Methods for planning power exchange between charging infrastructure and power grid

Abstract

A method for planning power exchange between charging infrastructure and a power grid, wherein the charging infrastructure has multiple charging terminals for connecting and charging electric vehicles, enabling the electric vehicles to exchange electrical power with the power grid via the charging terminals; each electric vehicle has an electrical storage device with a variable single state of charge for receiving and outputting electrical power; all electrical storage devices connected to the charging infrastructure form the total storage device of the charging infrastructure, characterized by a total storage capacity and a total state of charge, the total storage capacity being variable and the total state of charge being variable; a prediction of the arrival time of the electric vehicle to its charging terminal is created; a prediction of the total state of charge is created based on the prediction of the arrival time as a prediction of the total state of charge for a predicted time period; and the prediction of the total state of charge is created as a time-varying curve of the total state of charge.

Description

Technical Field

[0001] This invention relates to a method for planning power exchange between charging infrastructure and the power grid. The invention also relates to charging infrastructure, particularly vehicle fleets. Background Technology

[0002] A concept is known that enables electric vehicles to be network-enabled, not only by charging the electric vehicle's electrical storage device from the power grid when needed, but also by providing the option to temporarily supply power from this electrical storage device to the power grid. This technology is summarized as the concept of "Vehicle-to-Grid" (V2G). This concept is described, for example, in International Application WO 03 / 062018A2 or in its parent U.S. Application US2005 / 0127855 A1.

[0003] However, for multiple system services, a very high level of availability is required. Therefore, a vehicle's battery is only useful to the network when it is connected to the charging point at the correct time and in a defined state of charge. In the case of passenger cars (PKW), this is generally not guaranteed, as long as they are for private use. However, in the case of electric heavy-duty vehicles in the logistics sector, deviations in driving planning can lead to changes in state of charge and timing.

[0004] In its European priority application, the European Patent Office searched for the following prior art: KOU PENG et al., “Stochastic Coordination of Plug-In Electric Vehicles and Wind Turbines in Microgrid: A Model Predictive Control Approach” (IEEE TRANSACTIONS ON SMARTGRID, IEEE, USA, Vol. 7, No. 3, May 1, 2016, pp. 1537-1551), and US2015 / 0298565 A1 and US2018 / 0170202A1. Summary of the Invention

[0005] Therefore, the object of the present invention is to solve at least one of the aforementioned problems. In particular, the object of the present invention is to improve planning, especially the planning of the timing and scope of electric vehicle availability to support the power grid. At least one alternative to the solutions known to date is to be proposed.

[0006] According to the present invention, a method is proposed according to embodiments thereof. Therefore, the method involves planning power exchange between charging infrastructure and the power grid. Thus, via the charging infrastructure, electrical power can be fed into the power grid, but also extracted from the power grid.

[0007] Charging infrastructure has multiple charging terminals for connecting electric vehicles and charging them. These charging terminals can be spatially distributed; for example, they can be found at various stations (public transportation). (e.g., city cleaning, garbage collection), warehouses (postal and parcel service providers), ports (electric boats, ferries), airport terminals (airport infrastructure vehicles, air taxis in operation). Therefore, the electric vehicles can exchange power with the power grid via these charging terminals. In this way, the charging infrastructure exchanges power with the power grid. The charging terminal can control the charging or discharging of the power storage device, but it is also considered that control or part of the control may be performed within the electric vehicle itself, although the electric vehicle is connected to the corresponding charging terminal. However, such control in the electric vehicle can also receive control signals from the charging infrastructure, specifically control signals for the power value to be received or output.

[0008] Therefore, each electric vehicle that can be connected to the charging infrastructure has an electrical memory with a single variable state of charge in order to receive or output electrical power.

[0009] All electrical storage devices connected to the charging infrastructure form the total storage of the charging infrastructure, characterized by total storage capacity and total state of charge. In this regard, the total storage can be considered as a virtual storage, in which the storage capacities of the individual electric vehicles are summed.

[0010] Total storage capacity is variable because it depends on the type and number of electric vehicles connected to the charging terminal and, consequently, the charging infrastructure at each observation point. In other words, if an electric vehicle leaves the charging infrastructure, the total storage capacity decreases, while if an electric vehicle arrives and connects, the total storage capacity increases.

[0011] Furthermore, the total state of charge is variable. In addition, electric vehicles can reach charging infrastructure and can also leave, but their corresponding individual state of charge also changes because they are being charged or discharged, and thus this also causes a variable total state of charge.

[0012] It is also proposed to create predictions of the arrival times of electric vehicles to their charging terminals. In particular, it has been recognized that even in commercially viable fleets with driving plans, the arrival times of electric vehicles to charging infrastructure cannot be determined definitively, or can only be poorly or imprecisely determined. Road traffic can also play a significant role in these uncertainties. While departure times are also important for planning power exchange between charging infrastructure and the power grid, they can generally be well-predicted.

[0013] Therefore, it is proposed here that a prediction of arrival time be created, which will be described in more detail below.

[0014] The total state of charge (SBC) forecast is created based on the predicted arrival time as a prediction of the SBC for the predicted time period.

[0015] For the purposes of explanation, it is assumed that the charging infrastructure, specifically the electric vehicles in the fleet, returns to the charging infrastructure at the end of a cycle, specifically at the end of a day, i.e., arrives at their charging terminals, which occurs gradually. Each time an electric vehicle arrives and connects to its charging terminal and, consequently, the charging infrastructure, the total capacity and total state of charge of the total memory, i.e., the virtual total memory, increase. Therefore, a prediction of the total state of charge can be derived and thus created from the prediction of the arrival time. Here, the corresponding state of charge of the arriving electric vehicle is additionally included, particularly as a prediction of the state of charge.

[0016] This paper proposes creating a total state of charge (SBC) prediction as a time-varying curve of the SBC. This time-varying curve of the SBC thus reflects how or when the electric vehicle gradually arrives at its charging terminal and, consequently, the charging infrastructure. Therefore, this time-varying curve of the SBC can be used to plan, in particular, the power exchange between the charging infrastructure and the power grid. That is, power exchange can be planned so that the available energy predicted according to the time-varying curve of the SBC can be used for feeding, where the time-varying curve indicates when and how much power can be fed in. However, the SBC, given the total storage capacity, also provides information on how much energy can be extracted from the power grid, and further, how much power can be extracted from the power grid over a specific time period to store that power in the storage. From the SBC, it can be derived how much energy must be extracted from the grid to charge the storage.

[0017] In particular, it is proposed to create a total state of charge (SPC) prediction based on the driving plans of electric vehicles. This is based specifically on the concept of using a fleet to plan the switching power, and taking into account specific information about the planned driving of each electric vehicle (the concept of electricity, and in this process, specific information about the planned trips of each electric vehicle). This achieves a significant improvement over purely statistical observations of a large number of electric vehicles. Such fleets, or groups of electric vehicles from different organizations, can thus be included in grid planning as users of a common grid, while still allowing for individual consideration of each vehicle. For fleets as units, this enables high accuracy.

[0018] According to one perspective, a total state of charge (SOC) forecast should be created at least one day prior to the forecast period. It has been particularly recognized that, especially if electric vehicles are used commercially, their behavior is repetitive daily, and planning for one day is meaningful for the next, but larger intervals, such as a week, are also considered. However, it is also additionally recognized that power or energy planning for the power grid is typically performed at least one day in advance. This planning is usually done at midday or in the early afternoon, where consumers, in particular, also announce their share or even place direct orders. Especially in this case, charging infrastructure should contribute, and therefore, advance planning or forecasting a few minutes or hours before the relevant period is almost useless, as planning in the power grid has already been completed by then. Multi-day timeframes are feasible when multi-day cycles are available in the logistics sector.

[0019] Correspondingly, arrival time forecasts are made at least one day in advance. This typically means that almost no current data about the corresponding vehicles can be included in the forecast, i.e., whether the vehicles are currently encountering congestion or taking the wrong route. In advance planning of at least one day, the corresponding electric vehicles are usually not even activated on the relevant days for which the arrival time is to be predicted. Therefore, forecasts must be guided by other criteria, which will be explained below. However, this does not preclude the readjustment of existing forecasts over a period shorter than one day, in order to also revise existing plans.

[0020] According to one approach, a single state of charge prediction is created for each individual state of charge, specifically as a time-varying curve of the individual state of charge, and additionally, a total state of charge prediction is created based on the individual state of charge predictions.

[0021] In particular, such a single state of charge prediction can involve the state of charge assumed, or predicted, upon the arrival of the respective electric vehicle. In the simplest case, the time-varying curve can appear to show that the initial value is constant over time. However, changes after arrival are also considered, such as changes caused by small self-discharges or by functions that should be maintained even when the vehicle is parked. To give just one example, heating for frost-resistant operation can be included.

[0022] In each case, the individual state of charge prediction is included in the total state of charge prediction. The total state of charge prediction specifically causes the total state of charge to increase over time, i.e., the total state of charge caused by the separately arriving electric vehicles.

[0023] According to one aspect, a prediction of arrival time is created based on at least one of the following prediction information.

[0024] As information for predicting arrival time, information is provided regarding the travel plan for the corresponding electric vehicle, the travel plan having a planned travel time and optionally a planned route. The planned travel time also includes the planned arrival time, and the planned travel time can be used as a basis for prediction and then changed according to additional prediction information regarding the predicted arrival time. To name just a few examples, depending on the vehicle type, the travel plan can also represent ferry planning, flight planning, service planning, or deployed delivery routes.

[0025] As additional predictive information for arrival time, current traffic infrastructure information for the area associated with the corresponding electric vehicle is provided. This current traffic infrastructure information includes, for example, construction and / or detours. Temporary speed restrictions due to road damage can also be included, but so can the completion of detour routes. From this, it is possible to deduce whether the electric vehicle will arrive earlier or later. This can be combined with information about travel planning, for example, starting from the arrival time according to the travel plan, but then refining the arrival time based on the traffic infrastructure information.

[0026] As information for predicting arrival times, arrival times for the corresponding electric vehicle over the past few days or for similar routes are provided. In particular, an average of the arrival times over the past few days can be calculated, and variations in the arrival times are also considered if necessary. Additional information, such as the mentioned traffic infrastructure information, can be added to estimate whether the exemplarily mentioned average can be achieved, or whether delays or earlier arrivals can be expected.

[0027] Weather forecasts can be used as predictive information. The effect of weather forecasts on arrival time can be determined directly or indirectly. Indirect considerations include, for example, the expectation of higher traffic volume and thus delays in rainy weather, and lower traffic volume in dry, sunny weather. Direct information can be derived, for example, from ice or snowfall, because even in good traffic conditions, the electric vehicle may not be able to travel quickly.

[0028] As information for predicting arrival times, information about events affecting traffic volume in areas associated with the corresponding electric vehicles is presented. Such events could be announced demonstrations or major events such as large sporting events. The latter, in particular, can cause traffic volumes to rise or fall depending on the timing. Traffic volume is low if the event attracts a large audience, but high if a large number of spectators leave the event within a narrow timeframe. Correspondingly, this affects arrival times.

[0029] As predictive information, it can also provide information about the driver of the corresponding electric vehicle and / or about the driving behavior of the corresponding electric vehicle. Experience often leads to some drivers—and always will be—returning earlier, while others return later. This particularly affects the driving behavior of the corresponding electric vehicle when it is always assigned a driver. However, it should also be considered that vehicles may behave differently for reasons such as technicality. For example, it may be easier to navigate congestion using a smaller vehicle compared to using a larger one.

[0030] As can be seen and partially explained, all the aforementioned forecast information for arrival time can also be considered in combination. All of this forecast information is typically known one day in advance and therefore can also be included in forecasts that should be created more than one day prior to the forecast period.

[0031] According to one aspect, a single state of charge prediction is created based on at least one of the following prediction information for a single state of charge.

[0032] As predictive information for individual states of charge, information about the driving plan of the corresponding electric vehicle is provided, the driving plan having a planned driving time and optionally a planned driving route. From the driving time and, if necessary, the driving route, the amount of energy consumed by the electric vehicle can be derived. Correspondingly, it is possible to estimate how its corresponding individual state of charge changes and, further, what its approximate value will be when the electric vehicle returns to its charging station.

[0033] As predictive information for a single state of charge, information is presented regarding the characteristics of the electric vehicle, particularly the characteristics of its memory. The degree of memory degradation is considered in particular. This degradation can be derived from its behavior, but also, especially, from how many charge-discharge cycles the memory has performed. Information about these cycles is generally available to the charging control device and is therefore readily available.

[0034] As predictive information for individual states of charge (SOCs), information regarding the individual SOCs stored for the corresponding electric vehicle upon arrival at the charging terminal, based on detections over the past few days or along the route, is proposed. Therefore, the measured individual SOCs recorded at past arrival times of the electric vehicle are considered. Specifically, an average value can be formed over the SOCs at past arrival times. Furthermore, variations in these SOCs can be taken into account. As long as the boundary conditions remain constant, this average value is sufficient to form a good prediction for the corresponding individual SOC.

[0035] As predictive information for a single state of charge (SBC), information regarding a pre-settable SBC at the start of driving when the corresponding electric vehicle is disconnected from its charging terminal is proposed. Therefore, an initial SBC, existing at the start of driving, is considered, and the expected consumption is then subtracted from the initial SBC. Here, it is considered that the initial SBC cannot be detected in predictions exceeding one day, but must be assumed. As an assumption, a pre-settable SBC is used, which can also be pre-set because it is ultimately determined at the end of charging. Therefore, by correspondingly controlling charging, the SBC can be pre-set.

[0036] Therefore, driving should begin immediately before the predicted time period. However, it is preferable to create the prediction one day in advance, that is, one day before the start of the predicted time period, and further, before the start of driving considering a single pre-set state of charge.

[0037] As predictive information for individual states of charge, current traffic infrastructure information for the corresponding electric vehicle's region is proposed. It is recognized that not only arrival time or its delay may be affected by this traffic infrastructure information, such as construction or detours, but also the state of charge, as consumption will change accordingly.

[0038] As predictive information for a single state of charge, weather forecasts are considered. It is also recognized that weather can directly or indirectly affect consumption and thus the state of charge. The correlation can be similar to that described above for predictive information used for time of arrival.

[0039] As predictive information for individual states of charge, information about events affecting traffic volume in areas associated with the corresponding electric vehicles is also presented. These events, as described above, can also influence consumption and thus, consequently, the state of charge.

[0040] As predictive information for individual states of charge, information regarding the driver of the corresponding electric vehicle and / or the driving behavior of the corresponding electric vehicle is also proposed. This, as mentioned above, affects not only arrival time but also consumption.

[0041] As predictive information for a single state of charge (SOC), information about the state of the electric vehicle, particularly the state of its storage device, is also proposed. Specifically, it is considered that the storage device exhibits higher self-discharge or lower efficiency with increasing losses, thus affecting the SOC. Correspondingly, the degraded storage device also reduces its charging capacity, resulting in a smaller SOC in a new but identical storage device, leading to a smaller remaining SOC at the end of the trip. The temperature of the electric vehicle's storage device can also be used as predictive information. The temperature of the storage device can also affect its capacity, which can be taken into account.

[0042] As predictive information for individual states of charge (SOCs), information regarding model inaccuracies or errors in SOC detection is presented. This inaccuracy is particularly significant if it is based on an incorrect initial SOC, which then affects the final SOC upon return to the charging station. However, it is also important to consider, and consequently, to incorrectly or inaccurately consider, past SOCs in the return process. Therefore, it is proposed that this be considered in the prediction of individual SOCs.

[0043] Even when considering all this information, which involves possible predictive information for a single state of charge, consider combinations of information, such as those directly derived from the descriptions of the individual actions.

[0044] According to one approach, after creating a total state of charge (SPO) forecast, the SPO forecast is modified based on at least one change in information. This allows for readjustment of the SPO forecast. This is also considered at the end of the day. It is recognized that while forecasts for periods longer than one day are desirable, some changes are unavoidable, and therefore it is preferable to adjust the overall forecast in the short term and, in particular, inform the grid operator or power supplier accordingly.

[0045] Specifically, the overall state of charge (SOC) forecast is altered based on changes in the predicted arrival time and / or on changes in the forecast of individual SOCs. Thus, these changes, or information about these changes, can respectively form change information or a portion of change information.

[0046] Preferably, the overall forecast is created more than one cycle in advance, especially more than one day in advance, and is based, as described above, on various assumptions regarding the operation of the corresponding electric vehicle. These assumptions specifically anticipate traffic conditions, driving behavior, or the performance of the battery storage in a time period that has not yet begun. However, if changes occur now during the driving period shortly before the forecast time, especially significant changes, such as traffic accidents causing substantial delays in arrival time, this can be considered information regarding changes in the forecast of arrival time. Predictions of changes in arrival time can also be considered information regarding changes.

[0047] Furthermore, electric vehicles typically have sensors that continuously determine the state of charge (SBC). This information can be transmitted, particularly wirelessly, and can be verified to determine whether the predicted behavior of the corresponding individual SBC is consistent with, or at least substantially consistent with, this. If a large deviation exists, it is assumed that this also leads to a deviation in the individual SBC at the time of arrival. This assumption is therefore a change in the prediction of the individual SBC, and this forms change information; with the change information, the overall SBC prediction changes, i.e., it can be matched to the changed state.

[0048] According to one aspect, after creating a prediction of the arrival time, the prediction of the arrival time is changed based on at least one additional piece of information for the arrival time. Here, this is done after creating the total state of charge prediction, and therefore similarly to when changing the total state of charge prediction. As possible additional information for the arrival time, the following information is proposed for use individually or in combination with each other:

[0049] - Information regarding changes to driving plans, especially manual changes. It has been recognized that short-term driving plan changes should also be considered, for example, during special events or short detours, and therefore it is suggested that this be taken into account and that arrival time predictions be verified as necessary.

[0050] - Information regarding detected traffic conditions and / or short-term forecasts of traffic conditions. It is recognized here that short-term changes in traffic conditions can also occur, which should be considered here. Furthermore, accidents can also trigger short-term forecasts of traffic conditions, as congestion typically follows an accident, whether at the accident site or along known detour routes.

[0051] - Information regarding the current weather or short-term weather forecasts created after the creation of the total state of charge (SBC) forecast. It is also recognized that while current weather forecasts last for several days, they can still contain errors and may change within a day. This necessitates readjusting the time of arrival forecast. Current weather phenomena, such as snowfall, can be considered, and this can also be seen in short-term forecasts exceeding one or several hours, especially less than ten hours.

[0052] - Information regarding the deviation between the electric vehicle's current location and the planned location according to the driving plan. It is recognized that deviations can be identified during a day of executing the driving plan. Typically, delays cannot be recovered within the course of the day, or at least it can be assessed whether or to what extent such delays can still be recovered.

[0053] - Information regarding the loading or average occupancy of electric vehicles. Forecasts are also based on averages or occupancy in principle. However, if there are biases, this will affect arrival times. Especially in short-distance traffic where vehicles have many passengers, a high number of passengers indicates that vehicles stop more frequently and / or for longer periods to pick up and drop off. A similar consideration exists for the loading of electric vehicles, which correspondingly requires more time to unload. Vehicle driving dynamics may also be affected and thus suggested as additional information used to readjust arrival time forecasts.

[0054] One approach proposes that, after creating a single state of charge (SOC) prediction, the prediction is modified based on at least one additional piece of information for that SOC. It is also recognized that short-term changes occur after the prediction is created and can individually affect the individual SOCs. Therefore, a readjustment is proposed. For this purpose, the following additional information for the individual SOC is considered meaningful for both single and combined observations:

[0055] - Information regarding detected traffic conditions and / or short-term predictions of traffic conditions. It has been recognized that such short-term changes in traffic conditions can also be detected and can affect individual states of charge (SOC). Particularly sudden congestion leading to detours can affect SOC.

[0056] - Information regarding the current weather or short-term weather forecasts created after the creation of the total state of charge (PoC) forecast. These short-term weather changes can also affect the PoC and thus influence the matching of individual PoC forecasts. For example, cold weather may reduce battery efficiency and increase heater consumption. The same applies to hot and sunny weather, which may lead to higher consumption in air conditioning.

[0057] - Information regarding the deviation between the electric vehicle's current location and the planned location according to the driving plan. Driving plan deviations typically indicate that the vehicle is driving differently from the plan and thus incurring other losses, affecting the state of charge (SBC). Specifically, this is achieved by location monitoring, particularly using GPS, to determine the electric vehicle's location and compare it to the driving plan. The impact on the SBC can then be derived, for example, from empirical values. This change in location, thus the vehicle lagging behind the original driving plan, can also typically indicate that further delays and changes in driving plan execution are expected. This can also be used to adjust individual SBC processes.

[0058] - Information regarding the load or average occupancy of the electric vehicle. Both affect the weight of the electric vehicle and, consequently, its consumption. Therefore, readjusting individual state-of-charge forecasts is considered meaningful.

[0059] - Regarding the individual state of charge (SOC) of the electric vehicle, especially in conjunction with the associated time and / or associated location. This allows for comparison with the individual SOC or individual SOC change curve on which the original prediction is based. Accordingly, adjustments to the individual SOC prediction can be performed. In particular, real-time detection of the individual SOC can be performed and thus immediately used to adjust the individual SOC prediction.

[0060] However, it is also considered to record the time and / or location along with the individual state of charge. The time or location is the associated time or associated location. Together, it allows for better comparison with the driving plan. In particular, it is possible to examine when and at what location what actual state of charge occurs. For example, if the detected state of charge lags behind the predicted or assumed state of charge at that moment, but the vehicle is ahead relative to its position according to the driving plan, this has different consequences for the predicted individual state of charge than (ideally, no consequences) when the detected state of charge lags behind and the vehicle is otherwise on its way according to the driving plan, or even when the vehicle is behind and not following its driving plan. Therefore, it is possible to perform a readjustment of the prediction of the individual state of charge in the same way (which has different results for the predicted individual state of charge – ideally, especially no consequences – if the obtained state of charge trajectory and the vehicle are otherwise traveling according to a schedule or even a trajectory other than the schedule).

[0061] According to one aspect, predictions of arrival time, individual state of charge, and / or total state of charge are created and / or modified using corresponding prediction models, namely, prediction models for arrival time, prediction models for individual states of charge, or prediction models for total state of charge. Specifically, at least one of the aforementioned functionalities is implemented in such prediction models.

[0062] The prediction model for the total state of charge (SPC) can be specifically configured such that it takes the predicted arrival time and the predicted individual states of charge (IUCNs) as input variables. The SPC prediction model can specifically determine the SPC change curve by summing the associated IUCNs to the total SPC at the corresponding time based on the obtained arrival time. This allows for obtaining the change in the SPC change curve for each new arrival time. The initial value of the SPC change curve can be zero before summing the first value of the IUCNs.

[0063] Predictive models for arrival times can, for example, be based on empirical values ​​or average arrival times for the corresponding vehicles, plus a determined variation with a weighting factor that can also be 1. The variation can be the deviation between the last detected arrival time and the arrival time determined as an average to date, and this deviation is the variation to be considered. For the example described, this variation can be multiplied by a positive weighting factor <1 and then added.

[0064] A weighting factor of 0.1 can be used, for example, to cause an asymptotic approximation of the new value. If the same new arrival time is always checked ten times, the new value will have reduced its deviation by 63% after ten steps. Thus, a new value is derived, which can be assumed as a new average and thus gradually used as a basis in subsequent calculations. However, if special cases are considered, such as construction that results in predictable detours and requires well-calculated time, this can be directly added with a weighting factor of 1.

[0065] This can be done in exactly the same way in a single state of charge prediction and, consequently, in the prediction model. It is also possible to start with an initial average value, or alternatively with an empirical value as an initial value, and then gradually track that empirical value with random bias, or add to it according to the sign if there is a clearly calculable bias, or subtract it, of course.

[0066] The corresponding model thus maps the performance of the variable to be predicted to the performance of the input variables considered separately. This allows for the consideration of the performance used for prediction. The prediction model for arrival time specifically maps the performance of arrival time to the corresponding input variables, i.e., the corresponding information or additional information. The same applies to the prediction model for a single state of charge, which maps the performance of a single state of charge to the corresponding input variables, i.e., the corresponding information or additional information. These two models can be combined in the prediction model for the total state of charge.

[0067] According to one aspect, a method is proposed to create and / or modify arrival time predictions using a prediction model for arrival time. The prediction model considers at least one piece of prediction information for arrival time to create an arrival time prediction as model input, and the prediction model considers at least one piece of additional information for arrival time to modify the arrival time prediction as another input variable. In particular, it is proposed to consider at least one piece of additional information for arrival time as a disturbance variable.

[0068] Therefore, the forecasting model can perform forecasts, especially long before a cycle, particularly long before a day, based on at least one forecasting piece of information for arrival time, and additionally, adjust the forecasts based on at least one additional piece of information. In this way, these two aspects—creating a forecast for arrival time and changing or readjusting a forecast for arrival time—can be unified within the forecasting model. The forecasting model can output multiple forecasts, either separately or sequentially. Thus, forecasts that have not yet used additional information (because such additional information may not even exist) can be output for planning, i.e., for long-term planning exceeding one day, particularly for grid operators or electricity marketers. However, during this period, when a shorter time period begins, the forecast can be readjusted and then output again, i.e., as an adjusted forecast. The outputs can be sequential or performed in two separate output channels.

[0069] In this sense, according to one aspect, a single state of charge (SOC) prediction is created and / or modified using a prediction model for a single SOC. This involves considering at least one piece of prediction information for the single SOC to create the single SOC prediction as a model input, and considering at least one piece of additional information for the single SOC to modify the SOC prediction as another input variable, particularly as a disturbance variable. In this respect, the prediction model for the single SOC can operate in the same sense as the prediction model for the time of arrival, as described above.

[0070] This also applies to both models, considering the corresponding additional information as disturbance variables. Therefore, it can also be achieved using a control technique model, which obtains predictive information as input variables and additional information as disturbance variables. For such a controlled system, the additional information thus forms disturbance variables, but these disturbance variables are input into the model in the sense of input variables.

[0071] This is also based on the understanding that additional information actually functions as a confounding variable. Additional information can be unexpected and / or random. While much of the additional information in each observation may not be random in nature, it can, in general, become a random variable over many days, months, or even years. Therefore, it is proposed to consider it as a confounding variable.

[0072] Similarly, according to one aspect, a method is proposed to create and / or modify total state of charge (TSC) predictions using a prediction model for TSC, wherein the TSC prediction model considers at least one prediction of arrival time and / or individual state of charge as model inputs for the electric vehicle under consideration, and considers at least one additional piece of information for arrival time and / or individual state of charge as additional input variables, particularly as disturbance variables. Therefore, all of this information can be included in the overall model, i.e., the TSC prediction model.

[0073] Alternatively, the prediction model for the total state of charge (TBC) can be proposed to include prediction models for arrival time and individual states of charge for the electric vehicle under consideration, combined into a new model and / or simulated as a combined model. In this way, the prediction model for the TBC can be created in a simple manner, namely, by including prediction models for arrival time and individual states of charge. One prediction model from each of these models can be provided for each electric vehicle. However, it is also possible to propose using the same model for each electric vehicle, but with adjusted data sets. Furthermore, this model can be implemented in a process computer, not only for these individual models but also for the overall prediction model for the TBC.

[0074] According to one approach, the creation of an adaptively altered arrival time prediction is achieved by comparing the predicted arrival time for a given time period with the actual arrival time occurring within that time period and adjusting the future prediction based on the comparison. This adaptation is repeated, particularly every cycle and every day, i.e., the comparison is repeated, particularly every cycle and every day, and the adjustment is repeated, particularly every cycle and every day, in relation to the comparison.

[0075] It has been recognized that arrival times, based on the information and probabilistic considerations used, may still be imprecise, and therefore additional improvements in arrival time prediction can be achieved through the proposed adaptation. In particular, it is possible to determine the adaptation factor or the correction value of the adaptation, or to adjust the prediction model in its parameterization.

[0076] In particular, it allows for improvements or adaptations in each iteration. Specifically, the adaptation factors or additional correction values, or the parameters of the prediction model, can asymptotically approach the ideal values. This can be achieved, for example, with the proposed additional correction values, as follows.

[0077] When comparing the predicted arrival time and the detected arrival time, a difference is formed, and this difference is multiplied by a weighting factor between zero and 1, and the result is added to an existing additional correction value. A value of zero can be used as the initial value for the existing additional correction value. For example, if the weighting factor is set to 0.1, and if the same difference is always obtained between the predicted and actual arrival times without considering the additional correction value, then in this example, a correction value for the difference will be asymptotically approximated, where the correction value approaches the ideal value by 63% after ten steps when the weighting factor is 0.1, as exemplarily mentioned.

[0078] Similarly, this can be done with other parameters. In the case of the adaptation factor, for example, the additional correction value can be converted into an adaptation factor. To do this, the predicted arrival time can be added to the additional correction factor and divided by the predicted arrival time without the correction factor. This yields the adaptation factor, which is specifically in the range of 0.9 to 1.1, and is also related to the time base of the arrival time. For this purpose, a time base should be determined, defined as time zero, because unlike the state of charge, this time does not have a natural initial point. For example, a time one hour before the expected arrival time can be used as the time base. Therefore, the expected arrival time is one hour relative to the time base.

[0079] Generally, each adaptation correction based on the detected arrival time can be based on an arbitrary parameterization or a model whose parameters or degrees of freedom are optimized using recorded observation data and past forecasts.

[0080] Alternatively, it is proposed to adapt the creation of individual state of charge predictions by comparing the individual state of charge predictions for a prediction period with the individual state of charge that actually occurs during the prediction period and adjusting the future creation of the predictions based on the comparisons, wherein the adaptation is repeated, especially every cycle and especially every day, i.e., the comparisons are repeated, especially every cycle and especially every day, and the adjustments are repeated, especially every cycle and especially every day, in relation to the comparisons.

[0081] This can therefore be done similarly to the adaptation of arrival time predictions. Similar to a time reference, except that there is no need to define a reference for a single state of charge, because the state of charge actually has an absolute zero value, i.e., complete discharge. In other words, 0% state of charge is defined.

[0082] Therefore, the same recommendations are made here as for the adaptation of the time of arrival prediction, in particular for creating adaptation factors or additional correction values ​​for individual state of charge prediction, or additionally adjusting the parameters of the prediction model for individual state of charge.

[0083] According to one approach, the prediction model for arrival time is implemented as an adaptive model, which adjusts one or more model parameters based on a comparison between the predicted arrival time and the actual arrival time. As already described, this allows for improved predictions.

[0084] For example, model parameters can be adjusted so that they are calculated from the comparison between the predicted arrival time and the actual arrival time, thus determining, for example, how the model parameters must be set to obtain the actual arrival time in place of the predicted arrival time. This allows the derivation of a comparison or difference between the model parameters and the imagined ideal model parameters. In this way, fitness factors or additional corrections can be determined for the model parameters individually, as explained in the different contexts above. Alternatively, fitness factors or additional corrections can be determined for the model as a whole, for example, at the model output.

[0085] Similarly, as one aspect, it is proposed that the prediction model for a single state of charge is implemented as an adapted model, which adjusts one or more model parameters based on a comparison between the predicted single state of charge and the single state of charge that actually occurs. This can be done similarly in the adapted prediction model for arrival time.

[0086] Model parameters can also be adapted by identifying model parameters based on the detected input / output variables and the input / output performance of the corresponding model, and then adapting the current model parameters to the (newly) identified model parameters through a transfer function. In the linear case, model parameter identification can be performed by establishing equations for the model parameters and solving the system of equations using pseudo-inverses.

[0087] According to one approach, probability assessments are performed when creating a total state of charge (SBC) prediction and / or when creating an arrival time prediction and / or when creating a single SBC prediction. Therefore, the corresponding prediction is associated with a probability assessment, particularly a probability distribution. The probability assessment, particularly the probability distribution, can also form the prediction itself. For example, the probability distribution can be specified for the arrival time, and the probability distribution indicates the probability that the electric vehicle will arrive at the latest at the predicted or preset arrival time. The value could be, for example, 90%. For a later arrival time, such as a quarter hour later, the probability assessment indicates the probability that the electric vehicle will arrive at the latest at the second, i.e., 15-minute later arrival time. For this type of assessment, the value in the mentioned example must logically be higher than 90% and, for example, could be 95%. In this way and by this method, multiple values ​​can be determined and a probability distribution can be derived.

[0088] Similarly, it is possible to evaluate individual state of charge predictions by stating the probability that the electric vehicle will have at least the corresponding state of charge for different individual states of charge. At 0%, the value is logically 1, and for 100% it approaches zero. This can be based on the time the electric vehicle arrives at its charging terminal.

[0089] For total state of charge (TBC) prediction, a similar probability assessment and specific probability distribution are proposed. TBC prediction is slightly more complex in principle because it is set as a time-varying curve of the TBC. This allows for a probability assessment of the TBC at approximately each moment in the prediction period. Preferably, the time-varying curve of the TBC can be assigned to predetermined probability values, particularly p-90, p-95, p-98, or p-99.

[0090] The p-99 value indicates what state of charge the total charge memory has reached at least with a 99% probability at a given time. The same applies to the other values ​​mentioned. For the same time, since the probability distribution can also have state of charge values ​​for other probabilities, it is preferable to use only the state of charge for a specific probability value for each time moment, i.e., the p-99 value as exemplarily mentioned. Thus, the time-varying curve of the total state of charge reflects the time-varying curves for all p-99 values.

[0091] According to one approach, for each electric vehicle, an arrival time prediction is created together with a probability assessment, particularly with a probability distribution, and a single state-of-charge prediction is also created together with the probability assessment, particularly with the probability distribution, and these are combined separately to form a vehicle prediction. Therefore, the vehicle prediction includes both the arrival time prediction of the electric vehicle along with the probability assessment and the single state-of-charge prediction of the same electric vehicle along with the probability assessment.

[0092] Therefore, such vehicle predictions are created for multiple vehicles, especially all vehicles, in a fleet or charging infrastructure, and the total state of charge prediction is determined based on all the vehicle predictions together with probability assessments, especially with probability distributions.

[0093] Alternatively, it is proposed that the total state of charge (TBC) prediction be determined based on all vehicle predictions along with probability assessments for predetermined probability values. Therefore, the TBC prediction variation curve can be described with high probability, and based on this, switching power can be planned. In particular, it is also proposed here to describe the TBC prediction variation curve for the px values, especially for p-90, p-95, p-98, or p-99 values. The p-90, p-95, p-98, or p-99 values ​​also represent similar values ​​here.

[0094] According to one approach, power exchange planning is conducted to provide exchange power for communication with the power grid, wherein the exchange power is determined based on total state of charge (SPC) prediction. Specifically, an exchange power variation curve or exchange power band is provided and defined as the exchange power. The SPC prediction illustrates the time-varying curve and, based on this, the corresponding exchange power can be provided, ensuring that the SPC prediction, including the desired state of charge to be achieved at the end of the prediction period, can provide the corresponding power.

[0095] By pre-setting the power exchange band, fluctuations can also be considered if necessary. This concept is specifically based on the ability to consider the needs or pre-set requirements of the power grid, thus making it more likely that there will be power oversupply at some times and power demand at others. The power exchange, in particular, the power exchange variation curve or the power exchange band, can be matched to this and provide power accordingly. The predicted total state of charge (SPC) variation curve can form important boundary conditions for this, namely, power exchange to prevent excessively high or low SPC.

[0096] According to the present invention, a charging infrastructure is proposed, namely, a charging infrastructure configured for planning power exchange between the charging infrastructure and the power grid, wherein...

[0097] The charging infrastructure has multiple charging terminals for connecting and charging electric vehicles, enabling the electric vehicles to exchange electrical power with the power grid via the charging terminals.

[0098] - Each electric vehicle has an electrical memory with a single variable state of charge for receiving and outputting electrical power.

[0099] - All power storage devices connected to the charging infrastructure form the total storage of the charging infrastructure, characterized by total storage capacity and total state of charge, wherein

[0100] - The total memory capacity is variable, and

[0101] -The total state of charge is variable, where

[0102] -Charging infrastructure configuration is used, especially with control units for,

[0103] - Create predictions of the arrival time of electric vehicles at their charging terminals, and

[0104] - Create a total state of charge (SOC) forecast based on the predicted arrival time as a prediction of the SOC for the predicted time period, where

[0105] - Charging infrastructure configuration is used to create a total state of charge (TNC) prediction as a time-varying curve of TNC, and the charging infrastructure is specifically configured for...

[0106] - Create total state of charge predictions based on the electric vehicle's driving plan.

[0107] The charging infrastructure is therefore configured to execute at least one method according to one of the embodiments described above. For this purpose, a control unit can be provided, for example configured as a process computer and connected to the charging terminal, to exchange information with the charging terminal. This can be done wired or wirelessly. The charging infrastructure is thus configured to execute one of the methods, such that a corresponding program is implemented in the control unit. Furthermore, a connection for transmitting data to the electric vehicle, i.e., particularly a wireless connection, can exist during operation. At least one point is made that, with such a data connection, the control unit can receive information from the electric vehicle via the data connection. Attached Figure Description

[0108] The invention will now be described in detail below with reference to the accompanying drawings.

[0109] Figure 1 A flowchart illustrating the proposed method is shown.

[0110] Figure 2 Show Figure 1 The flowchart is used to illustrate a part of the range of states of charge and its applications.

[0111] Figure 3 The illustration shows a fleet of vehicles equipped with charging infrastructure.

[0112] Figure 4 The schematic diagram illustrates the structure used to predict the total state of charge.

[0113] Figure 5 Schematic and exemplary illustrations are shown according to Figure 4 A feasible structure for the prediction module.

[0114] Figure 6 Schematic and exemplary illustrations show that according to Figure 5 The feasibility of the adaptation module. Detailed Implementation

[0115] Figure 1The flowchart illustrates the principle of the proposed method. The concept of the method is to fully utilize the storage capacity of a fleet of electric vehicles to temporarily supply power to the power grid in a demand-related manner, and also to selectively control the extraction of power from the grid for charging the storage of the fleet's electric vehicles, taking the grid into account. Electricity price optimization can also be achieved with this method. Therefore, the fleet can be controlled to draw the required charging current as cheaply as possible while adhering to technical boundary conditions. In particular, it has been recognized that this depends on good planning of technical resources, and good planning can be achieved through the proposed method, especially through good forecasting.

[0116] The fleet of electric vehicles is represented by fleet module 102. This fleet module contains information about the electric vehicles. This information includes details about the memory of each respective electric vehicle, i.e., detailed information, i.e., individual information for each memory of each electric vehicle. Similarly, information about the state of charge of the electrical memory that the respective memory should have, particularly before starting the electric vehicle in the morning, is included. This concept is also specifically based on the use of a fleet of electric vehicles for better planning purposes, exemplified, for instance, as electric buses used for short-distance public transportation.

[0117] Correspondingly, information regarding the departure time of the respective electric vehicles can be obtained in the fleet module 102. The departure time specifically determines when the corresponding memory must reach its state of charge. However, the departure time also determines when the memory of the corresponding electric vehicle will no longer be connected to the power supply network.

[0118] However, the fleet module can also contain information about the maximum charging or discharging power of each of these memories. This information can be stored in the fleet module 102, and can also be updated.

[0119] Furthermore, the input data module 104 is illustrated in a diagrammatic manner. The input data module 104 specifically contains data or information that changes drastically, particularly information or data that can change daily and / or be updated daily. If the fleet's electric vehicles are electric buses, and if a driving plan exists that can serve as a bus driving plan, then this includes a driving plan for the electric vehicles. However, weather data, especially weather forecasts, can also be provided through the input data module 104. This data is input into the fleet module 102 and can be combined or offset with the data present in the fleet module 102. Here, based on this input data, such as the bus driving plan, it is possible to calculate or at least estimate when the corresponding electric vehicle will arrive at the fleet (Fuhrpark) and then provide information for receiving power from the power grid or for outputting power to the power grid.

[0120] At the moment the corresponding electric vehicle arrives at the fleet, especially when it connects to the charging infrastructure, the state of charge (SOC) of each power storage device can also be calculated. This can be done using data from input data module 104, and additionally, when the electric vehicle arrives at the fleet, other data, such as the SOC or expected SOC of the electric vehicle, can be calculated from existing data in fleet module 102. For this calculation, knowledge about the storage devices, particularly their size, can be used. This information is then retrieved from fleet module 102.

[0121] In this manner and method, forecasts or predictions are created. This can also be done in the fleet module 102. It should be noted that, in principle, the method can also be implemented differently from the absolute use of the described modules, such as fleet module 102 and input data module 104. These modules can, for example, be linked in other ways, or multiple input interfaces can exist to receive corresponding input data. For example, depending on the purpose, weather forecasts can be obtained from sources different from bus driving plans or another driving plan. In this regard, Figure 1 The flowchart is used for basic explanation.

[0122] Then, the charging infrastructure can provide forecasts or predictions to the fleet. This charging infrastructure is represented by infrastructure module 106. Infrastructure module 106 can contain information about the charging infrastructure, specifically the maximum charging power preset by the corresponding charging terminal. The charging terminal is also part of the charging infrastructure. The availability of power storage can also be known in infrastructure module 106, particularly by the charging terminal providing information on whether an electric vehicle is connected to the charging terminal, and, if necessary, information on the type of storage.

[0123] The data is also partially derived from the electrical memory and can be obtained from the fleet module 102 when necessary.

[0124] Infrastructure module 106 also obtains predictive data, namely, the arrival time of the electric vehicle to the charging terminal and various states of charge of the electric vehicle, once the electric vehicle arrives at the charging station and connects to it. This data is then transmitted and further processed as predictions.

[0125] In addition, a power grid connection module 108 is provided, which specifically provides information about the power grid and the power grid connection point used. The power grid connection point is the final connection point through which the charging terminal connects to the power grid. Here, the charging infrastructure can connect to the power grid via one or more power grid connection points.

[0126] The grid connection module 108 can provide data on maximum feed-in power, desired voltage, and desired reactive power. The maximum power, also known as maximum exchange power, is the maximum power that can be fed into the grid or extracted from the power supply grid. The desired voltage specifically refers to the voltage of the power supply grid at the corresponding grid connection point. The desired reactive power can be the reactive power preset by the grid operator and / or derived from data from the power supply grid. This can include the level of reactive power to be fed in, which can be determined based on the current grid voltage.

[0127] Able to provide all the aforementioned data, i.e., according to Figure 1 The flowchart at infrastructure module 106 provides all the data mentioned. However, such module division is not necessarily required here.

[0128] The grid connection module 108 also illustrates that not only the power provided by the electric vehicle or its memory, i.e., active power, but also the reactive power provided, can be output at one or more grid connection points so that it can be fed into or extracted from the power grid.

[0129] From the data, the maximum exchange power for maximum exchange with the power grid can be determined or received, particularly in infrastructure module 106. The maximum available active power can also be determined. The maximum available active power can depend particularly on the state of charge of the electric vehicle. The same applies to available reactive power, which, while requiring a small amount of storage capacity in the power storage device, still requires at least one small amount of storage capacity and also requires idle capacity for feeding the corresponding current into the power grid.

[0130] In order to coordinate the various storage devices of the electric vehicles in the fleet, and then to provide the fleet's common switching power to the power grid, a charging control unit is specially provided, which can also be called an aggregator or may contain an aggregator.

[0131] Using the above information, the charging control unit or aggregator creates a state of charge range. This is illustrated in the aggregation module 110.

[0132] The aggregator, represented by aggregation module 110, is used in the above discussion regarding... Figure 1 All the information described herein is used to determine the range of states of charge (SOC). In particular, a prediction is made for this SOC region. The SOC range is defined as a range or band extended by the sum of the SOC and time. This is indicated in the above diagram in aggregation module 110. There, the sum of the SOC is plotted on the vertical axis with respect to time on the horizontal axis. The SOC range is here formed by an upper and a lower limit. The sum of the SOC extends within this range, i.e., between the upper and lower limits.

[0133] Regarding this, and also regarding the lower diagram of the aggregation module, and also regarding the details of the switching power module 112 shown below the aggregation module 110, the following are related... Figure 2 Let's elaborate on this together.

[0134] In any case, the aggregator illustrated in the lower diagram of the aggregation module 110 can preset different switching powers and thus preset the switching energy at different times, which can be shown as a third dimension in the diagram. This feasible switching power or the proportion of switching power can be transferred to the switching power module 112. Similarly, the switching power module 112 can preset the switching power and thus preset the switching power variation curve at different times based on the availability or supply in the power grid. This is indicated in the diagram of the switching power module 112, thus providing different power levels related to time. However, the switching power variation curve does not have to be stepped out step by step. It can also be stepped out continuously.

[0135] This allows for the calculation or preset of power packets or power change curves for the exchanged power based on the state of charge range and the data provided by or in conjunction with the grid connection module 108 as described above.

[0136] In this regard, the power module 112 for exchanging power, or the variation curve exemplarily shown therein, can be understood as a result of the proposed method, or at least an intermediate or partial result. From data regarding the fleet according to the fleet module 102, data regarding the charging infrastructure according to the infrastructure module 106, together with data provided by the input data module 104, it is possible to determine, in particular, the predicted state of charge range.

[0137] Based on the state of charge range illustrated in the upper diagram of aggregation module 110, the curve of exchange power variation with time can be determined, preferably also as a prediction. For this purpose, data on the demand and supply of the power grid can be used, showing what level of exchange power is advantageous at what time. The state of charge range provides a range of variation for this purpose, which can also be called a flexibility space, and thus a predefined framework within which the curve of exchange power variation can be predefined. This predefined exchange power ensures that the total state of charge remains within the state of charge range.

[0138] Figure 2 Show Figure 1 The flowchart shows a portion of the process, namely the aggregation module 110 and the switching power module 112. The aggregation module 110 has a state of charge diagram 220 in the upper region, which illustrates the range of states of charge. A variant diagram 222 is shown in the lower region of the aggregation module 110. The variant diagram is based on the state of charge diagram 220 and additionally exemplarily illustrates a pair of variant possibilities for the switching power.

[0139] The exchange power graph 224 is shown in the exchange power module 112. This graph illustrates the possible changes in exchange power over time, i.e., the exchange power variation curve. All three graphs—the state-of-charge graph 220, the variant graph 222, and the exchange power graph 224—share the same time axis. Specifically, the starting charging time t for the state-of-charge range is plotted to illustrate this. S and target charging time t Z Furthermore, the stated time is also plotted on the other two charts using corresponding vertical lines. Start of charging time t S and target charging time t Z Therefore, the provision of time period T was extended. B The time period is provided only in the switching power module 112 for better overview. According to the way time is observed, the time period is the period during which switching power is provided. The switching power module 112 illustrates this. To plan this, a prediction is made for the time period, such that if a prediction is made, the time period is the predicted time period. The time period can be determined as a fixed, recurring period, or its initial time may fluctuate according to the prediction. Thus, the charging start time t... S It does not need to correspond to the initial time of the provided time period.

[0140] In the state of charge (SOC) chart 220, the total SOC is plotted essentially with respect to time t. A SOC region 226 is shown in the chart. SOC region 226 begins at a start charging point 228 and ends at a target charging point 230. The start charging point 228 is characterized by the value of the total SOC and the start charging time t. S Not only the total state of charge level but also the time t at which charging begins. S Both can be changed and are preferably determined by prediction, as is the case with... Figure 1 And there is a description of the data provided in conjunction with the input data module 104, which is specifically related to the fleet module 102.

[0141] Then, the total state of charge (SoC) extends from the starting charging point 228 to the target charging point 230. The target charging point 230 is characterized by the target charging time t. Z And it depends on the total state of charge of the SoC.

[0142] Start of charging time t S It is variable and related to when the electric vehicles actually return to the convoy, while the target charging time t Z It is possible to determine very precisely when the electric vehicle will leave according to the plan.

[0143] The trajectory of the total state of charge (SOC) of the SoC from the starting charging point 228 to the target charging point 230 is flexible. Recognizing this, and for this purpose proposed, only predefined limits are set, which extend the flexibility, i.e., the SOC range 226. The SOC range 226 here has a time-dependent upper limit 232 and a time-dependent lower limit 234. The time-dependent upper limit 232 can sometimes reach a value of 100%. Thus, all electrical memories are fully charged. Furthermore, the lower limit 234 can at least sometimes reach a lower value of 0%. However, this is for illustrative purposes, and generally, it is undesirable to completely discharge all memories, as this could damage them. Therefore, a value different from 0%, such as 20%, can also be chosen as the minimum value of the lower limit 234. The same applies to the upper limit 232; for the upper limit, for example, a maximum value of 90% can be chosen instead of a maximum value of 100%. The minimum and maximum values ​​can also be selected based on the corresponding values ​​of each electrical memory.

[0144] The state of charge (SCC) diagram 220 thus illustrates that the SCC range 226 is extended by the upper limit 232 and the lower limit 234, within which the total SCC can vary. For this purpose, a total SCC variation curve 236 is plotted. The total SCC variation curve can also be synonymously referred to as the total SCC variation curve. Therefore, the total SCC variation curve reflects the time-varying curve of the total SCC. The total SCC can be synonymously referred to as the total SCC.

[0145] Furthermore, as can be seen from the state of charge (SOC) chart 220, the SOC range 226 also changes with the change in the start-charge point 228, at least within its initial range near the start-charge point 228. Correspondingly, the SOC range 226 is also related to the prediction of the total SOC at the start of the SOC range 226, and the SOC range is also related to the prediction at the start-charge time t. S This is related to predictions in this area.

[0146] Variant chart 222 incorporates state-of-charge chart 220, where the coordinate axis of the total state of charge (SoC) points into the plotting plane. The time axis is retained, and a coordinate axis for power P is added. The power P, shown in variant chart 222, indicates at which inflection point of the upper limit 232 or lower limit 234 how much switching power can be output or received. Double arrows D1-D6 are drawn for this purpose.

[0147] Double arrow D1 relates to the starting charging point 228 and shows that it is currently possible to supply not only positive exchange power but also negative exchange power to the same extent. Double arrow D2 is drawn at the upper limit 232, where the exchange power can be increased further, but in particular, it is also possible to supply exchange power much more strongly in negative values. In the case of double arrow D3, and the same applies to double arrow D4, the upper limit 232 reaches its maximum value and then only negative exchange power can still be supplied. In the case of double arrow D4, not only can negative exchange power be supplied but must be supplied in order to also achieve the target charging point 230. However, the target charging point 230 can also be preset to a minimum value, and thus the total state of charge can also be higher than the target charging point 230, and thus double arrow D4 also only indicates possible negative exchange power. The exchange power can also be zero, but it cannot be positive, as double arrow D4 shows.

[0148] To reiterate, positive switching power is the power used to charge the memory; therefore, positive switching power is the power drawn from the grid and stored in the electrical memory.

[0149] The double arrows D5 and D6 indicate that only positive switching power is possible, since the lower limit 234 has reached its minimum value.

[0150] In this regard, the double arrows D1-D6 merely illustrate feasibility and help define the framework in which the power exchange can actually be altered.

[0151] The total state of charge (SBC) change curve 236 thus illustrates a series of multiple SBCs. For each of these SBCs, i.e., for each point on the SBC change curve 236, a range or minimum and maximum value can be preset. For each value, a trend of the SBC is also derived, which causes a change curve across the SBC range that enables additional power exchange. Continuous changes here can cause continuous change curves. However, if the electric vehicle leaves or arrives, the SBC changes abruptly.

[0152] The switching power variation curve is now exemplarily shown, namely, the switching power variation curve 238 in the switching power graph 224 of the switching power module 112. Therefore, at the start of charging time t... S The power exchange begins, and the power exchange change curve 238 starts with a negative value. Therefore, power is drawn from the power grid and used to charge the memory. Correspondingly, the total state of charge (SBC) or SBC change curve 236 rises. This can be seen not only in the SBC chart 220 but also in the variant chart 222. At time t1, the power exchange change curve 238 drops to zero, and the SBC change curve 236 correspondingly has a horizontal region.

[0153] At time t2, planning, i.e. forecasting, and then implementing the feeding of power into the power grid is carried out because the demand is particularly high at this time, and the demand can also be derived from this. At night, when only a small amount of power is needed, large generators are shut down so that the power demand can still be generated in the power grid.

[0154] The negative exchange power starting from time t2 can also be seen through the falling edge in the total state of charge change curve 236.

[0155] At time t3, the switching power returns to a positive value, and the total state of charge increases accordingly. At time t4, the switching power is increased again, causing the total state of charge change curve 236 to extend slightly steeper from t4 onwards. Finally, all memories are fully charged in this manner, thereby bringing the total state of charge change curve 236 to the target charging point 230.

[0156] Specifically, the power exchange variation curve 238, as exemplarily shown in the power exchange graph 224, is predetermined as a prediction. The grid operator or grid control unit can then operate based on this prediction, and the grid manager can accordingly set the parameters to match it.

[0157] Although especially in t S Within the range up to t4, there is a possibility of specifically altering the switching power, i.e., changing the truly preset switching power change curve 238. This can occur, for example, when undesirable power demand or power oversupply occurs in the power grid. This situation can also be identified, for example, by changing the grid frequency. Therefore, it is proposed that if the grid frequency of the power grid exceeds a predetermined boundary value, the switching power should be increased, i.e., more power should be extracted from the grid, and / or, if the grid frequency drops below a lower frequency value, the switching power should be decreased, i.e., more switching power should be fed into the power grid.

[0158] Figure 3 A fleet 340 is shown with charging infrastructure 342, which is connected to a power grid 344 via two grid connection points 346 and 347. The power grid has a grid control unit 348, which can control the power grid 344. The grid control unit 348 can also be operated by a grid operator.

[0159] The fleet 340 exemplarily has five electric vehicles 351-355 respectively connected to one of the charging terminals 361-365. The charging terminal can be connected to the power grid 344 via a grid connection point 346 or 347 through a distributor node 366 or 368.

[0160] Therefore, the three electric vehicles 351-353 can feed into or extract power from the power grid 344 via the grid connection point 346, while electric vehicles 354 and 355 feed into or extract power from the power grid 344 via the grid connection point 347.

[0161] A charging control unit 360 is still provided, which can operate each individual charging terminal 361-365 and thus control each memory of the electric vehicles 351-355. Furthermore, as an optional feasibility, the charging control unit can also operate distributor nodes 366 and 368. It is also proposed that the charging control unit 360 can communicate with the grid control unit 348. The charging control unit 360 can provide information to the grid control unit 348 and obtain information from the grid control unit. The charging control unit 360 can also be described as a control unit of the charging infrastructure, or part of a control unit.

[0162] In principle, the information connections between the charging control unit 360 and charging terminals 361-365, distributor nodes 366 and 368, and grid control unit 348 are shown as dashed lines. Power or energy can be transferred via the remaining lines depicted as solid lines.

[0163] The charging control unit 360 is capable of storing information about the characteristics of the charging infrastructure 342 and / or receiving it as current data. Furthermore, the charging control unit is capable of storing information about the characteristics of the electrical memory of the electric vehicles 351-355 and, in particular, obtaining and processing current information about the state of charge and, if necessary, other characteristics of the corresponding memory of the electric vehicles 351-355 via the charging terminals 361-365.

[0164] In this way, the charging control unit 360 can generally control the exchange power between the charging infrastructure 342 and the power grid 344. The exchange power is, in this sense, the sum of the partial exchange power exchanged with the power grid 344 via the grid connection point 346 and the grid connection point 347.

[0165] For illustrative purposes, electric storage devices 371-375 are also indicated in electric vehicle 351-355.

[0166] Figure 4 The structural schematic, based on the functional modules, illustrates how total state of charge prediction can be performed as a time-varying curve of the total state of charge. Therefore, Figure 4 The structure 400 is basically illustrated in two basic stages. Figure 4 The left-hand diagram illustrates the first stage, where, for each individual electric vehicle, predictions are made regarding its arrival time and its individual state of charge. Then, all these individual predictions are combined... Figure 4 The steps illustrated in the right part of the diagram are combined to form the total state of charge prediction.

[0167] Figure 4 Generally, this is based on n electric vehicles, labeled V1-Vn in their respective prediction modules. Each electric vehicle has a prediction module 402 or 402' for arrival time and a prediction module 404 or 404' for a single state of charge. For better overview, these modules are shown only for the first vehicle V1 and the nth vehicle Vn, respectively. In the diagram, the input and output variables for the first vehicle are represented by the number 1, while for the nth vehicle, the variable is represented by the letter 'n'. In this sense, the description of the modules for the first electric vehicle can be applied to the modules for the nth vehicle, and also to all modules not shown in between.

[0168] Therefore, a prediction module 402 is provided for the first vehicle, which obtains prediction information for the arrival time. I t1 is used as input data. The underscore indicates that this also applies. Figure 4 and 5 All other variables, which can be structured as a vector and can contain multiple individual variables. The prediction information for arrival time specifically includes information about the driving plan of the corresponding electric vehicle.

[0169] In addition, supplementary information for arrival time. Z t1 forms another input variable. Similarly, a measured value for arrival time t1m is provided. The measurement of arrival time can also be part of additional or predictive information, but for better overview, this value is additionally named and... Figure 5 and Figure 6 The text also elaborates on its significance.

[0170] Finally, the arrival time prediction module 402 outputs the predicted arrival time t1 as a result. However, because other information can be output along with this, such as in Figure 5 As also explained in the text, therefore Figure 1 Show vector t 1 is used as the output variable of prediction module 402. This is indicated by an underline.

[0171] Similarly, the prediction module 404 for a single state of charge obtains prediction information for that single state of charge. I C1, the prediction information specifically includes information about the driving plan of the corresponding electric vehicle in order to obtain a first estimate of the individual state of charge upon arrival.

[0172] In addition, there is supplementary information for individual states of charge. ZC 1 is used as an input variable, along with a measured value C1m for a single state of charge. The letter m should indicate the presence of a measurement. This also applies to the measured arrival time t1m.

[0173] The prediction module 404 for a single state of charge outputs a prediction for the single state of charge of the electric vehicle in question, which can be described as C1 and is labeled herein. C 1, marked with an underline, indicates that other values ​​can be included, namely, predictions of changes in individual states of charge and probabilistic assessments of individual states of charge.

[0174] The output variables of the prediction module 402 for the arrival time of the electric vehicle and the prediction module 404 for the single state of charge of the electric vehicle can be combined to form a vehicle prediction. In this sense, the two prediction modules 402 and 404 can also be combined into a module that can be called a vehicle prediction module.

[0175] Regardless, the output variables of prediction modules 402 and 404 for each electric vehicle are input into the total prediction module 406. In the total prediction module, the total prediction, i.e., the total state of charge prediction, can be determined as a time-varying curve of the total state of charge. This is output there as SoC(t). The variable t there does not describe the arrival time, but rather describes time as a variable in a common manner and method. In addition to the time-varying curve, an adjusted time-varying curve and a probability assessment can also be output here. The adjusted time-varying curve reflects subsequent changes in the time-varying curve. This should be indicated by an underline.

[0176] In particular, by superimposing over time, all input values ​​can be merged into the total prediction module 406. Specifically, the time-varying curves of individual states of charge can be summed to form the total state of charge. Probabilistic assessments can be performed on the corresponding probabilistic assessments of the total state of charge based on known considerations of probability theory.

[0177] Figure 5 Therefore, illustrative and schematic illustrations are shown. Figure 4 The prediction module 402. However, the structure represents not only the remaining prediction modules for arrival time, but also similarly the prediction module for a single state of charge.

[0178] The input variables of prediction module 402 are therefore related to Figure 4 Combined with the information described above, this is used for predicting arrival times. It1 is symbolically divided into various values ​​contained therein. Here, the first value I1 forms information about the driving plan. This information is contained in the driving plan module 510, from which the driving plan module determines the arrival time t0. In the simplest case, the value t0 is included in the driving plan as the arrival time and can be accepted accordingly. If no other factors need to be considered, this can already be the result of a prediction of the arrival time. However, it is not possible to precisely adhere to the arrival time according to the driving plan.

[0179] Correspondingly, additional prediction information is considered, where Ii represents various additional prediction information. Then, each of these prediction information is input into the calculation module 512, from which the deviation time or time deviation Δti is calculated respectively. The deviation time is added to the reference arrival time t0 at the summing point 514. However, it is also considered to evaluate the prediction information or at least some of the prediction information together, for example, by evaluating a neural network jointly trained on multiple prediction information.

[0180] This is performed on the total deviation time determined from the corresponding information used for the arrival time. The total deviation time can be added entirely to the reference arrival time t0 at the summing point 514. The result is then the predicted provisional arrival time t'. The predicted provisional arrival time t' can already be the arrival time to be output by the prediction module 402. However, here, a further improvement is proposed by means of the adaptation module 516. The adaptation module 516 also... Figure 6 The text describes how the adapter module outputs an additional correction value A, which is added to the temporary arrival time t' at the summing point 526, resulting in the predicted arrival time t, which is then output as t1 for the first electric vehicle.

[0181] Furthermore, not only the driving planning module 510 but also the calculation module 512, representing the calculation module, outputs a random assessment or a probability assessment. This is indicated therein as S0 or Si. This random information can be determined, for example, based on empirical values ​​such as fluctuations recorded over time. For example, the value of the fluctuation in arrival time can be considered for the probability assessment, while another prediction information is not responsible for this. In the other prediction information, the probability assessment can be determined in other ways if necessary. However, in principle, all prediction information and additional information are considered by means of empirical values, not only for arrival time but also for individual states of charge. This is the basic concept and applies to all implementations. The effects of the described information, additional information, or changing information on arrival time and / or individual states of charge and / or total states of charge can also be derived from empirical values ​​and included in the prediction. This is also the basic concept and applies to all implementations.

[0182] However, to give another example, weather forecasts can also be other information, and such weather forecasts typically include probability assessments submitted together. These probability assessments can then be considered in the corresponding calculation module 512. In the randomization module 518, the probability assessments are merged and output as total statistical information or as a total probability assessment S. The randomization module is specifically used for illustration. It is also considered that the probability assessments are linked to their forecast information, i.e., for example, to the deviation time determined from the weather forecast. Therefore, Δti and Si can maintain linked values. Thus, a link with the adaptation module 516 is also considered. Adaptation, for example, can benefit from probability information and can influence the output probability.

[0183] Furthermore, there are additional or subsequent variations in the predicted arrival time t. Therefore, additional information is used for the arrival time. Z t1 can be evaluated separately in the additional computing module 520, essentially as in computing module 512. The additional computing module 520 also represents multiple such additional computing modules, i.e., additional computing modules for additional information, respectively. Correspondingly, in Figure 5 The text also indicates additional information. Z The total t1 is decomposed into multiple individual additional information Zi.

[0184] In another design, the additional information can be processed directly together with or in the driving planning module 510 and the calculation module 512 and thus provided in the summing section 514. The adaptation module 516, represented here, can preferably be configured as a process only after processing the additional information. At least one variant proposes that weights be correlated with the additional information in the adaptation. For example, if t1m deviates significantly from t0, due to, for example, congestion known from one of the additional information pieces, or as input to the additional information, it is proposed that such deviation not be overly weighted. The adaptation should not be overly affected by such isolated events, and thus it is proposed that the adaptation consider the additional information together, at least some of it, i.e., operate according to the additional information. This is generally suggested not only for the illustrated implementation.

[0185] The result of the additional calculation module 520 is therefore an additional deviation time or time deviation ΔtZi, i.e., an additional deviation time or time deviation for each additional piece of information. The time deviation ΔtZi is summed at the summing point 522 to form the predicted arrival time t and to obtain the modified predicted arrival time ta.

[0186] For the additional information and the changes derived therefrom, namely the deviation time ΔtZi, a probability assessment is also performed, and the probability assessments can be aggregated in the additional stochastic module 524 into an additional probability assessment Sz. Here, the probability assessments can also be alternatively linked to their respective prediction information, so that the additional stochastic module 524 is also used in this context.

[0187] Finally, the calculated values—namely, the additional probability assessment Sz, the predicted change in arrival time ta, the predicted arrival time t, and the probability assessment S—are combined to form the output vector. t 1. And output. Then, the result can be as follows: Figure 4 As described in the text, it is transferred to the overall prediction module 406.

[0188] To improve the predicted temporary arrival time t', an adaptation module 516 is provided. Figure 6 The diagram is illustrative and is only shown as an example. Therefore, the predicted provisional arrival time t1' forms the input to the adaptation module 516. Following the sample-and-hold phase 640, the correction value A is summed to the predicted provisional arrival time in the summing section 630. The result is a corrected, i.e., the final predicted arrival time, which is used as an intermediate variable in the adaptation module. However, the same adaptation is... Figure 5 The addition is performed in part 514. For this purpose, the adapter module 516 outputs an additional correction value A.

[0189] To perform the adaptation proposal, the corrected arrival time t1 is compared with the actual measured arrival time t1m. The measured arrival time t1m is measured, particularly in earlier cycles, especially the most recent cycle, and / or in earlier rounds, particularly on the previous day or two days prior.

[0190] The comparison is performed, resulting in a difference in the summation portion 632. The result is therefore a difference in the sense of the adjustment error e. This difference is multiplied by a weighting factor according to the weighting module 634. The weighting module has a factor, i.e., the weighting factor, that should be between zero and 1. This factor should not be exactly zero, as it is logically no longer considered. However, the weighting factor may take the value 1. Thus, ideally, the fit is fully performed in one step.

[0191] However, since fluctuations are assumed in principle, lower values ​​are meaningful, such as, for example, 0.1. Thus, the correction factor A is essentially led to the final correction value via a first-order delay stage. In any case, the result of the weighting module 634 is a weighted adjustment error e'. This weighted adjustment error is summed at the summing point 636 to the correction value of the previous round. The result is therefore a new correction value A', which is provided via the holding stage 638 to form the current correction factor A. The correction factor A can therefore also be described as A(k) and the new correction factor A' as A(k+1).

[0192] In any case, it is possible to be as Figure 6 The adaptation is performed as described in the text. Figure 6 The structure here is an example of adaptation using an additional correction value, namely correction value A. If an ideal state occurs, i.e., the predicted arrival time corresponds to the measured arrival time and the adjustment error e is therefore zero, then the correction value A also retains its value. Therefore, the adaptation is performed such that the measurement of arrival time t1m is compared with the corresponding arrival time of past predictions. In other words, for the arrival time measured the previous day, there is also a prediction for that previous day, and these two times are compared. The correction value A determined here can still be used for the current prediction, for which, due to the principle, no measurement exists yet.

[0193] Correspondingly, in Figure 6 The sample-and-hold loop 640 is plotted to show that, for the fit, the current provisional predicted value t1' is not used, but rather the previous value is used. However, this should be understood schematically and also takes into account that an earlier value, i.e., the one corresponding to the measured value t1m, may also be taken. Correspondingly, for illustration, the input variable of the sample-and-hold loop 640 is shown as t1'(k+1) and the output value is plotted as t1'(k). The hold time in the sample-and-hold loop 640 can be one day or more days, particularly two days.

[0194] The model parameters of the calculation module 512 and the additional calculation module 520, and even the driving planning module 510, can be adapted in a similar manner. However, in this model adaptation, output variables or states are not adapted; instead, parameters, i.e., special factors, are adapted.

[0195] This adaptation can be performed such that, in the corresponding model to be adapted, one parameter or multiple parameters are changed, either identically or according to another relationship, until the output corresponds to a comparison measurement. The resulting parameters can then be compared with the parameters that existed before the change. Differences can be generated separately, and these differences can be used to... Figure 6The adjustment error e at the output of the summing section 632 is handled in the same way. Therefore, the deviation is multiplied by a weighting factor between zero and 1, and the result is added to an earlier correction value that could be the correction factor. This summation is given via a holding link so that it can then be added again at the summing section 636 for the next variation curve. The process is repeated in the next variation curve, starting again with a change in the parameter that thus forms the model parameter.

[0196] Alternatively, adaptation for prediction can also be performed, as is known from the adjustment techniques used for adapting regulators. Here, the regulator is thus replaced, and the prediction model is adapted. Self-tuning methods are particularly considered here, where the system and its changes are identified based on observations of the input and output variables. The model can then be adjusted accordingly based on the changes thus identified. In this sense, the prediction information and additional information considered separately are the input variables and the predicted arrival time, or, in the case of a single state of charge prediction, the predicted single state of charge; the output variables and the corresponding model used to determine the corresponding arrival time or the corresponding single state of charge can be considered as the system in this sense.

[0197] The objective of this invention is not only to predict the time as well as as well as the state of charge, including the probability distribution, as well as as well as possible.

[0198] It has been recognized that alternative sites can also forecast charging energy demand and charging periods, specifically with variable start times and fixed end times.

[0199] It has been recognized that such forecasting may be necessary when mobile storage is aggregated into a common storage capacity that can provide flexibility.

[0200] In particular, it relates to the provision of information for aggregation into highly available memory consortia.

[0201] The present invention also aims to implement and utilize probability-dependent state of charge (SoC) forecasting and arrival forecasting for electric vehicle applications in line operation, particularly for buses and logistics. Applications in car-sharing or bicycle-sharing are also considered.

[0202] This concept focuses on predicting the arrival time (including probability distribution) and state of charge (including probability distribution) of vehicles at charging points. During aggregation (see attached figures, specifically...) Figure 2 and Figure 3 The aforementioned forecast is required. This allows for the use of a very large share of battery capacity, ensuring high availability for global and local system services and / or power arbitrage operations. The forecast is created in advance, particularly one day in advance, i.e., before the initial charging energy procurement, and can be dynamically updated.

Claims

1. A method for planning power exchange between charging infrastructure (342) and power grid (344), wherein - The charging infrastructure (342) has multiple charging terminals (361-365) for connecting electric vehicles (351-355) and charging the electric vehicles, so that the electric vehicles (351-355) can exchange electrical power with the power grid (344) via the charging terminals (361-365). - Each electric vehicle (351-355) has an electrical memory (371-375) with a variable single state of charge for receiving and outputting electrical power. - All power storage devices connected to the charging infrastructure (342) form the total storage of the charging infrastructure (342), characterized by a total storage capacity and a total state of charge (236), wherein - The total memory capacity is variable, and - The total state of charge (236) is variable, wherein - Create a prediction of the arrival time (t1, tn) of the electric vehicles (351-355) to their charging terminals (361-365), and - Based on the predicted arrival times (t1, tn), a total state of charge prediction is created as a prediction of the total state of charge (236) for the predicted time period, wherein - Create the total state of charge prediction as a time-varying curve of the total state of charge (236). The total state of charge prediction is created based on the driving plan of the electric vehicle, and the driving plan includes the planned driving time and the planned driving route.

2. The method according to claim 1, characterized in that, The total state-of-charge prediction is created at least one day before the prediction time period (T B ) begins.

3. The method according to claim 1 or 2, characterized in that, - Create a single state of charge prediction for each individual state of charge, and - Additionally, the total state of charge prediction is created based on the individual state of charge prediction.

4. The method according to claim 1 or 2, characterized in that, Based on at least one prediction information for arrival time ( I t1) Create a prediction of the arrival time, the prediction information being selected from the following list: - Information regarding the driving plan of the corresponding electric vehicles (351-355), - Current traffic infrastructure information for the corresponding electric vehicle-related areas. - For the corresponding electric vehicle's arrival time stored over the past several days. - Weather forecast, - Information regarding events affecting traffic volume in the corresponding areas associated with the electric vehicles, and - Information regarding the driver of the respective electric vehicle (351-355) and / or regarding the driving behavior of the respective electric vehicle (351-355).

5. The method according to claim 1 or 2, characterized in that, A single state of charge prediction is created based on at least one prediction information (IC1) for a single state of charge, the prediction information being selected from the following list: - Information regarding the driving plan of the corresponding electric vehicles (351-355), - Information regarding the characteristics of the electric vehicle. - Information regarding the individual state of charge stored for the corresponding electric vehicle upon arrival at the charging terminal, based on detections over the past several days. - Information regarding the preset individual state of charge of the electric vehicle (351-355) at the start of driving when the electric vehicle is disconnected from its charging terminal. - Current traffic infrastructure information for the corresponding electric vehicle-related areas. - Weather forecast, - Information regarding events affecting traffic volume in the corresponding areas associated with the electric vehicles. - Information regarding the drivers of the respective electric vehicles (351-355) and / or regarding the driving behavior of the respective electric vehicles (351-355), - Information regarding the status of the electric vehicles (351-355), and - Information on the model's inaccuracy or error in detecting a single state of charge.

6. The method according to claim 3, characterized in that, - After creating the total state of charge prediction, the total state of charge prediction is changed based on at least one change information.

7. The method according to claim 1 or 2, characterized in that, - After creating the arrival time prediction, the arrival time prediction is changed based on at least one additional piece of information (Zt1) selected from the following list: - Information regarding changes to the aforementioned driving plan. - Information about detected traffic conditions and / or short-term forecasts of traffic conditions. - Information regarding the current weather following the creation of the total state of charge forecast, or short-term weather forecasts created for the forecast space. - Information regarding the deviation between the electric vehicle's current location and the location planned according to the driving plan. - Information regarding the load or average number of people occupied by the electric vehicle.

8. The method according to claim 1 or 2, characterized in that, - After creating the single state of charge prediction, the single state of charge prediction is modified according to at least one additional piece of information (ZC1) for the single state of charge, the additional information being selected from the following list: - Information about detected traffic conditions and / or short-term forecasts of traffic conditions. - Information regarding the current weather following the creation of the total state of charge forecast, or short-term weather forecasts created for the forecast space. - Information regarding the deviation between the electric vehicle's current location and the location planned according to the driving plan. - Information regarding the load or average number of people occupying the electric vehicle. - Information about the individual state of charge of the electric vehicle.

9. The method according to claim 1 or 2, characterized in that, With the help of - A predictive model for arrival time. - Predictive models for a single state of charge, or - Prediction model for the total state of charge (236) - Create and / or change - The predicted arrival time, - The single state of charge prediction, and / or - The total state of charge prediction.

10. The method according to claim 1 or 2, characterized in that, - Creating and / or modifying the arrival time prediction using a prediction model for arrival time, the prediction model for arrival time - Consider at least one or more predictions for the arrival time to create a prediction of the arrival time as model input, and - Consider at least one additional piece of information (Zt1) or one of the additional pieces of information used for arrival time to modify the prediction of the arrival time as an additional input variable, and / or - Creating and / or modifying the single state of charge prediction using a prediction model for a single state of charge, the prediction model for a single state of charge... - Consider at least one prediction piece of information for a single state of charge, or one prediction piece of information for a single state of charge, to create a single state of charge prediction as model input, and - Consider at least one additional piece of information (ZC1) for a single state of charge, or one of the additional pieces of information for a single state of charge, to modify the state of charge prediction as an additional input variable, and / or - Create and / or modify the total state of charge prediction using a prediction model for the total state of charge (236), the prediction model for the total state of charge - For each electric vehicle under consideration - Consider at least one or more predictions for arrival time and / or for a single state of charge as model input, and - Consider at least one additional piece of information (Zt1) or one of the additional pieces of information for arrival time and / or at least one additional piece of information (ZC1) or one of the additional pieces of information for a single state of charge as additional input variables, and / or - The prediction models for the total state of charge (236) include, respectively, for the electric vehicles under consideration. - Predictive models for arrival time, and - Predictive models for individual states of charge. Combine them into a new model and / or simulate them as a combined model.

11. The method according to claim 1 or 2, characterized in that, The creation of the arrival time prediction is adapted by comparing the predicted arrival time for the predicted time period with the actual arrival time occurring within the predicted time period, and adjusting the future creation of the prediction based on the comparison, wherein the adaptation is repeated, i.e., the comparison is repeated and the adjustment is repeated in relation to the comparison, and / or The creation of the individual state of charge prediction is adapted by comparing the individual state of charge prediction for the prediction time period with the individual state of charge that actually occurs in the prediction time period, and adjusting the future creation of the prediction based on the comparison, wherein the adaptation is repeated, i.e. the comparison is repeated and the adjustment is repeated in relation to the comparison.

12. The method according to claim 1 or 2, characterized in that, - A prediction model for arrival time used to create and / or modify the prediction of the arrival time, or the prediction model for arrival time implemented as an adapted model, the adapted model adjusting one or more model parameters based on a comparison between the predicted arrival time and the actual arrival time, and / or - A prediction model for a single state of charge, or a prediction model for a single state of charge implemented as an adapted model, for creating and / or modifying the prediction of the single state of charge, wherein the adapted model adjusts one or more model parameters based on a comparison between the predicted single state of charge and the single state of charge that has occurred thereto.

13. The method according to claim 3, characterized in that, - In creation - The total state of charge prediction, - The predicted arrival time, and / or - When predicting a single state of charge, respectively - Execution probability assessment (S).

14. The method according to claim 1 or 2, characterized in that, - For each electric vehicle (351-355) - Together with probability assessment, a prediction of the arrival time is created, and - Create a single state of charge prediction together with probability assessment, and - Combined into vehicle prediction, and - Determine the total state of charge prediction based on all vehicle predictions along with probability assessments, and / or - The total state of charge prediction is determined for a predetermined probability value based on all vehicle predictions along with probability assessments.

15. The method according to claim 1 or 2, characterized in that, - Perform power switching planning so that - To provide switching power for exchange with the power grid (344), wherein - Determine the switching power based on the total state of charge prediction.

16. The method according to claim 3, characterized in that, - For each individual state of charge, a single state of charge prediction is created as a time-varying curve of that individual state of charge.

17. The method according to claim 5, characterized in that, The list contains information about the characteristics of the electric vehicle's memory.

18. The method according to claim 5, characterized in that, The list contains information about the status of the memory of the electric vehicles (351-355).

19. The method according to claim 6, characterized in that, - Based on the predicted changes in the arrival time, and / or - Changes predicted based on the individual state of charge This is used to change the total state of charge prediction.

20. The method according to claim 7, characterized in that, The list has: - Information regarding the manual alteration of the driving plan.

21. The method according to claim 8, characterized in that, The list contains information about a single state of charge of the electric vehicle along with the associated time and / or associated location.

22. The method according to claim 10, characterized in that, - Consider at least one additional piece of information (Zt1) or one of the additional pieces of information used for arrival time to change the prediction of the arrival time as a confounding variable.

23. The method according to claim 10, characterized in that, - Consider at least one additional piece of information (ZC1) for a single state of charge or one additional piece of information for a single state of charge to change the state of charge prediction as a disturbance variable.

24. The method according to claim 10, characterized in that, - Consider at least one additional piece of information (Zt1) or one additional piece of information for arrival time and / or at least one additional piece of information (ZC1) or one additional piece of information for a single state of charge as interference variables.

25. The method according to claim 11, characterized in that, The adaptation is repeated daily, that is, the comparison is repeated daily and the adjustment is repeated in relation to the comparison.

26. The method according to claim 13, characterized in that, - In creation - The total state of charge prediction, - The predicted arrival time, and / or - When predicting a single state of charge, respectively - Assignment probability distribution.

27. The method according to claim 14, characterized in that, - For each electric vehicle (351-355) - Create the prediction of the arrival time together with the probability distribution.

28. The method according to claim 14, characterized in that, - For each electric vehicle (351-355) - Create a single state of charge prediction together with the probability distribution.

29. The method according to claim 14, characterized in that, - The total state of charge prediction, along with the probability distribution, is determined based on all vehicle predictions and probability assessments.

30. The method according to claim 14, characterized in that, - The total state of charge prediction is determined based on all vehicle predictions along with probability assessments, with a probability of 90%, 95%, 98%, or 99%.

31. The method according to claim 15, characterized in that, - In order to exchange with the power grid (344), an exchange power variation curve or exchange power band is provided.

32. The method of claim 31, wherein the switching power variation curve or the switching power band is determined based on the total state of charge prediction.

33. A charging infrastructure (342) configured for planning power exchange between the charging infrastructure (342) and the power grid (344), wherein - The charging infrastructure (342) has multiple charging terminals (361-365) for connecting electric vehicles (351-355) and charging the electric vehicles, so that the electric vehicles (351-355) can exchange electrical power with the power grid (344) via the charging terminals (361-365). - Each electric vehicle (351-355) has an electrical memory (371-375) with a variable single state of charge for receiving and outputting electrical power. - All power storage devices connected to the charging infrastructure (342) form the total storage of the charging infrastructure (342), which is characterized by total storage capacity and total state of charge (236), wherein - The total memory capacity is variable, and - The total state of charge (236) is variable, wherein - The charging infrastructure (342) is configured for, - Create a prediction of the arrival time of the electric vehicles (351-355) to their charging terminals (361-365), and - Based on the predicted arrival time, a total state of charge prediction is created as a prediction of the total state of charge (236) for the predicted time period, wherein - The charging infrastructure is configured to create the total state of charge prediction as a time-varying curve of the total state of charge (236). The charging infrastructure (342) is configured to create the total state of charge prediction based on the driving plan of the electric vehicle, the driving plan having a planned driving time and a planned driving route.

34. The charging infrastructure (342) according to claim 33, characterized in that, The charging infrastructure (342) is configured to perform the method according to any one of claims 1 to 32.

35. The charging infrastructure (342) according to claim 33, characterized in that, The charging infrastructure (342) has a control unit.

36. The charging infrastructure (342) according to claim 35, characterized in that, The control unit is configured to perform the method according to any one of claims 1 to 32.

Images