Method and device for assigning charging points to vehicles to optimize vehicle utilization and waiting times at the charging points

Abstract

The invention relates to a method, in particular a method that is at least partially computer-implemented, for operating a charging point infrastructure (1) with multiple charging points (2) for charging electric vehicles (4), comprising the following steps: - Assigning (S1) one or more charging points (2) to the electric vehicles (4) depending on the current state of charge of the electric vehicles (4); - Performing (S4) an optimization depending on a cost function to assign exactly one charging point (2) to each electric vehicle (4), where the cost function maximizes the amount of energy transferred by all charging points (2) for a predetermined period; - Transmitting (S5) the exactly one assigned charging point (2) to each of the electric vehicles (4).

Description

Technical field

[0001] The invention relates to the allocation of charging points within a charging point infrastructure to electric vehicles, particularly when a large number of electric vehicles are competing for charging points. The invention further relates to a method for achieving optimal utilization of the charging point infrastructure. Technical background

[0002] Electric vehicles need to be recharged at regular intervals. This is often done at publicly accessible charging points that the vehicles can drive to. At the charging points, the electric vehicle is connected to the charging point to recharge. Charging points can be charging stations or other charging devices.

[0003] Often, several vehicles compete for a single charging point, even though other charging points are available and accessible nearby. This frequently leads to waiting times at one charging point, as electric vehicles can only be charged one after the other. Meanwhile, other charging points have low occupancy.

[0004] Furthermore, it is possible to determine individual driver usage patterns and use this information to reduce future battery strain. This allows for a precise estimate of both the range and the charging time.

[0005] Document WO 2013 045 449 A1 discloses a method for charging electric vehicles by charging stations, comprising the steps a) assigning electric vehicles to different electric vehicle supply units of the charging stations, and b) charging the electric vehicles according to electric vehicle charging information and provided charging power by the electric vehicle supply units of the charging stations, wherein a comparison of electric vehicle charging information, preferably load profiles, of electric vehicles and charging power information, preferably provided charging power, of different charging stations is predicted on the basis of electric vehicle information and a charging station parameter, and wherein steps a) and b) are carried out on the basis of the predicted comparison. Disclosure of the invention

[0006] According to the invention, a method for operating a charging point infrastructure according to claim 1, as well as a corresponding device and a charging point infrastructure according to the dependent claims, are provided.

[0007] Further details are specified in the dependent claims.

[0008] According to a first aspect, a procedure for operating a charging point infrastructure with multiple charging points is planned, comprising the following steps: - Assigning one or more charging points to electric vehicles depending on the current state of charge of the electric vehicles; - Performing an optimization based on a cost function to assign exactly one charging point to each electric vehicle, where the cost function maximizes the amount of energy transferred by all charging points for a predetermined period; - Transmitting the one and only assigned charging point to each of the electric vehicles.

[0009] Furthermore, the assignment of one or more charging points to electric vehicles can be carried out in such a way that, in particular depending on a usage pattern that indicates historical usage behavior, the respective charging point is reached with a predetermined minimum probability with a state of charge greater than a predetermined minimum state of charge.

[0010] Furthermore, the cost function can be designed to minimize the waiting time of electric vehicles at charging points and / or to minimize the overall aging of all vehicle batteries of the electric vehicles through optimization.

[0011] In particular, the waiting time can be determined from the expected arrival time at the charging point as the time that must be waited until one or more other vehicles have completed their charging processes at that charging point, the arrival time depending on the route and usage patterns.

[0012] Furthermore, after transmitting the one and only assigned charging point to each of the electric vehicles, the electric vehicle can be controlled to drive to the assigned charging point or to guide the driver to the charging point.

[0013] It can be provided that for each electric vehicle, a usage pattern indicating historical usage behavior for that vehicle is provided in the form of a usage pattern model in the form of a Hidden Markov Model (HMM). A predicted usage pattern for a prediction horizon is created using the usage pattern model. Depending on the predicted usage pattern and the distance between the current location of the electric vehicle and the location of a relevant charging point, an arrival time is determined. The optimization procedure then assigns each electric vehicle to the charging point that results in the shortest waiting times, based on the arrival times. The use of such a Hidden Markov Model is described, for example, in "C. Simonis, 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, doi: 10.1109 / EBCCSP.2017.8022813. Using the HMM, probabilistic modeling of driver-specific usage scenarios or patterns is possible by describing transition probabilities, also with regard to including charging behavior or interaction with the charging infrastructure.

[0014] For the operation of electric vehicles, each vehicle can be assigned a charging point within a charging infrastructure, maximizing the utilization of all charging points while simultaneously minimizing waiting times for electric vehicle drivers at a single charging point. Beyond optimizing the time-dependent energy output of the charging infrastructure's charging points, it is also beneficial to distribute the electric vehicles among the charging points in such a way that the risk of an electric vehicle failing to reach its charging point due to a rapidly depleting battery remains below a predefined threshold, thus minimizing battery aging for all vehicle batteries operated within a given period.

[0015] This task is particularly complex because, depending on pre-existing conditions and usage patterns, vehicle batteries are aged and stressed to varying degrees, and during a charging process, they experience different aging or degradation depending on the initial state of charge and the initial charging temperature.

[0016] The above procedure involves assigning charging points within a charging infrastructure to electric vehicles, thus allocating a charging point to each vehicle. The charging points must be within range of the respective vehicle, and a time slot can be assigned to each vehicle. This time slot is chosen so that the charging point is either unoccupied during this period or the waiting time before charging begins is as short as possible, and a specific amount of energy is available to charge the vehicle.

[0017] To increase the overall energy throughput of the charging points and, in an advantageous design, to further minimize vehicle waiting times at charging points, the above procedure is provided. This procedure makes it possible to assign a charging point to each electric vehicle within whose range charging points of a charging infrastructure are located.

[0018] First, vehicles located within the coverage area of ​​a charging infrastructure are assigned the charging points within their range. This assignment can be done by assigning a charging point to a vehicle if the vehicle can reach the charging point with a certain predefined minimum probability (e.g., more than 99%), given a predefined minimum state of charge (SOC) of, for example, 10%. This is based on a calculated remaining range, which can be predicted based on the current state of charge and the vehicle's usage pattern. The prediction of the range or SOC can be performed probabilistically as a function of distance and time to quantify the uncertainty accordingly.

[0019] Furthermore, only those charging points that are not yet booked or occupied at the time of the expected arrival at the charging point should be assigned to the vehicles. The expected arrival time is determined by the distance to be traveled to reach the charging point and, if applicable, by the usage pattern, if its influence on the travel speed is to be taken into account. In this way, available charging points are assigned to each electric vehicle.

[0020] Furthermore, for each individual vehicle within whose range there are available charging points of the charging point infrastructure, a usage-specific load prediction is made.

[0021] The usage pattern for the vehicles under consideration, resulting from the user's driving behavior, is now predicted based on past usage patterns. This can be done, for example, using a usage pattern model in the form of a hidden Markov model, which evaluates driving cycles, parking cycles, and charging cycles, and makes it possible to predict a future usage pattern of the vehicle and, in particular, a future driving behavior of the user.

[0022] The usage pattern model can be used to determine how the user is likely to use their vehicle and when a charging process is expected to occur. Simultaneously, the vehicle's usage pattern also yields a battery state prediction, specifically an expected state-of-charge profile. Depending on the usage pattern, artificial operating parameter profiles can be predicted that result in a battery load essentially corresponding to the past battery load. The battery state can also be predicted. In particular, the predicted aging at the point of reaching the charging point for the next expected charging process can be determined.

[0023] The state of health (SOH) is a key parameter for determining the remaining battery capacity and remaining lifespan of device batteries. The state of health represents a measure of the battery's aging. In the case of a device battery, battery module, or battery cell, the state of health can be expressed as the capacity retention rate (SOH-C). The SOH-C, i.e., the capacity-related state of health, is expressed as the ratio of the measured instantaneous capacity to the initial capacity of the fully charged battery and decreases with age. Alternatively, the state of health can be expressed as the increase in internal resistance (SOH-R) relative to the initial resistance of the device battery. The relative change in internal resistance (SOH-R) increases with the battery's aging.

[0024] The aging state, and thus the corresponding change in the aging state, can be simulated or determined using conventional and well-known models for determining the aging state. In particular, an electrochemical aging state model can be used, which is fundamentally based on an electrochemical battery model. Such an electrochemical battery model can comprise a system of differential equations that, based on differential equations parameterized via model parameters, models internal battery states, especially equilibrium states and, if applicable, kinetic states, using a time integration method and provides a relationship between the temporal operating parameters of the device battery, namely battery current, battery voltage, battery temperature, and state of charge, and the internal battery state.

[0025] Such electrochemical battery models are known, for example, from US20230305073A1, US20220179009A1, US20220334191A1, US20220099743A1, US 2016 / 023,566, US 2016 / 023,567, and US 2020 / 150,185. An aging state can be derived from the internal battery states. Data-based or hybrid models can also be used as aging state models to determine the change in the aging state.

[0026] Using an optimization method, scenarios can now be determined in which an electric vehicle is assigned to a charging point within a specific time window. These scenarios are selected according to a combinatorial optimization procedure, where a cost function is optimized that maximizes the energy output / throughput of all charging points over a given period, minimizes the waiting time of vehicles at the charging points, and minimizes the overall change in the aging state of all vehicle batteries.

[0027] Furthermore, the amount of energy transferred by all charging points can correspond to the sum of the amount of energy transferred by an electric vehicle assigned to a charging point, whereby the amount of energy of the electric vehicle assigned to a charging point is determined depending on a state of charge upon reaching the charging point and the desired state of charge after the charging process, wherein the state of charge upon reaching the charging point depends on a current state of charge, a usage pattern of the electric vehicle in question and the distance between the current location of the electric vehicle and the location of the charging point.

[0028] The cost function can be modeled using quantiles or percentiles, as this allows the uncertainty of the prediction to be exploited. It enables the inclusion of risks in the decision-making process via probabilities. For example, the costs C(F) for driving a vehicle D to a single charging point F and subsequently charging at that point could be as follows: C(F)=1Q50(E(D(S,U)))+k1 W(S,U)+k2ΔSOH(S,U) where Q is the quantile or 50% quantile: Q 50The energy throughput E of a vehicle D charging at this charging point F, U the predicted usage behavior of the driver, the waiting time W at the charging point in question, weighted by a predefined factor k1 if another vehicle is charging there, and the expected change in the state of aging ΔSOH, weighted by a predefined factor k2, resulting from approaching the charging point and carrying out the charging process with a typical current profile representing the driver's individual usage pattern. The energy throughput depends on the distance S from the current vehicle position to the charging point and the usage behavior (which determines the state of charge upon reaching the charging point). The waiting time is determined by the expected arrival time (e.g.,50% quantile of the arrival times) of the vehicle at the charging point, which depends on the route and usage patterns, as the time that must be waited until one or more other vehicles have finished their charging processes at that charging point.

[0029] The optimization is performed using multiple predefined charging points, which can have a variety of charging stations and / or charging devices, and a variety of vehicles for a predetermined duration. Vehicle selection is exclusive, meaning that a specific vehicle can only be assigned to one charging point. The cost function for the multiple predefined charging points is as follows: Csum=∑i=1nCi(D1…Dn) where n is the number of vehicles in the area of ​​the multiple charging points. The cost function C sum is minimized in this process.

[0030] The cost function C sumFrom the perspective of the charging infrastructure, this is defined as follows: The expected energy throughput is maximized by ranking or prioritizing across the large number of vehicles. Alternatively, a ranking can be created, the "n" best vehicles D are selected, and reservations are made for the n charging stations.

[0031] To solve the optimization problem, a gradient-free approach, such as Bayesian optimization, can be used. Alternatively, a gradient-based approach can be chosen, where, for example, the gradients of the optimization problem are estimated in the central processing unit using auto-difference to achieve a global optimum or a sufficiently good local optimum. Heuristic optimization methods, such as GridSearch, can also be used. Theoretically, brute-force methods are also conceivable, which would identify a multitude of possibilities and determine the best combinatorial solution.

[0032] Each vehicle is then assigned exactly one charging point for a specific time window. When optimizing the assignment of charging points to vehicles, the ambient temperatures at the charging points can be taken into account, as these influence the battery load during the charging process. The ambient temperature at the charging point is also predicted based on weather data for the time of charging of the assigned vehicle.

[0033] The assignments are transmitted to the respective vehicles, so that the relevant charging point is identified and its location is known in each vehicle.

[0034] If necessary, the vehicle battery temperature can be predictively conditioned to minimize charging time when the relevant charging point is reached. Conditioning can be achieved by adjusting the battery temperature through heating or cooling, depending on the ambient temperature, thus enabling particularly high charging currents. This increases the energy throughput at the charging point and reduces waiting time during the charging process. Brief description of the drawings

[0035] The embodiments are explained in more detail below with reference to the accompanying drawings. These show: Fig. 1 a schematic representation of a charging point infrastructure with multiple charging points and vehicles with their respective range polygons; Fig. 2 a schematic representation of the communication between vehicles and a central unit of the charging point infrastructure; Fig. 3. A flowchart illustrating a procedure for optimizing vehicle use of charging points; and Fig. 4 a schematic representation of a hidden Markov model to illustrate a vehicle usage pattern for predicting a likely charging process; Fig. 5 a representation of the course of the ambient temperature along with the confidence interval. Description of embodiments

[0036] Fig. Figure 1 schematically shows a charging point infrastructure 1 with charging points 2 for charging electric vehicles 4. The charging points 2 can correspond to charging stations where electric vehicles can be connected to an electrical energy source of the supply network. The electric vehicles 4 can move along a road network 3 and have a battery charge level at any given time that enables a certain range within the road network 3.

[0037] For each vehicle, a surrounding range polygon is specified, indicating the maximum range of the electric vehicle in question, i.e., an area in which the electric vehicle 4 can be located at maximum driving distance, where the maximum driving distance is determined, for example, as the driving distance when reaching a minimum charge level, e.g., 10%.

[0038] The range polygon is determined in a known manner, in particular by calculating every possible route based on a usage pattern prediction using a probabilistic forecast of the vehicle user (e.g., driving 60 km / h + / - 5 km / h), especially using a sampling or Monte Carlo approach. In the example shown (60 km / h + / - 5 km / h), a range polygon can be drawn around the driver's current coordinates. This polygon is discretized and evaluated at the support points, which are represented by drivable roads. Interpolation can be performed between these support points to outline the theoretical range along these routes.

[0039] Fig. Figure 2 schematically shows several vehicles that are in communication connection with a central unit 10 of the charging point infrastructure 1. The electric vehicles 4 each have a vehicle battery 41, which is intended for operating the traction motor 42. The vehicle battery 41 is equipped with a battery management system 43. The electric vehicles 4 have a communication device 44 to transmit measured values ​​and operating parameters of the vehicle battery and the vehicle to the central unit 10.

[0040] The electric vehicles 4 transmit the operating parameters F to the central unit 10, which specify at least parameters that influence the aging state of the vehicle battery 41. The operating parameters F include, for example, battery current, battery voltage, battery temperature, and state of charge at the cell, module, and pack levels for the vehicle battery, as well as vehicle speed, ampere-hour throughput per kilometer, driving distances, distances traveled depending on the calendar date and time of day (and day of the week), and other driving characteristics for the vehicle, such as statistical data on longitudinal and lateral acceleration, gear selection, motor speed, battery currents, vehicle load mass, environmental conditions such as weather or traffic conditions, GPS position, etc.

[0041] The operating parameters F are recorded at a rapid time interval of 1 Hz to 100 Hz and can be regularly transmitted to the central processing unit 10 in uncompressed and / or compressed form. Furthermore, the time series can be transmitted to the central processing unit 10 in blocks at intervals of several hours to several days, utilizing compression algorithms to minimize data traffic.

[0042] The central processing unit 10 includes a data processing unit 11, in which the procedure described below can be executed, and a database 12 for storing data points, model parameters, states and the like.

[0043] In Fig. Figure 3 shows a flowchart illustrating a procedure for operating the charging point infrastructure 1. The procedure is primarily executed in the central unit 10.

[0044] In step S1, each of the electric vehicles 4 is assigned charging points that are located within the respective remaining range of the electric vehicle 4 (range polygon) and are available at a possible arrival time. To determine the arrival time, the travel time until reaching the charging point can be calculated based on the distance of the respective charging point from the current location of the electric vehicle 4, using a known function of a conventional navigation system. This calculation yields the arrival time.

[0045] The range of a vehicle 4 can be determined by ensuring that it can be achieved with a given probability, e.g. 99%, with a minimum state of charge of SOC >= 10%.

[0046] Many charging points can be booked, meaning that a reservation exists for a specific vehicle (4) at a future time to perform a charging process. Each electric vehicle is thus assigned several charging points that are within its current range and available at the time of arrival. Charging points can also be assigned to multiple vehicles, even if their availability overlaps.

[0047] In the example of the Fig. Figure 1 shows that vehicle A with a state of charge (SOC) of 40% has five possible charging points within its range, of which two are expected to be occupied at the time of charging (shown with dashed lines). Three charging points are not yet reserved and are still available and reservable. Vehicle B with a state of charge (SOC) of 28% has four possible charging points within its range, of which one is expected to be occupied at the time of charging and another is not on the planned route. These are shown with dashed lines. Two charging points are not yet reserved and are still available.

[0048] Subsequently, in step S2, a future vehicle usage is generated in the form of a predicted usage pattern based on the historical usage patterns of a driver of vehicle 4. This usage can take into account operational parameters, in particular vehicle speed, torque, current, battery voltage, battery temperature, distances traveled, and navigation inputs, to create a usage prediction. Specifically, a hidden Markov model can be used as the usage pattern model. This model learns distributions and transition probabilities between operating cycles such as parking without charging, charging, driving, discharging (V2G mode), etc., and determines a sequence of the vehicle's operating cycles by randomly selecting the transition probabilities.

[0049] The predicted usage pattern is composed, according to the above procedure, from combinations of operating cycle profiles, idle cycle profiles, and charging cycle profiles of the vehicle battery 41, each representing a specific usage pattern for the vehicle battery. For this purpose, a sequence of operating cycles, idle cycles, and charging cycles for a past period, such as one month, three months, or half a year, is determined by evaluating the historical operating parameters. An operating cycle is defined as a period of virtually continuous current draw from the vehicle battery, an idle cycle as a period without current flow to or from the vehicle battery, and a charging cycle as a period with current input by the vehicle battery. The operating cycles can be assigned to one or more operating cycle profiles, the idle cycles to one or more idle cycle profiles, and the charging cycles to one or more charging cycle profiles.

[0050] By assigning the cycles determined from the historical farm size trend to the respective cycle profiles, a sequence of cycle profiles can be determined that represents the historical usage pattern.

[0051] From this, frequencies of sequences of two cycle profiles each within a past period, e.g., 3 to 12 months, can be determined. This results in a frequency distribution of cycle profile sequences that corresponds to a probability distribution for the occurrence of periods with specific usage patterns within the considered time period.

[0052] Determining the frequency distribution of sequences consisting of two cycle profiles within the resulting sequence of cycle profiles can involve weighting the proportions of cycles in the frequencies of the assigned cycle profiles according to their temporal distance from a current point in time. Thus, the frequency distribution can give less weight to profile combinations from the more distant past than to profile combinations that are only recently associated with certain cycles. When creating the frequency distribution, profile combinations for cycles further in the past can therefore be weighted less than profile combinations for more recent cycles.

[0053] To create a predicted usage pattern, the sequences of cycle profiles can be assembled based on a probability specified by a particular frequency distribution, especially by sampling, particularly by random selection from the frequency distribution. Thus, cycle profile combinations can be sampled according to the frequencies specified by the probability distribution to generate a sequence of cycle profiles.

[0054] Each of the cycle profiles occurring in this sequence can now be assigned a profile-operating parameter curve as a representative curve of the operating parameters. For this purpose, a representative operating parameter curve is assigned to each cycle profile. This representative operating parameter curve can, for example, be derived from the operating parameter curve of the most recently recorded cycle that can be assigned to the corresponding cycle profile.

[0055] Thus, a predicted operational size profile can be generated from an artificial usage pattern, with which a correspondingly predicted aging state or predicted aging state profile can be determined according to the aging state model.

[0056] The combination of profiles can be arranged according to the probability weighting from the frequency distribution, e.g. until a desired prediction horizon of, for example, one month, three months, half a year or one year is reached.

[0057] The cycle profiles can be assigned to the cycles from recorded operating parameter profiles according to aggregated characteristics. For operating cycles, these characteristics can include one or more of the following: converted Ah throughput, maximum discharge current, duration of the operating cycle, mean battery temperature, and other parameters that characterize a usage pattern of the vehicle battery and the vehicle 4 and thus influence the development of the aging state of the vehicle battery 41 and the expected start of charging.

[0058] Similarly, idle cycles can be classified according to their duration and assigned to corresponding idle cycle profiles. Charging cycles can be assigned to a corresponding charging cycle profile depending on the charging strategy used, i.e., normal charging or fast charging.

[0059] The identified cycles can be grouped into cycle profiles using a clustering method based on their characteristics. In particular, the assignment of operating cycles, idle cycles, and charging cycles to their respective cycle profiles can be achieved by applying suitable clustering methods.

[0060] Furthermore, the profile operating parameter profiles assigned to the cycle profiles can each correspond to the operating parameter profile of the cycle that is assigned to the cycle profile in question and that is closest to the centroid of the associated cluster.

[0061] The above method advantageously enables the generation of artificial operating size profiles that project the usage pattern of the vehicle battery from the past into the future.

[0062] Fig. Figure 4 shows a representation of a hidden Markov model with nodes representing cycle profiles and edges representing transition probabilities, used to predict usage patterns and generate artificial operational size profiles. This allows for the prediction of an aging state. The prediction is based on a time-integration-based aging state model capable of evaluating operational size profiles.

[0063] The hidden Markov model is continuously refined. For this purpose, the operating parameters within a predefined period up to the current point in time (e.g., one month, three months, six months, etc.) are segmented according to cycles: operating cycles, idle cycles, and charging cycles. Periods of continuous operation, i.e., a continuous discharge current greater than 0 A, are identified as operating cycles. Interruptions of a discharge current of less than a predefined minimum duration, such as 60 seconds, can be disregarded to exclude traffic light stops or recuperation phases and still assign them to the respective operating cycle. Furthermore, idle cycles can be identified as periods with a battery current of 0 A that exceed the predefined minimum duration. Charging cycles can be identified as those periods during which a charging current flows into the vehicle battery.

[0064] Operating cycle profiles can be classified according to the vehicle battery's usage, which may differ in certain characteristics. These characteristics can be aggregated usage data derived from the battery current and temperature profiles. They can also include parameters such as Ah throughput over the duration of the operating cycle, average battery temperature, maximum discharge current, and the cycle duration itself. For example, operating cycle profiles can be defined for highway driving, city driving, short trips, long-distance driving, and so on.

[0065] The assignment of operating cycles to the individual operating cycle profiles B1, B2, B3 can be carried out using a clustering method. This involves grouping the previously recorded operating cycles into operating cycle profiles according to the aggregated characteristics and assigning one operating cycle profile to each cluster.

[0066] The idle cycles, characterized by a period without current flow, can be assigned to a corresponding idle cycle profile according to their duration. For example, idle cycle profile R1 can be defined as an idle cycle between 1 and 15 minutes, idle cycle profile R2 as an idle cycle with a duration of 15 minutes to 2 hours, idle cycle profile R3 with a duration of 2 to 10 hours, and idle cycle profile R4 as an idle cycle with a duration of more than 10 hours.

[0067] Charging cycles can be defined as time periods during which a charging current is supplied. These cycles can be classified according to the selected charging strategy or charging profile, e.g., into L1, L2, etc. The charging profile essentially specifies a maximum permissible charging current for a given state of charge.

[0068] Each cycle profile is assigned a time duration.

[0069] When determining the frequency distribution, a weighting of the cycle profile sequences can also be considered, with a higher weighting the closer the time of occurrence of the respective profile combination is to the current time, i.e., the more recent the underlying farm size profile is. Accordingly, to account for any changes in usage patterns, more recent cycle profiles can be given greater consideration, for example, by weighting their contribution to the frequency distribution with a factor of two or higher.

[0070] Based on the frequency distribution of the profile combinations, a predicted usage pattern can be created for a given prediction horizon. This is achieved by sampling, i.e., by random selection, according to the probability distribution of sequences of cycle profiles as determined by the frequency distribution. This results in a constructed sequence of cycle profiles.

[0071] The sequence of randomly selected cycle profiles is determined by simple sampling, based on the frequencies derived from the frequency distribution. The cycle profiles are then sequenced for a predetermined prediction horizon or until the expected end of the vehicle battery's service life is reached.

[0072] Subsequently, corresponding operating parameter profiles are assigned to the sequenced cycle profiles. Each cycle profile is assigned a profile operating parameter profile. The assigned profile operating parameter profiles can, for example, each correspond to the operating parameter profile most recently assigned to the respective cycle profile.

[0073] Alternatively, each cycle profile can be assigned the operating parameter profile that, when clustering operating cycles, idle cycles, and charging cycles, lies closest to the centroid of the respective cluster assigned to the corresponding cycle profile. By concatenating profile operating parameter profiles according to the sequence of cycle profiles in the constructed usage pattern, a predicted artificial operating parameter profile can be created that best replicates a possible operating parameter profile.

[0074] Using the operating parameters determined in this way, battery degradation can be calculated as a progression of its aging state. The predicted usage pattern can be used to determine when the driver is likely to switch to a new charging cycle. Furthermore, load prediction is employed to predict driver-specific usage patterns, such as estimating when the driver will travel a certain distance or use the vehicle on weekends. Thus, a probabilistic prediction of usage behavior is provided for each vehicle, from which an operating parameters profile is derived, and which distance the driver is likely to travel within a given time horizon.

[0075] Furthermore, manual input from the driver (such as setting an expected departure time) can be used to refine the modeling of the driver prediction or to minimize uncertainty.

[0076] In step S3, a state prediction of the vehicle battery is performed based on the individual predicted usage patterns. For this purpose, the artificial operating characteristic curves of the vehicle battery are used with the electrochemical battery model, as described, for example, in US20220334191A1, US20230305073A1, US20230016228A1, US20220170995A1, US20220099743A1, and US20210373082A1. This results in an expected aging for each vehicle and its associated charging points, i.e., a difference between the current state of aging and the state of aging upon reaching the charging point, the expected state-of-charge curve of the vehicle battery, and, if applicable, a charging curve optimized for the charging process.

[0077] The driver-specific usage pattern and the driver-specific battery model can each be retrained and reapplied in the central unit 10. The decrease in the state of charge can also be determined, which can result from the probabilistic estimation of the sequence of operating cycles with a degree of uncertainty, as for example in Fig. 3 is shown.

[0078] In an optimization step S4, a cost function can now be minimized which is defined such that the amount of energy charged within a predetermined period is maximized across all charging points of the charging point infrastructure, the total (cumulative) waiting time of the electric vehicles at the charging points is minimized, and as a constraint, the total aging of the vehicle batteries across all vehicles is minimized.

[0079] Using the optimization method, scenarios can be identified in which an electric vehicle is assigned to a specific charging point. These scenarios are iteratively selected according to a combinatorial optimization procedure and evaluated according to a cost function that maximizes the amount of energy transferred / energy output / energy throughput of all charging points over a given period, minimizes the waiting time of vehicles at the charging points, and minimizes the overall change in the aging state of all vehicle batteries.

[0080] Furthermore, the amount of energy transferred by all charging points can correspond to the sum of the amount of energy transferred by an electric vehicle assigned to a charging point, whereby the amount of energy of the electric vehicle assigned to a charging point is determined depending on a state of charge upon reaching the charging point and the desired state of charge after the charging process, wherein the state of charge upon reaching the charging point depends on a current state of charge, a usage pattern of the electric vehicle in question and the distance between the current location of the electric vehicle and the location of the charging point.

[0081] For example, the costs C(F) for driving to a single charging point F with a vehicle D and the subsequent charging at the charging point with predefined weighting factors k1 and k2 can be as follows: C(F)=1Q50(E(D(S,U)))+k1 W(S,U)+k2ΔSOH(S,U) where Q is the quantile or 50% quantile: Q 50The energy throughput E of a vehicle D charging at this charging point F, U the predicted usage behavior of the driver, W the waiting time at the charging point if another vehicle is charging there, and ΔSOH the expected change in the vehicle's aging state due to approaching the charging point and performing the charging process. The energy throughput depends on the current state of charge, the distance S from the vehicle's current position to the charging point, and the usage behavior (which determines the state of charge upon reaching the charging point).

[0082] The waiting time is determined from the expected arrival time (e.g., 50th percentile of arrival times) of the vehicle at the charging point, which depends on the route and usage patterns, as the duration that must be waited until one or more other vehicles have completed their charging processes at this charging point.

[0083] The optimization is performed using multiple predefined charging points, which can have a variety of charging stations and / or charging devices, and a variety of vehicles for a predetermined duration. Vehicle selection is exclusive, meaning that a specific vehicle can only be assigned to one charging point. The cost function for the multiple predefined charging points is as follows: Csum=∑i=1nCi(D1…Dn) where n is the number of electric vehicles in the area of ​​the multiple charging points. The cost function C sum is minimized in the optimization process.

[0084] To solve the optimization problem, a gradient-free approach, such as Bayesian optimization, can be used. Alternatively, a gradient-based approach can be chosen, where, for example, the gradients of the optimization problem are estimated in the central processing unit using auto-diff to achieve a global optimum or a sufficiently good local optimum. Alternatively, heuristic optimization methods, such as GridSearch, can also be used.

[0085] To solve the optimization problem, a gradient-free approach, such as Bayesian optimization, can be used. Alternatively, a gradient-based approach can be chosen. As a result of the optimization, exactly one charging point is assigned to each vehicle.

[0086] In step S5, this charging point is communicated to the assigned vehicle, and a recommendation is issued to the driver to approach the charging point in question.

[0087] Preferably, the ambient temperature of the charging points is taken into account when determining aging during the charging process. The constraint of minimizing the overall aging of all vehicle batteries in the vehicles under consideration accounts for the temperature load factor accordingly. A possible ambient temperature profile, along with its confidence interval, for one of the charging points as a function of arrival time is shown in the diagram. Fig. Figure 5 illustrates this. The confidence interval influences the uncertainty of the aging prediction for the charging process when the predicted aging state is determined using an electrochemical battery model.

[0088] Once the electric vehicle reaches the charging point, a smart contract can be automatically concluded between the charging point operator and the vehicle's driver. This allows for the trading of an energy quantity within a defined time window.

[0089] Furthermore, the vehicle battery can be heated or cooled before reaching the charging point by means of temperature management known in itself, in order to enable charging that is as gentle on the battery as possible or as fast as possible within an optimal temperature window.

[0090] Within the vehicle, the conversion and regulation of the battery temperature to the target temperature trajectory now takes place in a closed control loop for the purpose of optimal battery conditioning. QUOTES INCLUDED IN THE DESCRIPTION

[0000] This list of documents cited by the applicant was automatically generated and is included solely for the reader's convenience. The list is not part of the German patent or utility model application. The DPMA accepts no liability for any errors or omissions. Cited patent literature

[0000] WO 2013 045 449 A1

[0005] US 20230305073A1 [0025, 0076] US 20220179009A1

[0025] US 20220334191A1 [0025, 0076] US 20220099743A1 [0025, 0076] US 2016 / 023,566

[0025] US 2016 / 023,567

[0025] US 2020 / 150,185

[0025] US 20230016228A1

[0076] US 20220170995A1

[0076] US 20210373082A1

[0076] Zitierte Nicht-Patentliteratur

[0000] C. Simonis, 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, doi: 10.1109 / EBCCSP.2017.8022813

[0013]

Claims

[1] Method, in particular a method, at least partially computer-implemented, for operating a charging point infrastructure (1) with multiple charging points (2) for charging electric vehicles (4), comprising the following steps: - Assigning (S1) one or more charging points (2) to the electric vehicles (4) depending on the current state of charge of the electric vehicles (4); - Performing (S4) an optimization depending on a cost function to assign exactly one charging point (2) to each electric vehicle (4), where the cost function maximizes the amount of energy transferred by all charging points (2) for a predetermined period; - Transmitting (S5) the exactly one assigned charging point (2) to each of the electric vehicles (4). [2] Method according to claim 1, wherein the cost function is designed to minimize the waiting time of electric vehicles (4) at the charging points (2) and / or to minimize the overall aging of all vehicle batteries (41) of the electric vehicles (4) by means of optimization. [3] Method according to claim 2, wherein the waiting time is determined from the expected arrival time at the charging point (2) as the time that must be waited until one or more other electric vehicles (4) have completed their charging processes at this charging point (2), wherein the arrival time depends on the distance traveled and the usage pattern. [4] Method according to one of claims 2 to 3, wherein for the charging process, when the charging point (2) is reached, a current profile is specified depending on the current profile underlying the determination of aging. [5] Method according to one of claims 1 to 4, wherein the assignment of one or more charging points to electric vehicles (4) is carried out in such a way that, in particular depending on a usage pattern which indicates a historical usage behavior, the respective charging point (2) is reached with a predetermined minimum probability with a state of charge greater than a predetermined minimum state of charge. [6] Method according to any one of claims 1 to 5, wherein for each electric vehicle (4) a usage pattern indicating historical usage behavior for the vehicle in question is provided in the form of a usage pattern model in the form of a hidden Markov model (S2), wherein a predicted usage pattern for a prediction horizon is created using the usage pattern model, wherein an arrival time is determined depending on the predicted usage pattern and depending on the distance between the current location of the electric vehicle (4) and the location of a relevant charging point (2), wherein the optimization method assigns to the electric vehicles (2) each the charging point (2) which results in the shortest waiting times depending on the arrival times. [7] Method according to any one of claims 1 to 6, wherein the optimization method corresponds to a combinatorial optimization method. [8] Method according to any one of claims 1 to 7, wherein the amount of energy transferred by all charging points (2) corresponds to the sum of the amount of energy transferred by an electric vehicle (4) assigned to a charging point (2), wherein the amount of energy of the electric vehicle (4) assigned to a charging point (2) is determined depending on a state of charge (2) upon reaching the charging point (2) and the desired state of charge after the charging process, wherein the state of charge upon reaching the charging point (2) depends on a current state of charge, a usage pattern of the electric vehicle (4) in question and the distance between the current location of the electric vehicle (4) and the location of the charging point (2). [9] Method according to any one of claims 1 to 8, wherein the amount of energy transferred is determined depending on the usage patterns of each electric vehicle (4), wherein the cost function is modeled probabilistically to take into account any risk or uncertainty or probability of prediction by the usage patterns. [10] Method according to any one of claims 1 to 9, wherein when one of the electric vehicles (4) is assigned to a charging point (2), a smart contract is created for the amount of energy to be transferred. [11] Device for carrying out one of the methods according to any one of claims 1 to 10. [12] Computer program product comprising instructions which, when the program is executed by at least one data processing device, cause it to perform the steps of the method according to any one of claims 1 to 10. [13] Machine-readable storage medium comprising instructions which, when executed by at least one data processing device, cause it to perform the steps of the method according to any one of claims 1 to 10.

Images