Method for adapting an operating strategy of an electric vehicle in order to optimise a range

Abstract

The invention relates to a method for adapting an operating strategy of an electric vehicle in order to optimise a range, said method comprising: receiving (208) vehicle data (204) which comprise: information about an initial state of charge (310) of an energy store of the electric vehicle, and / or an initial, estimated remaining range of the electric vehicle; receiving historical usage data (214) for the use of the electric vehicle by the driver; determining (210) a usage prediction (212), which indicates an estimated future usage of the electric vehicle by a (110) driver, based on the historical usage data (214); receiving (216) surroundings data (218) which comprise information about vehicle-external parameters which potentially influence the range; determining (220), based on the surroundings data (218), one or more surroundings predictions (222), each of which indicates an estimated future surroundings behaviour; determining (224), based on the usage prediction (212) and the one or at least one of the plurality of environment predictions (222), and using a prediction model (226), a state prediction (228) which indicates an estimated future state of charge (320) and / or an estimated future remaining range; determining (230), based on the state prediction (228) and with the solving of an optimisation problem, adaptation data (232) which comprise values for parameters of the electric vehicle which influence the range; and transmitting (234) the adaptation data to the electric vehicle.

Description

[0001] Description

[0002] title

[0003] Method for adapting an electric vehicle's operating strategy to

[0004] Optimizing a reach

[0005] The present invention relates to methods for adapting an operating strategy of an electric vehicle to optimize its range, as well as a computing unit, an electromechanical system and a computer program for carrying them out.

[0006] Background of the invention

[0007] Electric vehicles, i.e., vehicles powered by an electric motor, typically have an energy storage device such as a battery. Depending on the battery's state of charge and various other factors, such as the driver's driving style, the electric vehicle has a certain remaining range. This remaining range can be estimated to inform the driver whether they will reach their destination or whether they need to recharge.

[0008] Disclosure of the invention

[0009] According to the invention, methods, in particular computer-implemented methods, for adapting an operating strategy, as well as a computing unit and a computer program for carrying them out, with the features of the independent claims, are proposed. Advantageous embodiments are the subject of the dependent claims and the following description. The invention relates to electric vehicles, i.e., vehicles that are powered by an electric motor and have an energy storage device such as a battery, and in particular to determining or estimating the state of charge of the energy storage device or the remaining range of the electric vehicle, as well as their optimization. Good or even optimal energy management allows predictive conditioning of the operating strategy of an electric vehicle, so that a user can experience optimal range and energy management, and in particular without range anxiety.

[0010] This involves the provision or reception of vehicle data from the electric vehicle. This vehicle data can be recorded or determined, in particular, within the electric vehicle or its processing unit, and then transmitted to and received by a central computing system (e.g., in the cloud). This can be done, for example, via a suitable wireless communication connection.

[0011] The vehicle data includes information about the initial state of charge (SoC) of the electric vehicle's energy storage system, as well as information about the initial, estimated remaining range of the electric vehicle. The initial state of charge is read or determined directly from the electric vehicle. Similarly, the initial, estimated remaining range is determined primarily within the electric vehicle. This remaining range is described as "estimated" because it depends on various factors, such as future driving behavior, and is therefore not known exactly in advance. Such an estimate can be implemented in the electric vehicle, although it is often not very accurate.

[0012] Furthermore, historical usage data regarding the driver's use of the electric vehicle is provided or received. This historical usage data includes and / or is based on, for example, information about one or more operating parameters of the electric vehicle; these can be parameters such as speed, torque, currents, and voltages, which can be measured at various points in the vehicle. In particular, these operating parameters can be presented as time series or trends.

[0013] Furthermore, a usage prediction is determined, indicating an estimated future use of the electric vehicle by a driver; this is based on historical usage data. This historical usage data specifically relates to a particular driver, who can be identified, for example, by an identifier. A driver can also be permanently linked to a specific electric vehicle, meaning the historical usage data is specifically assigned to an electric vehicle that is considered to belong to a particular driver, even if another user drives the electric vehicle at times. Likewise, the historical usage data can also be specifically assigned to the driver themselves, possibly in combination with the electric vehicle.

[0014] Based on historical usage data, specifically driver-specific time series such as vehicle speed, torque, current, voltage, temperature, previously driven routes, and navigation inputs, a usage prediction is created for the specific driver. For this purpose, a so-called hidden Markov model can be used, which, based on historical data, determines the distributions and transition probabilities between...

[0015] Parking, driving, charging, etc., are learned. More details on this specific aspect can be found, for example, in C. Simonis and N. Bajcinca, "On model-based source coding for dynamical systems," 2017 3rd International Conference on Event-Based Control, Communication and Signal Processing (EBCCSP), Funchal, Portugal, 2017, pp. 1-4, doi: 10.1109 / EBCCSP.2017.8022813. This enables probabilistic modeling of driver-specific usage patterns, for example, by describing transition probabilities.

[0016] In addition, environmental data is provided or received. This environmental data includes information about one or more vehicle-external parameters that potentially affect the range. Examples include predictive route data, V2X, traffic data, weather data, and status information (such as Aqua Planning warnings). Various cloud APIs from different providers can be used to obtain such environmental data.

[0017] Scenario-based environmental predictions can be derived in this way. For example, it can be assessed and calculated whether a traffic jam is developing, such as during rush hour, or whether free-flowing traffic is expected; this affects the expected speed trajectory of the electric vehicle depending on time and location. Similarly, it can be used to assess and calculate how the ambient temperature will behave depending on time and location, as well as other information that may be relevant for energy management.

[0018] Based on the environmental data, one or more environmental predictions (i.e., concrete scenarios of environmental behavior or scenario-based environmental predictions) are then determined, each indicating an estimated future environmental behavior. Such scenarios or environmental predictions can be integrated into a simulation environment. This can be done by passing data or metadata to a prediction model.

[0019] Based on the usage prediction and one or at least one of the several environmental predictions, and using a prediction model based on a state prediction, in particular a Gaussian model, a state prediction is then determined. This state prediction provides an estimated future state of charge of the electric vehicle's energy storage system and / or an estimated future remaining range of the electric vehicle.

[0020] Advantageously, a model of the environment is used as the prediction model, providing various pieces of information depending on time and location, taking into account (ideally) all environmental influences, depending on the scenario or environmental prediction. Such information can include, for example, the probability of the environmental prediction occurring (e.g., 80% or 100%), a distance or its temporal (nth) derivative (e.g., velocity, i.e., the first derivative) as a function of time and / or location, a temperature as a function of time and / or location, and optionally other energy management-relevant parameters.

[0021] The primary task of the prediction model is to perform a highly accurate, driver-specific calculation of state of charge (SoC) and / or range. This can be achieved by improving or adapting the SoG and / or range calculations from the electric vehicle's data, or by directly modeling the SoC and / or range.

[0022] The prediction model includes, for example, a probabilistic model, such as the previously mentioned Gaussian regression process. In the first case, the prediction model is trained on the residual of the SoC model from the electric vehicle to improve it (in this case, for example, a hybrid overall model for range prediction is available). In the second case, the prediction model is trained directly on a label that contains SoC or range information, advantageously in aggregated form. A label can correspond to a SoC after battery relaxation, allowing the label to be determined as accurately as possible. More detailed information on this can be found, for example, in US 11,912,159 B2.

[0023] In one embodiment, historical usage data is used as training data to train or parameterize the prediction model. This allows, for example, driver-specific characteristics to be represented. A driver who consistently drives fast, has a strong preference for acceleration, and a high energy demand from auxiliary systems would likely be systematically overestimated by the vehicle's range prediction with regard to the remaining range.

[0024] However, because the range prediction is constantly evaluated, e.g. in the cloud, and the result is preferably also stored in a database, the prediction model can easily learn this driver-specific effect or the systematic range overestimation of the model based on historical data, so that in the future this driver can be predicted much more accurately with regard to the expected range and the future SOC trend.

[0025] The prediction model may have been trained based on historical usage data and will, in particular, be repeatedly retrained; the latter is done with increasingly recent historical usage data that is gradually acquired. "Historical" data, therefore, refers to data that was collected and available prior to a current point in time. The prediction model can be retrained regularly and automatically (e.g., weekly), with a check performed before each training session, based on the latest (current) and historical data, to determine whether the driver is known or if there has been a driver change. This allows the latest data and information to be utilized. The modeling can be performed, for example, using a Gaussian process or a Bayesian neural network. The state prediction for the vehicle range or state of charge (SoC) is provided, for example, including confidence levels.

[0026] Based on state prediction and by solving an optimization problem, adaptation data is then determined. This adaptation data comprises values ​​for one or more parameters of the electric vehicle that influence its range. These parameters can be selected from various parameter sets (or clusters), such as comfort settings of the electric vehicle, operating strategies of components or systems of the electric vehicle, particularly concerning driving dynamics, or navigation.

[0027] Comfort settings include, for example, the amplitude (strength) or target value (e.g., a target temperature) of systems such as air conditioning, steering wheel heating, defrosting functions (e.g., for the rear window), media systems, and ventilation systems. Operating strategies for components or systems include, for example, maximum acceleration, maximum speed, recuperation strength, and chassis settings. For navigation, a location or time for the next charging stop might be considered. The adaptation data is then provided and / or transmitted to the electric vehicle, where it can be received and configured; the latter can be initiated, for example, by a processing unit within the electric vehicle.

[0028] To solve the optimization problem, for example an available black-box algorithm can be used, i.e. an algorithm that solves the optimization problem, but of which it is not necessarily known how the problem is specifically solved.

[0029] In one embodiment, however, it is provided that when determining the adaptation data, pre-selection adaptation data predetermined for the (specific, current) driver are first determined, and that these pre-selection adaptation data are then adapted, in particular iteratively, to obtain the (final) adaptation data at least as required.

[0030] For example, a pre-selection can be made for each driver based on usage predictions, followed by fine-tuning. In an initial step, it is checked whether the destination can be reached with a defined probability, e.g., at least 95%, within the current scenario or environmental prediction. If so, the optimization process terminates and no further measures are required; otherwise, a pre-selection with adjustments follows.

[0031] During the initial selection process, a driver-specific assessment or ranking can be made to determine how important each energy control lever (i.e., which parameters affecting range) is to the individual driver. For example, a sporty driver might not want to limit their maximum acceleration or speed, but would be open to adjusting the recuperation strength so that more energy is fed back into the energy storage system or battery during braking. Energy savings can also be achieved through various comfort settings, such as reducing the air conditioning and ventilation settings.

[0032] During the (potentially iterative) fine-tuning, the energy levers (parameters) can then be quantified predictively, for example, based on a quantile (e.g., the 5th percentile) of the range or state of charge (SoC) prediction, so that the target (a specific location) can be reached with a 95% probability. For this purpose, the parameters or their settings can be iteratively fine-tuned or adapted until a sufficiently high probability of reaching the target location is achieved; for example, the air conditioning temperature can be increased from 19°C to 22°C.

[0033] To solve the optimization problem, a gradient-free approach, such as the aforementioned Bayesian optimization, can be used. Alternatively, a gradient-based approach can be chosen, where, for example, in the cloud, the gradients of the optimization problem are estimated using auto-diff to achieve a global optimum or a sufficiently good local optimum. Heuristic optimization methods, such as grid search, can also be used. In any case, the existing models from the previous procedural steps should be considered when solving the optimization problem.

[0034] The adaptation data obtained in this way, i.e., the measures derived from the optimization, can, as mentioned, be transferred to the electric vehicle and at least partially applied or interpreted there. This can be done directly and automatically.

[0035] In one embodiment, however, a user prompt is displayed (in the electric vehicle) asking whether the adaptation data should be set; this can be done, for example, visually on a display and / or audibly. User input is then received indicating whether the adaptation data should be set; this can be done, for example, via a button press, a touchscreen, or audibly. At least some of the adaptation data is then set, specifically only if the adaptation data is to be set according to the user input, i.e., if the driver (or another user in the vehicle) agrees.

[0036] It may therefore be intended that the explicit consent of the driver or passenger is obtained visually via an HMI or a central display, or alternatively via audio communication, e.g. in the following manner:

[0037] "To reach destination XYZ without a charging stop, the air conditioning will be set from 19°C to 22°C. Is that OK?" The destination can advantageously be linked to or associated with GPS data.

[0038] Once the driver or another user confirms, the measure can be implemented, and the vehicle will then, for example, implement predictive energy management in a closed-loop control system. This occurs continuously, for example, with a check every five to ten minutes, so that the latest information that has occurred in the last few minutes (e.g., accidents, severe weather, road closures, etc.) can also be taken into account.

[0039] In particular, the entire process is logged, i.e., stored in a database. This allows the aforementioned historical usage data, or at least a portion of it, to be preserved. Specifically, the following data can be stored: driver-specific SoC and / or range prediction (embedded, i.e., determined as it is in the electric vehicle) with its prediction horizon; driver-specific SoC and / or range prediction from the cloud (i.e., determined as described above) with its prediction horizon; the actual SoC and / or range history; data relating to the solved optimization problem; and the implementation of the optimization (embedded, i.e., which adaptation data or settings are actually used), including driver feedback on preferences.

[0040] This type of historical information on model performance can be used to continuously improve the method by retraining the prediction model. A computing unit according to the invention, e.g., a computer, server, or central computing system, or a control unit of a motor vehicle, is configured, particularly through programming, to carry out a method according to the invention.

[0041] Implementing a method according to the invention in the form of a computer program or computer program product with program code for carrying out all method steps is also advantageous, as this incurs particularly low costs, especially if an executing control unit is already available for other tasks. Finally, a machine-readable storage medium is provided with a computer program stored on it as described above. Suitable storage media or data carriers for providing the computer program are, in particular, magnetic, optical, and electrical storage media, such as hard drives, flash memory, EEPROMs, DVDs, etc. Downloading a program via computer networks (Internet, intranet, etc.) is also possible. Such a download can be wired or wireless (e.g., via a WLAN network, a 3G, 4G, 5G, or 6G connection, etc.).

[0042] Further advantages and embodiments of the invention will become apparent from the description and the accompanying drawing.

[0043] The invention is schematically illustrated in the drawing using an exemplary embodiment and is described below with reference to the drawing.

[0044] Brief description of the drawings

[0045] Figure 1 schematically shows an electric vehicle in which a method according to the invention can be used.

[0046] Figure 2 schematically shows a process according to the invention in a preferred embodiment. Figure 3 schematically shows a diagram to illustrate the invention.

[0047] embodiment(s) of the invention

[0048] Figure 1 schematically and by way of example shows an electric vehicle 100 in which a method according to the invention can be used. The electric vehicle has, by way of example, an electric motor 102 for propulsion and an energy storage device 104 designed as a battery for supplying the electric motor with electrical energy. In addition, the vehicle 100 has, by way of example, a computing unit 106 designed as a control unit.

[0049] Furthermore, a driver 110 of the electric vehicle 100 is shown as an example. Additionally, a road 120 is shown, purely as an example, along which the driver 110 can drive the electric vehicle 100.

[0050] Furthermore, a computing unit 150, designed as a central or higher-level computing system, is shown, which is intended to serve as an example of the so-called cloud, in which data can be processed. Data can be exchanged between the vehicle 100 and the computing system 150 via a wireless communication link that is only indicated.

[0051] Figure 2 schematically illustrates a process according to the invention in a preferred embodiment, namely for adapting an operating strategy of an electric vehicle to optimize the range of the electric vehicle, such as the electric vehicle 100 according to Figure 1.

[0052] After the procedure is started, vehicle data can be determined in step 202, either in the electric vehicle or in an executing computing unit there. This vehicle data includes information about the initial state of charge of the electric vehicle's energy storage system, as well as information about the initial, estimated remaining range of the electric vehicle. Furthermore, the vehicle data includes information about one or more operating parameters of the electric vehicle, such as speed, torque, currents, and voltages.

[0053] This vehicle data 204 is then transmitted to the higher-level computer system in step 206 and received and made available there in step 208. The higher-level computer system is also indicated here by 150 to illustrate which steps can be carried out there in particular.

[0054] Furthermore, historical usage data 214 is received or provided. In a step 210, a usage prediction 212 is then determined, which indicates an estimated future use of the electric vehicle by a driver. This is done based on historical usage data 214 regarding the driver's use of the electric vehicle. This historical usage data 214 can, for example, be stored on a storage unit 152 of the computing system 150. As already mentioned, this data is regularly updated. Likewise, the historical usage data 214 can, for example, be transmitted from the vehicle to the higher-level computing system together with the vehicle data 204, where it is received and provided. As mentioned, the historical usage data 214 can also include and / or be based on information about one or more operating parameters of the electric vehicle, such as speed, torque, currents, and voltages.

[0055] In step 216, environmental data 218 is provided or received. This environmental data 218 comprises information about one or more vehicle-external parameters that potentially influence the range; examples include information 218.1 on traffic congestion, 218.2 on weather, and 218.3 on road conditions. This environmental data 218 can be obtained or queried from one or more external sources. The environmental data 218 is primarily current or live data. In step 220, based on the environmental data 218, one or more environmental predictions 222 are determined or created, each indicating an estimated future environmental behavior (so-called scenarios).

[0056] In step 224, based on the usage prediction 212 and one or at least one of the several environment predictions 222, and using a prediction model 226, a state prediction 228 is determined which indicates an estimated future state of charge of the electric vehicle's energy storage and / or an estimated future remaining range of the electric vehicle.

[0057] The prediction model is based on a machine learning model, specifically a Gaussian model. As mentioned, it is preferably retrained regularly, or alternatively, according to one or more criteria, repeatedly to incorporate more recent historical usage data.

[0058] In step 230, based on the state prediction 228 and by solving an optimization problem, adaptation data 232 are determined. The adaptation data 232 comprise values ​​for one or more parameters of the electric vehicle that influence its range. As mentioned, these parameters can be selected, for example, from the parameter areas of the electric vehicle's comfort settings, operating strategies of components or systems of the electric vehicle, particularly concerning driving dynamics, and navigation.

[0059] As already explained in detail, predetermined preselection adjustment data can first be determined, which can then be iteratively adjusted (fine-tuning) if necessary, in order to obtain the (final) adjustment data 232.

[0060] This is illustrated in a diagram in Figure 3. Here, a state of charge (300) is plotted against a time (302). Figure 310 represents the initial state of charge of the energy storage system, as determined within the electric vehicle itself. During fine-tuning, the energy lever (parameter) can then be quantified predictively, for example, based on a quantile (e.g., the 5th percentile) of the range or state of charge prediction, so that the target (314) (a specific location) can be reached with a 95% probability. For this purpose, the parameters and their settings can be iteratively fine-tuned or adapted until a sufficiently high probability of reaching the target location is achieved; for example, the air conditioning temperature can be increased from 19°C to 22°C.

[0061] Figure 3 shows a curve of the estimated future state of charge of the energy storage system (320), and figures 312.1 and 312.2 show an upper and lower limit of the mentioned 95% interval within which the estimated future state of charge will lie.

[0062] The adaptation data 232 determined above are then, in step 234, provided and transmitted to the electric vehicle. There, they are received in step 236. In step 238, the electric vehicle is then instructed to configure at least some of the adaptation data. As mentioned, this may occur after a user query 240 and subsequent user input 242.

[0063] Furthermore, the data obtained in this way are stored in step 244, e.g. on the storage unit 152 of the computer system 150, thus ending the procedure, step 244.

Claims

Claims 1. Method for adapting an operating strategy of an electric vehicle (100) to optimize the range of the electric vehicle, comprising: providing or receiving (208) vehicle data (204) of the electric vehicle, wherein the vehicle data comprise: Information about an initial state of charge (310) of an electric vehicle's energy storage system, and / or Information about an initial, estimated remaining range of the electric vehicle, which has been determined specifically in the electric vehicle; Providing or receiving historical usage data (214) for the use of the electric vehicle by the driver; Determine (210) a usage prediction (212) that indicates an estimated future use of the electric vehicle by a (110) driver, based on historical usage data (214); Providing or receiving (216) environmental data (218), wherein the environmental data includes information about one or more vehicle-external parameters that potentially affect the range; Determine (220), based on the environmental data (218), one or more environmental predictions (222), each indicating an estimated future environmental behavior; Determine (224), based on the usage prediction (212) and one or at least one of the several environment predictions (222), and using a prediction model (226) based on a machine learning model, in particular a Gaussian model, a state prediction (228) that specifies an estimated future state of charge (320) of the energy storage of the electric vehicle and / or an estimated future remaining range of the electric vehicle; Determine (230), based on the state prediction (228), and by solving an optimization problem, of fitting data (232), which Values ​​for one or more parameters of the electric vehicle that affect the range include; and Providing or transmitting (234) the adaptation data to the electric vehicle.

2. Method according to claim 1, wherein the prediction model has been trained based on historical usage data and is repeatedly retrained.

3. Method according to claim 1 or 2, wherein the adaptation data (232) comprise several parameters of the electric vehicle that affect the range, which are selected from at least two of the following parameter ranges: Comfort settings of the electric vehicle, operating strategies of components or systems of the electric vehicle, in particular regarding driving dynamics, navigation.

4. Method according to any of the preceding claims, wherein determining (230) the adjustment data (232) comprises: Determining pre-selection adaptation data predetermined for the driver, and if necessary, adapting, in particular iteratively, the pre-selection adaptation data to obtain the adaptation data.

5. The method of claim 4, wherein the adjustment of the preselection adjustment data comprises: adjusting one or at least one of the several parameters, based on a predetermined quantile of the state prediction, such that a predetermined goal of the electric vehicle is achieved with at least a predetermined probability.

6. Method according to one of the preceding claims, wherein the initial state of charge and / or the initial estimated range has been determined based on a model, wherein the prediction model has been trained on a residual of the model, and wherein determining the state prediction, using the prediction model, includes a correction of the initial state of charge and / or the initial estimated remaining range to obtain the estimated future state of charge and / or the estimated future remaining range.

7. Method according to any one of claims 1 to 5, wherein the prediction model has been directly trained, and wherein determining the state prediction, using the prediction model, comprises directly determining the estimated future state of charge and / or the estimated future range.

8. Method according to one of the preceding claims, wherein the determination of the usage prediction (212) based on the historical usage data is carried out using a hidden Markov model.

9. Method for adapting an operating strategy of an electric vehicle (100) to optimize the range of the electric vehicle, comprising: Providing or transmitting (206) vehicle data (204) of the electric vehicle, wherein the vehicle data includes: Information about the initial charge level of an electric vehicle's energy storage system, and / or Information about an initial, estimated remaining range of the electric vehicle, which has been determined specifically in the electric vehicle; Providing or transmitting historical usage data (214) relating to the driver’s use of the electric vehicle; Received (236) adaptation data (232) which include values ​​for one or more parameters of the electric vehicle that affect the range; and Causing (238) the electric vehicle to set at least some of the adaptation data.

10. The method of claim 9, further comprising: Display a user prompt (240) asking whether the customization data should be set; and Receiving a user input (242) as to whether the customization data should be set; wherein at least a part of the customization data is set, in particular only if the customization data is to be set.

11. Method according to claim 9 or 10, wherein the adaptation data have been determined according to any one of claims 1 to 8.

12. Computing unit (106, 150) configured to perform all process steps of a process according to any of the preceding claims.

13. Computer program that causes a computing unit (150) to perform all the process steps of a method according to any one of claims 1 to 11 when executed on the computing unit (150).

14. Machine-readable storage medium with a computer program stored thereon according to claim 13.

Images