Method for predicting the aging state of a battery

Abstract

The present invention relates to a method for predicting the aging state of a battery. In addition, the present invention relates to a vehicle including at least one battery, predicting the aging state of the battery according to the method of the present invention and / or improving the aging characteristics of the battery by means of the aging state predicted according to the method of the present invention. The present invention also relates to a prediction system configured to implement the method of the present invention.

Description

Technical Field

[0001] The present invention relates to a method for predicting the aging state of a battery.

[0002] Furthermore, the present invention relates to a vehicle including at least one battery, predicting the aging state of the battery according to the method of the present invention and / or improving the aging characteristics of the battery by means of the aging state predicted according to the method of the present invention.

[0003] The present invention also relates to a prediction system configured to implement the method of the present invention. Background Art

[0004] The aging state (SOH, State of Health in English) of a battery depends on different influencing factors. These influencing factors are, for example, the amount of current passing through the battery, the number and depth of the charging and discharging cycles of the battery, the maximum charging and discharging current of the battery, the thermal circuit for the battery, the operating temperature of the battery, the state of charge (SOC, State of Charge in English) of the battery, and so on. Since all these factors determine the SOH value in a battery that usually operates individually in different forms and weights, it is difficult to accurately determine this value.

[0005] Document WO 2019 / 017991 A1 describes a battery management system for a motor vehicle, which includes a model for estimating the state of a rechargeable battery. Summary of the Invention

[0006] A method for predicting the aging state of a battery is proposed.

[0007] The battery is preferably configured as a lithium-ion battery and includes a plurality of battery cells connected in series or in parallel with each other.

[0008] Herein, first, data related to different usage attributes of the battery are provided. The data of the battery includes, for example, the voltage change curve of the battery, the current change curve of the battery, and the operating temperature change curve of the battery. Of course, the data can also include parameters characterizing the battery. The parameters characterizing the battery include, for example, temperature, charging current, energy throughput, state of charge, or a combination of states, such as a high state of charge at high temperature for a long time. The parameters characterizing the battery of the battery also include chemical data of the battery. "Usage attributes" especially refer to the load and charging attributes of the battery.

[0009] Subsequently, the data of the battery is transmitted to a storage institution. Here, the data of multiple comparison batteries are stored in the storage institution. In the storage institution, the aging state values and / or aging state change curves of the comparison batteries, preferably related to the usage attributes of the corresponding comparison batteries, are also stored. The storage institution can be assigned to a data-driven fleet model (Flottenmodell). Preferably, the storage institution includes the data-driven fleet model. Here, the data stored in the storage institution is processed and analyzed with the aid of the data-driven fleet model.

[0010] Thereafter, the aging state value of the battery is determined according to defined times and / or defined events with the aid of a battery-reference model. The battery-reference model includes the relationship between the data of the battery, the aging state value of the battery, and the usage attributes of the battery here. Preferably, the signal or signal change curve of the battery is preprocessed with the aid of a preprocessing model before determining the aging state value.

[0011] If the parameters characterizing the battery of the battery, such as tags, the measured aging state reference for machine learning, are known, the aging state value can be directly and precisely calculated with the aid of the battery-reference model. If there are tags, the battery-reference model is made into an artificial intelligence-based model. If there are no tags yet, the battery-reference model corresponds to, for example, the value calculated in a vehicle control device.

[0012] If the parameters characterizing the battery of the battery are not known, the aging state value of the battery can be calculated, for example, with the aid of a battery management system. The aging state value calculated by the battery management system then serves as a base value for the battery-reference model and is associated with the parameters characterizing the battery of the battery, which are indicators for the degree of aging of the battery.

[0013] Next, the aging state change curve of the battery is formed according to the determined aging state value of the battery.

[0014] Thereafter, the predicted aging state value and / or the predicted aging state change curve of the battery are determined. Here, the predicted aging state value and / or the predicted aging state change curve of the battery can be determined by extrapolation of the formed aging state change curve.

[0015] It is also possible to determine the predicted aging state value and / or the predicted aging state change curve from the data stored in the storage mechanism of the comparison battery by means of the allocation relationship between the formed aging state change curve and the identified aging state change curve, wherein a plurality of aging state change curves are identified from the data stored in the storage mechanism of the comparison battery. The identification work can be carried out by means of a clustering scheme of the battery load (English: Clustering), such as the nearest neighbor heuristic or the k-means algorithm, taking into account the aging state change curves of different usage attributes.

[0016] In order to determine the predicted aging state value and / or the predicted aging state change curve, the extrapolation and allocation relationship of the formed aging state change curve can be used together.

[0017] Preferably, the load prediction of the battery is implemented. The load prediction can be carried out, for example, on the basis of the predictive mileage data of the vehicle, the navigation data of the vehicle and other information from the vehicle control device. Here, the predictive mileage data can be determined by means of vehicle-to-X communication, especially vehicle-to-vehicle communication. Vehicle-to-X / vehicle communication represents the exchange of information and data between the vehicle and the environment / vehicles, which has the background of reporting critical and dangerous situations to the driver in advance.

[0018] Preferably, the predicted aging state value and / or the predicted aging state change curve of the battery are calibrated from the actually determined aging state value by the battery reference model using the actual aging state value and / or the actual aging state change curve of the battery. It is also possible to calibrate the predicted aging state value and / or the predicted aging state change curve of the battery using the stored aging state value and / or the stored aging state change curve from the storage mechanism of the comparison battery, wherein the stored aging state value and / or the stored aging state change curve extrapolate the individual aging state change curves of the corresponding comparison battery on the one hand and predict the confidence interval with respect to the aging state value by means of the statistical distribution of the comparison battery on the other hand. Preferably, the two calibrations are implemented.

[0019] Preferably, the battery reference model is constructed as an artificial intelligence-based model. However, it is also conceivable that the battery reference model is constructed as a physical model, a meta-model or a data model, which can be the comprehensive characteristic curve of the battery. Of course, multiple of the above-mentioned models can be used to form the battery reference model.

[0020] Preferably, the battery-reference model is calibrated based on the data stored in the storage mechanism for the comparison battery. Without knowing the parameters characterizing the battery, the battery-reference model can be parameterized based on the aging state value of the battery calculated by the battery management system and used as a base value for the battery-reference model, and the data stored in the storage mechanism for the comparison battery.

[0021] The aging state value of the battery is preferably determined for different prediction ranges (“Prädiktionshorizonte”).

[0022] A “prediction range” can for example refer to a specific duration associated with a specific minimum number of driving cycles of the vehicle. Thereby, the calendar and cyclic aging of the battery with current and temperature loads is also taken into account. Compared with the predicted aging state values in the medium term, such as 4 weeks and / or 20 driving cycles and the predicted aging state values in the long term, such as 8 weeks and / or 40 driving cycles, in particular the predicted aging state values in the short term, such as 2 weeks and / or 10 driving cycles, take into account very different driving patterns that may occur, for example, when changing the driver or owner of the vehicle. Thereby, the prediction accuracy is much higher.

[0023] Preferably, the data of the battery is transmitted to the storage mechanism via a wireless network system. Here, the wireless network system can be configured as a WLAN network system. Preferably, the wireless network system is configured as a wireless mobile communication network, such as a UMTS- or LTE-network.

[0024] Preferably, the storage mechanism is configured as a cloud memory. However, it is also conceivable that the storage mechanism is configured as a storage medium, such as the memory of the battery control device or an external memory.

[0025] Preferably, the data of the battery transmitted to the storage mechanism is continuously monitored and evaluated.

[0026] Preferably, the data stored in the storage mechanism for the battery or the corresponding comparison battery is continuously verified and evaluated.

[0027] Preferably, a model based on artificial intelligence is used to monitor, evaluate and verify the data transmitted to the storage mechanism and the data stored in the storage mechanism.

[0028] When a deviation occurs between the predicted and the actual aging state values at the prediction time, it is always necessary to match the actual values determined by the battery-reference model and take them into account for future predictions. Possible systematic deviations are included in another predicted aging state change curve and corrected so that the prediction residuals are normally distributed.

[0029] Preferably, an operating strategy for the battery is applied based on the predicted aging state values or the predicted aging state change curve, and the goal pursued by the operating strategy is to suppress the aging characteristics of the battery and thereby extend the service life of the battery. This can be achieved by, for example, placing the battery in an economic operating state by varying power limits, charging characteristics, operating temperature, or similar parameters. Thereby, the battery can leave the original poor aging state change curve and transition to a better aging state change curve.

[0030] If the medium-term or long-term aging state values of the battery show a significant deterioration, a slow reduction in the aging state can be achieved by changing the operating strategy. The measures that take effect, both passive and active measures, for example, involve current reduction, temperature regulation, or suggestions for avoiding rapid charging cycles, etc. The strategy adjustment can also be used as a pre-specification for the optimal aging characteristics of other batteries connected to the storage device with similar aging state change curves. Due to load changes caused, for example, by more driving in urban or long-distance traffic or due to changes in charging characteristics, the aging state values can also be reduced more moderately and transition to other aging state change curves without operating strategy intervention. Through these measures, the extension of the service life of the battery is ensured.

[0031] In the opposite case where the battery is not operated under optimal conditions (which would lead to increased aging), a change in the operating strategy may also result in, for example, higher power or more deep discharge cycles providing individual fast charging processes for the battery or similar advantages in terms of the battery's aging characteristics.

[0032] Furthermore, a vehicle including at least one battery is proposed, and the aging state of the battery is predicted according to the method of the present invention and / or the aging characteristics of the battery are improved by means of the aging state predicted according to the method of the present invention.

[0033] Other information or parameters of the vehicle, such as the usage attributes of the vehicle and the driving style of the driver, can also be transmitted to the storage device here.

[0034] A prediction system is also proposed, which is set up to implement the method according to the invention for predicting the aging state of a battery.

[0035] The prediction system can, for example, have a battery-reference model, a storage mechanism with a data-driven model, and a fusion model. The battery-reference model represents the basic battery technology here. The data-driven model heuristically depicts the actual characteristics of a comparison battery with respect to battery aging. The fusion model combines these two approaches into a highly accurate calculation and prediction of the aging state of the battery. Preferably, the fusion model is constructed as an artificial intelligence-based model. The prediction system can also have a preprocessing model for signal preprocessing and a load prediction model for predicting the load on the battery.

[0036] Advantages of the Invention

[0037] The advantages according to the invention allow aging determination to be taken into account without individual aging state correlations.

[0038] With the method according to the invention, no adjustment of the software in the battery management system is required for aging state determination and prediction.

[0039] Furthermore, the battery-reference model can be calibrated with the data from the storage mechanism and thereby the accuracy of the predicted aging state values can be improved.

[0040] Advantageously, the method according to the invention allows, through short-term prediction, a rapid response to changing load characteristics of the battery, for example, with respect to systematic changes or anomalies in the usage characteristics.

[0041] With the method according to the invention, an operating strategy of the battery can be applied, which causes an extension of the battery service life and / or an improvement in efficiency.

[0042] Advantageously, with the method according to the invention, changes in the load and charging characteristics of the battery can be automatically recognized and thus an operating strategy intervention may not be required.

[0043] With the method according to the invention, non-obvious contributing factors for battery aging can be analyzed, modeled, and verified across vehicles, for example, by means of big data methods and artificial intelligence in the storage mechanism, and can then be directly used for predicting the aging state of the battery. Thereby, the prediction of the aging state is also enriched by additional information from the storage mechanism, which further improves the accuracy of the aging prediction.

[0044] For an unknown battery, there are no parameters for characterizing the battery. For the unknown battery, big data volumes can be analyzed, modeled, and verified across vehicles by artificial intelligence in the storage mechanism. Through the newly identified interaction relationships, new product generations and / or software upgrades can be continuously optimized with the knowledge from the storage mechanism.

[0045] Furthermore, with the method according to the invention, an online-verified battery-reference model for an unknown battery can be derived from the data stored in the storage mechanism. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Embodiments of the present invention are explained in detail with the aid of the drawings and the following description. Among them:

[0047] Figure 1 shows a flowchart of a method according to the invention for predicting the aging state of a battery; and

[0048] Figure 2 shows a schematic diagram of a prediction system for implementing the method according to the invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0049] In the following description of the embodiments of the present invention, the same or similar elements are denoted by the same reference numerals, and in some cases, the repeated description of these elements is omitted. The drawings only schematically show the subject matter of the present invention.

[0050] Figure 1 shows a flowchart 100 of a method according to the invention for predicting the aging state of a battery.

[0051] In a first step 101, data of the battery related to different usage attributes of the battery are provided. The data of the battery includes, for example, the voltage change curve of the battery, the current change curve of the battery, and the operating temperature change curve of the battery. Of course, the data can also include parameters characterizing the battery. "Usage attributes" especially refer to the load and charging attributes of the battery.

[0052] In a second step 102, the data of the battery is transmitted to a storage mechanism 240 (see Figure 2 ). Here, data of a plurality of comparison batteries are stored in the storage mechanism 240. In the storage mechanism 240, aging state values and / or aging state change curves of the comparison batteries, preferably related to the usage attributes of the corresponding comparison batteries, are also stored. The storage mechanism 240 can be assigned to a data-driven population model 242 (see Figure 2). Preferably, the storage mechanism 240 includes a data-driven population model 242. The data stored in the storage mechanism 240 is processed and analyzed by means of the data-driven population model 242.

[0053] In the third step 103, the aging state value of the battery is determined by means of the battery-reference model 230 (see Figure 2 ), at defined times and / or according to defined events. The battery-reference model 230 includes the relationship between the data of the battery, the aging state value of the battery, and the usage attributes of the battery. Preferably, before determining the aging state value, the signal or the signal change curve of the battery is preprocessed by means of a preprocessing model 220 (see Figure 2 ).

[0054] If the battery parameters characterizing the battery are known, the aging state value can be calculated directly and precisely by means of the battery-reference model 230.

[0055] If the battery parameters characterizing the battery are not known, then, for example, the aging state value of the battery can be calculated by means of a battery management system. The aging state value calculated by the battery management system is then used as a base value for the battery-reference model 230 and associated with the battery parameters characterizing the battery, which are indicators of the degree of aging of the battery.

[0056] In the fourth step 104, an aging state change curve of the battery is formed based on the determined aging state value of the battery.

[0057] In the fifth step 105, the predicted aging state value and / or the predicted aging state change curve of the battery are determined. The predicted aging state value and / or the predicted aging state change curve of the battery can be determined by extrapolation of the formed aging state change curve.

[0058] The predicted aging state value and / or the predicted aging state change curve can also be determined from the data stored in the storage mechanism 240 of the comparison battery by means of the assignment relationship between the formed aging state change curve and the identified aging state change curve, where multiple aging state change curves are identified from the data stored in the storage mechanism 240 of the comparison battery. The identification work can be carried out by means of a clustering scheme (English: Clustering) for identifying the load of the battery, such as the nearest neighbor heuristic or the k-means algorithm, taking into account the aging state change curves of different usage attributes.

[0059] To determine the predicted aging state value and / or the predicted aging state change curve, extrapolation and assignment of the formed aging state change curve can be used together.

[0060] Figure 2 A schematic illustration of a prediction system 200 is shown, which is set up to carry out the method according to the invention.

[0061] The prediction system 200 hereby includes a preprocessing model 220, a battery-reference model 230, a storage means 240 with a data-driven population model 242, a load prediction model 250, and a fusion model 260.

[0062] Here, first, data of a battery (not shown) of the vehicle 210, which are related to different usage attributes of the battery, and all signals of the vehicle 210 are provided. The data of the battery include, for example, the voltage curve of the battery, the current curve of the battery, and the operating temperature of the battery. Of course, the data can also include parameters characterizing the battery. "Usage attributes" particularly refer to the load and charging attributes of the battery.

[0063] The signals of the battery and the vehicle 210 are preprocessed by the preprocessing model 220 for further use. The preprocessed signals are then transmitted to the battery-reference model 230, the storage means 240, and the load prediction model 250.

[0064] In the storage means 240, data of a plurality of comparison batteries are stored. In the storage means 240, the aging state values and / or aging state change curves of the comparison batteries, preferably related to the usage attributes of the respective comparison batteries, are also stored. Signals or information of other vehicles are also stored in the storage means 240. The data stored in the storage means 240 are hereby processed and analyzed by means of the data-driven population model 242.

[0065] With the aid of the load prediction model 250, for example, the load of the battery can be predicted. The load prediction can be carried out, for example, on the basis of predictive mileage data of the vehicle 210, navigation data of the vehicle 210, and additional information from the control device of the vehicle 210. Here, the predictive mileage data can be determined by means of communication, such as vehicle-to-X / vehicle communication, and transmitted between the load prediction model 250 and the storage means 240.

[0066] Based on the data of the battery, the aging state value of the battery is determined according to the defined moments and / or defined events by means of the battery-reference model 230. The battery-reference model 230 includes the relationship between the data of the battery, the aging state value of the battery, and the usage attributes of the battery here. When determining the aging state value of the battery, the result of the load prediction model 250 is also taken into account.

[0067] The aging state value determined by the battery-reference model 230, the result of the load prediction model 250, and the data stored in the storage mechanism 240 are transmitted to the fusion model 260. The fusion model 260 is constructed as an artificial intelligence-based model here. The aging state of the battery is predicted by means of the fusion model 260.

[0068] The battery-reference model 230 is calibrated by means of the fusion model 260 based on the data of the comparison battery stored in the storage mechanism 240. The fusion model 260 thus integrates the model-based aging state calculation and the prediction of the battery-reference model 230, as well as the data-driven aging state calculation and the prediction of the data-driven population model 242.

[0069] Based on the predicted aging state of the battery, an operating strategy for the battery or the vehicle 210 is developed by means of the fusion model 260. The goal pursued by the operating strategy is to suppress the aging characteristics of the battery and thereby extend the service life of the battery. This can be achieved by, for example, placing the battery in an economic operating state by varying power limits, charging characteristics, operating temperature, or similar parameters. Thereby, by means of the operating strategy, the battery can deviate from the original poor aging state change curve and transition to a better aging state change curve.

[0070] The present invention is not limited to the embodiments described here and the aspects emphasized therein. Rather, within the scope defined by the claims, a large number of alternative solutions within the scope of those skilled in the art are feasible.

Claims

1. A method for predicting the aging state of a battery, the method comprising the following steps: - providing data related to various usage attributes of the battery; - transmitting the data to a storage institution (240), where data of multiple comparison batteries are stored, processed, and analyzed; - determining the aging state value of the battery according to defined times and / or events by means of a battery-reference model (230), wherein the battery-reference model (230) includes the relationship between the data of the battery, the aging state value of the battery, and the usage attributes of the battery; - forming an aging state change curve of the battery according to the determined aging state value; - determining the predicted aging state value and / or the predicted aging state change curve of the battery by means of extrapolation of the formed aging state change curve and by means of the assignment relationship between the formed aging state change curve and the aging state change curves found from the data stored in the storage institution (240) of the comparison batteries, wherein multiple aging state change curves are found from the data stored in the storage institution (240).

2. The method according to claim 1, characterized in that the battery-reference model (230) is constructed as an artificial intelligence-based model.

3. The method according to claim 1 or 2, characterized in that the battery-reference model (230) is calibrated according to the data of the comparison batteries stored in the storage institution (240).

4. The method according to claim 1 or 2, characterized in that the aging state value of the battery is determined for different prediction ranges.

5. The method according to claim 1 or 2, characterized in that the data of the battery is transmitted to the storage institution (240) by means of a wireless network system.

6. The method according to claim 1 or 2, characterized in that the storage institution (240) is constructed as a cloud memory.

7. The method according to claim 1 or 2, characterized in that the data of the battery transmitted to the storage institution (240) is continuously monitored and evaluated.

8. The method according to claim 1 or 2, characterized in that the data of the battery and the comparison batteries stored in the storage institution (240) are continuously verified and evaluated.

9. A vehicle (210) including at least one battery, the aging state of the battery is predicted according to the method according to any one of claims 1 to 8, and / or the aging characteristics of the battery are improved by means of the aging state predicted according to the method according to any one of claims 1 to 8.

10. A prediction system (200), the prediction system being set up to implement the method according to any one of claims 1 to 8.

Images