only under no-current conditions can the battery cell voltages be accurately measured, and these voltages serve as the basis for determining the charge levels of each individual battery cell.

[0011] However, because these balancing cycles place a high load on the components of the vehicle battery pack, the number of balancing cycle activations should be minimized as much as possible. In particular, the number of balancing cycle activations should be limited by avoiding activation within a time window that is too short for the complete execution of a balancing cycle. Since balancing cycles typically last for a minimum duration between 15 and 60 minutes, there is a risk that the vehicle may begin driving or the battery pack may begin charging during this balancing cycle, i.e., begin current extraction or supply.

[0012] The idea behind this method is to provide an operational behavior model that describes the vehicle's operational behavior and is used to determine what downtime is expected to be long enough to perform a balancing cycle, and what downtime is not. This operational behavior model can be provided in several ways. The operational behavior model is designed or trained to describe information about the average disconnection time of the vehicle's battery pack based on calendar time readings, that is, date, day of the week, clock time, etc.

[0013] The operational behavior model can be specified or learned automatically based on monitoring of vehicle driving and usage data. Therefore, when the vehicle stops operating and the start of an idle phase (the state of no current in the vehicle's battery pack) is identified, a query in the operational behavior model can determine whether sufficient time is expected to be available to complete a full balancing cycle.

[0014] In this way, the occurrence of the balancing cycle being interrupted shortly after its activation and having to restart during the subsequent idle phase can be reduced, resulting in lower loads on the vehicle battery pack components and thus longer component lifespan. Furthermore, the duration for performing the balancing cycle to reduce battery charge deviations is maximized. Because the operating behavior model learns the individual operating behavior of the corresponding vehicle—that is, learns the individual operating behavior of the vehicle in relation to the vehicle—it can also better account for special usage conditions, such as those of taxis, ambulances, etc.

[0015] It can be stipulated that if the expected duration of the idle phase is less than the minimum balancing duration, the execution of the balancing cycle should be prevented. This is especially applicable when the balancing cycle is supposed to start at regular intervals or at periodic points according to the operating mode. Then, the above-described steps prevent the activation of the balancing cycle.

[0016] In one implementation, observation datasets can be collected during the operation of the vehicle battery pack, wherein these observation datasets assign calendar time readings of the start of detected idle periods to the duration of those idle periods, and these observation datasets are used to create or train operational behavior models.

[0017] The operational behavior model may in particular include a clustering method in which the observed dataset is clustered into one or more clusters, wherein each cluster defines the expected duration of a calendar time period and an idle period, specifically by means of the distance from the centroid of the relevant cluster and the expected duration of the idle period assigned to the centroid of the relevant cluster; wherein the expected duration of the idle period of the relevant cluster is determined for the current calendar time reading if the current calendar time reading falls within the calendar time period of one of these clusters.

[0018] Furthermore, if the current calendar time reading is outside the calendar time period determined by these clusters, a balancing cycle can begin after the idle phase has started and after a specified relaxation period.

[0019] In particular, the balancing cycle can only begin after a specified relaxation period following the start of the idle phase, when the change in the battery charge of the vehicle's battery pack exceeds a specified increase level.

[0020] Alternatively, if the current calendar time reading is outside the calendar time period determined by these clusters, the balancing loop cannot be executed or the execution of the balancing loop can be prevented.

[0021] It can be stipulated that the balancing cycle is only executed when the change in the battery charge of the vehicle's battery pack exceeds a predetermined level.

[0022] It can be stipulated that if the number of observation datasets outside the clusters determined by this clustering method exceeds a specified outlier threshold, the running behavior model is discarded.

[0023] According to another aspect, an apparatus for performing one of the above methods is provided. Attached Figure Description

[0024] The embodiments are then described in more detail with reference to the accompanying drawings. Wherein: Figure 1 A schematic diagram of an electric motor vehicle is shown, which has a battery pack system having a vehicle battery pack. Figure 2 A flowchart illustrating methods for operating vehicle battery packs, particularly for performing balancing cycles, is shown; and Figure 3 A diagram illustrating the clustering of the observed dataset for k-nearest neighbor classification of the start time and duration of the idle phase for motor vehicles is shown. Detailed Implementation

[0025] Figure 1 An electric motor vehicle 1 is schematically shown, which has an electric drive motor 2, a vehicle battery pack 3, and a control unit 4. The vehicle battery pack 3 provides the electrical energy required to operate the drive motor 2. The control unit 4 may correspond to a conventional microprocessor-controlled control unit. The control unit 4 may also be part of the battery pack management system of the vehicle battery pack 3.

[0026] The vehicle battery pack 3 has multiple battery cells 31, which are connected in parallel and / or in series with each other in a suitable manner to provide operating voltage for the operation of the drive motor 2.

[0027] For the vehicle battery pack 2 to function properly, the battery cells 31 need to have the same or substantially the same battery charge to avoid excessive load on each battery cell during operation, i.e., when extracting or supplying electrical energy. For this reason, a balancing function is implemented in the control device 4. This balancing function compensates for the charge of the battery cells 31 after activating the balancing cycle, so that the battery cells 31 with higher charge are discharged through one or more balancing resistors 32, so that the battery charge of the relevant battery cell 31 is balanced with the battery charge of the remaining battery cells 31 (with lower charge).

[0028] The discharge of the relevant battery pack cells 31 is achieved via corresponding balancing resistors 32, which are connected to these battery pack cells 31 via appropriate balancing switches (semiconductor switches) 33. If a balancing cycle is activated, a discharge current flows through the balancing resistors 32, and electrical power is converted into heat loss. Due to the high current flowing through the balancing switches 33 and balancing resistors 32 during this balancing process, the lifespan of these balancing switches 33 is limited to the number of switching cycles. Therefore, to extend the lifespan of the vehicle battery pack 3, one solution is to reduce the number of balancing cycle activations to avoid premature wear of the balancing switches 33.

[0029] The execution criteria can be implemented in control unit 4. Figure 2 The flowchart describes the method in more detail. This method can be implemented in a control unit in hardware and / or software.

[0030] In step S1, the charge level of each battery cell 31 in the vehicle battery pack 3 is first measured. As a result, a charge level reading is prepared for each battery cell 31, indicating the charge level of the corresponding battery pack.

[0031] Battery capacity measurement can be achieved, for example, by measuring the terminal voltage of the battery pack in a no-current state, wherein the terminal voltage is, if necessary, also allocated to the battery pack capacity in a manner known per se, based on the measured battery pack temperature. Methods for determining the battery pack capacity are also known, and these methods can be performed during the active operation of the vehicle battery pack 3. This step can be performed periodically, particularly during each idle phase of the vehicle battery pack 3.

[0032] In step S2, it is first checked whether battery pack charge balancing of these battery packs 31 is necessary. This can be confirmed, for example, if the deviation in battery pack charge between the battery pack 31 with the highest charge and the battery pack 31 with the lowest charge exceeds a predetermined percentage of the maximum charge, such as between 5% and 10%. The necessity of performing a balancing cycle can also be explained by the variance of the distribution of battery pack charge of these battery packs 31, especially when the variance exceeds a specified threshold.

[0033] If it is confirmed in step S2 that a balanced loop is needed (either of two options: yes), the method continues to step S3; otherwise, it jumps back to step S1.

[0034] In step S3, it is checked whether a balancing cycle can be performed. Typically (that is, when the execution of a balancing cycle is not particularly urgent), a balancing cycle is only permitted during idle periods, i.e., during the period when the vehicle battery pack 3 is in a current-free state. In particular, a balancing cycle cannot be performed during the operation of the vehicle 1, during the charging process, or during a predetermined period after the end of vehicle operation or the end of the charging process (due to the relaxation time). The predetermined time since the last current extraction from or supply to the vehicle battery pack 3 can be, for example, between 15 and 45 minutes, preferably about 30 minutes, and is used to provide sufficient relaxation time to these battery pack cells 31.

[0035] If it is confirmed in step S3 that the execution of the balanced loop is allowed (either option: yes), the method continues to step S4; otherwise (either option: no), it jumps back to step S3.

[0036] In step S4, it is checked whether there is sufficient training level or sufficient empirical data to support the use of the running behavior model. If there is sufficient training level for the running behavior model or the running behavior model is available (either one: yes), the method continues to step S5; otherwise, the method continues to step S11.

[0037] Subsequently, the operational behavior model is used to check whether sufficient remaining downtime or inactivity time for vehicle battery pack 3 can be accounted for. To this end, the operational behavior model is evaluated, which can explain, in a data-driven manner, whether the current idle phase lasts long enough based on current calendar time readings.

[0038] The operational behavior model can be designed as a trained, data-based model that describes the average duration of the vehicle's battery pack 3's idle periods based on calendar time readings, such as time of day, calendar date, day of the week, day of the month, or day of the year.

[0039] This operational behavior model can be trained on an observation dataset derived from past vehicle usage. This observation dataset can be created by recording the calendar time readings (clock time, day of the week, day of the month, day of the year, etc.) at which the vehicle entered a dormant state (i.e., was placed in a dormant state) and the duration until the vehicle is reactivated (the duration of the idle phase) each time the vehicle stops (enters a dormant state) and begins operation (exits a dormant state), such as starting to drive or begin charging. Therefore, the observation dataset assigns calendar time readings to the duration of the idle phase.

[0040] In a preferred embodiment, the operational behavior model may include a clustering method that clusters the observed dataset using clustering methods such as k-means clustering, Gaussian mixture models, affinity propagation, etc. The invention is then described based on the k-means clustering method. For this purpose, the observed dataset is clustered to obtain a certain number of cluster locations. For example, the result of this clustering method is that the observed dataset can be clustered, such as in… Figure 3 As shown graphically in the chart.

[0041] Figure 3 A graph is shown illustrating the clustering of observation dataset B and observation dataset C, which represents the duration of idle periods by the day of the week WT (Monday-Sunday) and the time of day TZ (0:00-24:00), as well as the clustering of observation datasets in which idle periods of a certain duration frequently occur.

[0042] For each such determined cluster in the observed dataset, a centroid can be determined, which describes the calendar time reading and the average expected duration of the idle period assigned to that calendar time reading.

[0043] A quantity threshold can be applied to the centroids of cluster C within the clustered observation dataset B to determine if the number of observation datasets assigned to these clusters is sufficient for use in the methods described above. By specifying this quantity threshold, clusters with fewer than the assigned number of observation datasets are removed. Furthermore, only clusters containing a minimum number of observation datasets exceeding the quantity threshold are considered.

[0044] If no cluster has the minimum number of observations, the runtime behavior model is not sufficiently trained to make judgments about the execution of the balanced loop and is therefore unusable. If at least one of the found clusters satisfies the selection criteria, the runtime behavior model is sufficient to make judgments about the execution of the balanced loop. This can be used accordingly in method step S4.

[0045] That is, if it is confirmed in step S3 that there are conditions for performing a balancing cycle, the expected duration of the idle phase of the vehicle battery pack 3 can be determined in step S5 by using the calendar time point when the vehicle is parked or the current calendar time point.

[0046] Therefore, in step S5, it is checked whether the calendar time point can be assigned to one of the detected clusters of the running behavior model. This can be identified, for example, by comparing it with a specified distance threshold. Thus, it can be determined whether the calendar time point can be assigned to a cluster of the running behavior model based on the Euclidean distance between the calendar time reading and the calendar time reading of the centroid of one of the found clusters. In particular, when the Euclidean distance between the time and day of the day of the calendar time reading and the time and day of the current day (the current calendar time reading) is less than the specified distance threshold, it is confirmed that the cluster can be assigned to the current calendar time reading, and the running behavior model can be applied.

[0047] Therefore, the operational behavior model defines the time periods within which the expected duration of idle phases can be determined using the operational behavior model. Outside of these time periods, the operational behavior model cannot provide information about the expected duration of idle phases.

[0048] If it is confirmed in step S5 that the calendar time point can be assigned to one of the clusters of the running behavior model (either option: Yes), the method continues to step S6. Otherwise (either option: No), the method continues to step S11.

[0049] In step S6, the expected duration of the idle phase allocated to the centroids of the assigned clusters is determined.

[0050] In step S7, check whether the expected (still remaining) duration of the idle phase is greater than the expected or known duration (minimum balancing duration) of the balancing cycle to be executed. If so (either / or: yes), then the balancing cycle is started or continued in step S8. Otherwise (either / or: no), the balancing cycle is not started or is blocked (e.g., due to periodic activation caused by the operating mode), and the method continues with step S1.

[0051] In step S9, it is checked whether the balancing cycle has been executed for a sufficient duration, i.e., the minimum balancing duration, or whether there is activation of vehicle 1 or vehicle battery pack 3 (end of idle phase). If so (either option: yes), the balancing cycle ends in step S10 and the method continues with step S1; otherwise (either option: no), the method continues with step S8.

[0052] If it is confirmed in step S4 that the operating behavior model still does not have enough database (either option: no) to be applied, then in step S11, the balancing cycle can be started in the conventional manner after the vehicle battery pack 3 is deactivated, during the idle period after a predetermined time, without knowing the expected duration of the idle phase.

[0053] Alternatively, a balancing cycle can be initiated through an evaluation of the operational behavior model only when the high urgency of performing the balancing cycle is identified due to the discrepancy between the battery charge levels of the battery packs, without knowing the expected duration of the idle phase.

[0054] For example, a high degree of urgency can be identified when the difference between the battery pack battery 31 with the highest battery pack capacity and the battery pack battery 31 with the lowest battery pack capacity is greater than a specified battery pack capacity threshold, such as between 10% and 15% of the maximum battery pack capacity, or when the variance of the distribution of the battery pack capacity of these battery pack batteries 31 is greater than the specified battery pack capacity threshold.

[0055] The determination of the observation dataset and the training or improvement of the running behavior model can be performed continuously in parallel with the methods described above. Specifically, when the start of an idle phase is confirmed, the current calendar time reading can be stored, and the duration of the idle phase can be assigned to that calendar time reading at the end of the idle phase. These observation datasets are stored and clustered according to the clustering methods described above.

[0056] Other types of operational behavior models, such as data-based models like neural networks or Gaussian process models, can be applied similarly. The availability of such an operational behavior model for a given current calendar time reading then depends on the confidence level of the expected duration allocated to the idle phase. If this confidence level is too low, i.e., below a specified confidence interval threshold, the operational behavior model cannot be used for the current idle phase.

[0057] By using outlier checks, for example, to confirm that a large dataset of observations contains data where the distance from the centroid of the found cluster is greater than a specified outlier threshold, it can be determined whether the previously trained behavioral model has been discarded and can be retrained using future observation datasets. Outlier detection on a large dataset of anomalous observations can indicate that the vehicle's behavior has changed significantly, for example, due to a change in driver.

[0058] The above method has been described as being executed within the vehicle's control unit. Since the computational cost for evaluating the data collection points can become very high, it may be stipulated that the model creation and evaluation of this operational behavior model be performed outside the vehicle, such as in a central unit outside the vehicle, particularly in a cloud system.

Claims

1. A method for operating a vehicle battery pack (3) to perform a balancing cycle, wherein in the balancing cycle, at least one battery pack cell of the vehicle battery pack having a higher battery pack capacity than at least one battery pack cell of the remaining battery pack cells is discharged to a specified degree, the method comprising the steps of: - Confirm (S3) the start of the idle phase of the vehicle battery pack (3), during which the vehicle battery pack (3) is in a no-current state; - Based on the current calendar time reading, the expected duration of the idle phase (S6) is determined by means of a data-based operational behavior model; - If the expected duration of the idle phase exceeds the specified minimum balancing duration, then execute (S8) the balancing cycle.

2. The method of claim 1, wherein if the expected duration of the idle phase is less than the minimum balancing duration, the execution of the balancing cycle is prevented.

3. The method according to claim 1 or 2, wherein an observation dataset is acquired during the operation of the vehicle battery pack (3), wherein the observation dataset assigns a calendar time reading of the start of a detected idle phase to the duration of the idle phase, wherein the observation dataset is used to create or train the operation behavior model.

4. The method of claim 3, wherein the operational behavior model includes a clustering method in which the observed dataset is clustered into one or more clusters, wherein each cluster defines a calendar time period and an expected duration of the idle phase; wherein if the current calendar time reading falls within a calendar time period of one of these clusters, the expected duration of the idle phase for the relevant cluster is determined for the current calendar time reading.

5. The method of claim 4, wherein each cluster defines the calendar time period and the expected duration of the idle period by means of the distance from the centroid of the related cluster and the expected duration of the idle period allocated to the centroid of the related cluster.

6. The method of claim 4, wherein if the current calendar time reading is outside the calendar time period determined by these clusters, the balancing cycle begins after a specified relaxation period following the start of the idle phase.

7. The method of claim 6, wherein the balancing cycle is initiated only after a specified relaxation period following the start of the idle phase, when the change in the battery charge of the battery pack (31) of the vehicle battery pack (3) exceeds a specified increase.

8. The method of claim 4, wherein if the current calendar time reading is outside the calendar time period determined by these clusters, the balancing loop is not executed or the execution of the periodically activated balancing loop is prevented.

9. The method according to claim 1 or 2, wherein the balancing cycle is performed only when the change in the battery charge of the battery pack (31) of the vehicle battery pack (3) exceeds a predetermined level.

10. The method of claim 1 or 2, wherein the running behavior model is discarded if the number of observation datasets outside the clusters determined by the clustering method exceeds a specified outlier threshold.

11. An apparatus for performing one of the methods according to any one of claims 1 to 10.

12. A computer program product comprising instructions that, when executed by a computer, cause the computer to perform the steps of the method according to any one of claims 1 to 10.

13. A machine-readable storage medium comprising instructions that, when executed by a computer, cause the computer to perform the steps of the method according to any one of claims 1 to 10.

Images