Methods for predicting vehicle energy demand

Abstract

The invention relates to a method (1) for predicting the energy demand of a vehicle. It is important that an artificial function model (2) is generated and / or trained with the aid of an artificial intelligence (3) in the vehicle model, which is referred to as KI, and in a subsequent step, the energy demand of the vehicle is predicted by providing at least one data recorder arranged on a data bus of the vehicle for collecting bus data (5) exchanged via the data bus, wherein the bus data (5) are collected using at least one data recorder, wherein an artificial function model (2) previously generated and / or trained according to the method (1) is provided on the vehicle and the collected vehicle bus data (5) is provided to the artificial function model, wherein the artificial function model (2) subsequently provides predicted total power demand data (7), which represent the predicted total power demand of the vehicle or the predicted power demand of a single component of the vehicle.

Description

Technical Field

[0001] The invention relates to a method for predicting the energy demand of a vehicle according to the preamble of claim 1 . Background Art

[0002] The above method is known from US 2021 / 116907 A1. Another method for predicting energy demand is known from DE 20 2018 106 059 U1. Specifically, the method teaches a system for detecting the power demand of bus users of a bus system in the field of building technology.

[0003] Reference is also made to the documents DE 10 2013 109348 A1, DE 20 2015 106567 U1, US 2016 / 358475 A1, CN 203 405 557 U, CN 101 519 073 A and WO 2019 / 017991 A1. Summary of the invention

[0004] It is an object of the present invention to provide or indicate an improved or at least different method of predicting energy demand.

[0005] In the invention, this object is achieved in particular by the subject matter of independent claim 1. Advantageous embodiments are the subject matter of the dependent claims and the description.

[0006] The core idea of ​​the invention is to collect power demand data and bus data on a car model of a certain model, for example, used in the process of car development, and to provide these data to an artificial intelligence, so that the artificial intelligence can generate and / or train a so-called artificial functional model. The artificial functional model generated and / or trained in this way can then be prepared at the factory and then integrated into a marketable car of this model or a similar model or implemented in a controller of the car. According to the invention, in order to be able to provide the car user with a forecast of the car's energy demand, it is only necessary to provide the bus data collected on the car to the artificial functional model, which converts these data into a corresponding forecast of the car's energy demand.

[0007] To this end, the present invention proposes a method for predicting the energy demand of a vehicle, within the scope of which the following steps are performed:

[0008] 1) Generate and / or train an artificial function model with the help of artificial intelligence (hereinafter referred to as KI) of a car model, which is implemented by configuring a measuring sensor and at least one data logger for the car model when providing it, wherein the measuring sensor is arranged on the current fuse or electrical equipment of the car model for detecting the power demand of each current fuse of the car model, and the data logger is arranged on the data bus of the car model for collecting bus data exchanged through the data bus. Then, the power demand at each current fuse can be detected by the measuring sensor, and power demand data representing / reflecting the power demand can be provided, wherein the bus data is collected and provided with the help of at least one data logger. Then, the acquired bus data and power demand data are stored in a memory, in particular in a memory of the car model and / or the car, and synchronized. Then, the stored synchronized bus data and power demand data can be provided as so-called training data to the above-mentioned KI of the car model, wherein the KI generates and / or trains the artificial function model based on the provided bus data and power demand data. The artificial function model here or thereby represents / reflects the correlation between the bus data and the power demand data. In the subsequent step 2), the energy demand of the vehicle is predicted, which is implemented by configuring at least one data recorder for the vehicle, the data recorder is arranged on the data bus of the vehicle, and is used to collect bus data exchanged through the data bus of the vehicle, and the bus data of the vehicle is collected and provided by means of at least one data recorder of the vehicle, wherein an artificial function model generated and / or trained according to step 1) is provided on the vehicle, and the collected vehicle bus data is provided to the artificial function model. In this way, the artificial function model can provide predicted total power demand data (hereinafter referred to as PGS data), which represents the predicted total power demand of the vehicle or the predicted power demand of a single / individual current fuse and / or electrical device of the vehicle.

[0009] In other words, the method of the present invention proposes that in step 1), an artificial function model is first generated and / or trained with the aid of KI, wherein data of a vehicle model of a predetermined vehicle model are utilized. Therefore, step 1) of the method of the present invention can preferably be performed on site by the manufacturer during the vehicle model development process. According to step 2), the artificial function model generated and / or trained by the manufacturer is provided on a vehicle of this model or a similar model, for example wirelessly via radio, or wired in a workshop during vehicle service, or further, for example, wired or by radio during mass equipment / standardized assembly of the vehicle during production. Within the scope of step 2), the total power demand of the vehicle can then be predicted with the aid of the bus data collected on the corresponding vehicle and the provided artificial function model, by converting these bus data into corresponding predictions about the energy demand of the vehicle by the artificial function model. Advantageously, the power demand of individual systems and / or components of the vehicle can be predicted in particular.

[0010] The present invention understands the term "energy demand assessment" as an approximation or estimation of the energy demand of the vehicle. Furthermore, the present invention understands the term "providing" as meaning, in particular, integrating and / or implementing.

[0011] According to the invention, it is also proposed that during the generation and / or training of the KI according to step 1) and / or during the prediction of the energy demand of the vehicle according to step 2), the acquired bus data is used to improve the artificial function model. This allows so-called deep reinforcement learning (from the English “Deep Reinforcement Learning”), i.e. the artificial function model installed in the vehicle model and / or the vehicle to be continuously trained and / or adjusted. This makes it possible, for example, for the artificial function model to react to the driving style of different users of the vehicle model and / or the vehicle, so that the predicted energy demand of the vehicle model and / or the vehicle can be individually adjusted, calculated, estimated or approximated.

[0012] In order to be able to make the above-mentioned improvements to the artificial function model, it is also proposed according to the present invention that the acquired bus data has total power demand data, which will be referred to as GS data below for simplicity. This data represents the actual total power demand of the car. Then the above-mentioned PGS data is provided with the help of the artificial function model, which represents the predicted total power demand of the car, wherein the GS data is compared with the PGS data, for example, in a comparator of the car or the artificial function model. Here, the artificial function model is configured to measure the deviation between the GS data and the PGS data, that is, the deviation between the actual total power demand of the car and the predicted total power demand of the car. In addition, the artificial function model is configured to influence the deviation of this determination, that is, to minimize it. In order to achieve the desired improvement to the artificial function model, when the GS data deviates from the PGS data, the artificial function model minimizes the deviation between the GS data and the PGS data. This can improve the learning effect of KI. Its advantage is mainly that, for example, under the condition of a given remaining capacity of the car energy storage device, the remaining cruising range of the car can be better predicted. In addition, KI can more accurately estimate the energy requirements of various components of the car based on previous driving behavior and the use of various systems of the car during driving, so as to relatively accurately predict the remaining range of the car. In addition, specific energy-saving suggestions can be provided to car users, such as deactivating certain electrical devices of the car to extend the remaining range of the car. It is also conceivable that other electrical devices of the car can be automatically controlled or adjusted based on the current prediction of KI, for example, by the controller of the car.

[0013] Further advantageously, the PGS data are provided and / or used at a control unit of the vehicle. Advantageously, the KI and / or the artificial function model can also be provided to the control unit of the vehicle and / or integrated there. This makes it possible to provide the method of the invention to the vehicle in a convenient and cost-effective manner.

[0014] The artificial functional model is preferably used in a single car model and / or in a development environment. However, as mentioned above, the artificial functional model can also be applied to cars of similar models or simply another model. That is, the artificial functional model can be used across models to provide an estimate of the energy demand of a single car. This means that the artificial functional model proposed by the present invention has a relatively high cost-effectiveness in use.

[0015] Further advantageously, the PGS data of the vehicle are provided and / or used at the back end, in particular at a fixed station remote from the vehicle, for example on a server. Thus, the energy requirements of a group of vehicles of a certain model can be detected and monitored externally.

[0016] Advantageously, the artificial function model can be implemented by an artificial neural network or a regression model or an alternative machine learning model. Thus, a variety of advantageous artificial function models are provided. Of course, other categories, types and methods of artificial function models and / or artificial neural networks can be advantageously included, and preferably all categories, types and methods of artificial function models and / or artificial neural networks can be included.

[0017] In summary, the invention advantageously relates to a method for predicting the energy demand of a vehicle. It is important that an artificial function model is generated and / or trained with the aid of an artificial intelligence called KI of a vehicle model and in a subsequent step the energy demand of the vehicle is predicted, which is implemented by providing at least one data recorder, which is arranged on a data bus of the vehicle for collecting bus data exchanged via the data bus, wherein the bus data is collected with the aid of at least one data recorder, an artificial function model previously generated and / or trained according to the method is provided on the vehicle and provided with the collected vehicle bus data, and subsequently the artificial function model provides predicted total power demand data, which represents the predicted total power demand of the vehicle or the predicted power demand of a single component of the vehicle.

[0018] Further important features and advantages of the invention emerge from the dependent claims, the drawings and the associated figure description based on the drawings.

[0019] It is to be understood that the features mentioned above and those yet to be explained below can be used not only in the respectively specified combination but also in other combinations or alone, without departing from the scope of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Preferred exemplary embodiments of the invention are shown in the drawings and are explained in more detail in the following description, wherein the same reference numerals refer to identical or similar or functionally identical components.

[0021] The accompanying drawings schematically show:

[0022] Figure 1 A flowchart showing generation and / or training of an artificial function model according to a preferred embodiment of the method of the present invention is shown.

[0023] Figure 2 Another flow chart showing vehicle energy demand prediction according to a preferred embodiment of the method of the present invention is shown, and

[0024] Figure 3 Another flow chart showing a preferred embodiment of the method according to the present invention using an artificial function model to perform so-called deep reinforcement learning (from the English “Deep Reinforcement Learning”). DETAILED DESCRIPTION

[0025] Figure 1 A flowchart for generating and / or training an artificial function model 2 according to a preferred embodiment of a method 1 according to the invention, the method being used to predict the energy demand of a vehicle model not shown in the figure. The illustrated method 1 comprises step 1), according to which an artificial function model 2 represented by a box is generated and / or trained with the aid of an artificial intelligence 3 (hereinafter referred to as KI) of the vehicle model. To this end, when a vehicle model is provided, it is equipped with measuring sensors and at least one data logger, the measuring sensors being arranged on the current fuses of the vehicle model for detecting the power demand of each current fuse of the vehicle model and providing power demand data 4 representing the detected power demand, and the data logger being arranged on the data bus of the vehicle model for collecting bus data 5 exchanged via the data bus. Within the scope of method 1, the power demand of each current fuse is detected with a measuring sensor, and further power demand data 4 representing these detected power demands are provided. Within the scope of method 1, bus data 5 are also collected and provided with the aid of at least one data logger. The bus data 5 and the power demand data 4 are then stored in a memory not shown in the figure and time-synchronized, in particular in such a way that time-related power demand data 4 are assigned to each bus data 5. Figure 1 In the figure, the power demand data 4 and the bus data 5 are both represented by a simple box. In order to generate and / or train the artificial function model 2, the stored time-synchronized bus data 5 and power demand data 4 are provided as so-called training data to the KI 3 of the vehicle model and processed, so that the KI 3 can finally generate and / or train the artificial function model 2 based on the provided bus data 5 and power demand data 4.

[0026] Figure 2 Another flow chart shows the energy demand forecasting of a marketable vehicle (not shown in the figure) according to a preferred embodiment of the method 1 of the invention. The illustrated method 1 advantageously comprises a step 2), according to which the energy demand of the vehicle can be predicted by first providing the vehicle with at least one data recorder, which is arranged on the data bus of the vehicle and is used to collect bus data 5 exchanged via the data bus of the vehicle. Within the scope of step 2) of the method 1, during the actual operation of the vehicle, the bus data 5 of the vehicle are collected and provided by means of at least one data recorder of the vehicle. The collected bus data 5 can now be provided to the artificial function model 2 (which is also provided on the vehicle) that was pre-generated and / or trained within the scope of step 1) of the method 1, so that it can then acquire and provide the predicted total power demand data 7, which are referred to as PGS data in the following. These PGS data 7 represent the predicted total power demand of the vehicle or the predicted power demand of a single current fuse of the vehicle.

[0027] Figure 3Another flow chart shows the so-called deep reinforcement learning (from the English “Deep Reinforcement Learning”) performed by the artificial function model 2 according to a preferred embodiment of the method 1 of the present invention. It is provided that the bus data 5 acquired according to step 1) and / or step 2) of the method 1 of the present invention are used to train or at least adjust or optimize the artificial function model 2 previously provided on the vehicle model and / or vehicle according to step 1) and / or step 2) of the method 1 of the present invention. For this purpose, it is provided, for example, that the acquired bus data 5 have or include total power demand data 6, which are called GS data, which represent the actual total power demand of the vehicle model or the vehicle, and that PGS data 7 are provided with the help of the artificial function model 2, which, as mentioned above, represent the predicted total power demand of the vehicle model and / or the vehicle. The GS data 6 are then compared with the PGS data 7 in a comparator 8. For example, the artificial function model 2 is provided to minimize the deviation 9 between the GS data 6 and the PGS data 7, in particular in a way that the determined deviation 9 is fed back from the comparator 8 to the artificial function model 2. Specifically, if the GS data 6 actually deviates from the PGS data 7, the artificial function model 2 minimizes the measured deviation 9 between the GS data 6 and the PGS data 7, thereby achieving deep reinforcement learning.

Claims

1. A method for predicting the energy demand of a vehicle (1), The following steps are involved: 1) Generate and / or train an artificial function model (2) with the help of artificial intelligence (3) of the vehicle model, The implementation is: - providing the vehicle model with a measuring sensor and at least one data logger, the measuring sensor being arranged on a current fuse of the vehicle model for detecting a power demand of the current fuse and providing power demand data (4) representing the detected power demand, the data logger being arranged on a data bus of the vehicle model for collecting bus data (5) exchanged via the data bus, - detecting the power demand at the current fuse or at the electrical consumer by means of the measuring sensor, and providing power demand data (4) representing the detected power demand, - the bus data (5) are acquired and made available by means of the at least one data logger, - then time-synchronizing the bus data (5) and the power demand data (4) and storing them in a memory as training data, - said training data is then provided to said artificial intelligence of said car model (3), - the artificial intelligence (3) generates and / or trains an artificial function model (2) based on the provided bus data (5) and power demand data (4), the artificial function model representing the correlation between the bus data (5) and the power demand data (4), and 2) Predict the energy demand of the car, which is achieved by: - the motor vehicle is provided with at least one data logger, which is arranged on a data bus of the motor vehicle and is used to record bus data (5) exchanged via the data bus of the motor vehicle, - using at least one data logger of the motor vehicle to record and make available bus data (5) of the motor vehicle being operated, - providing the artificial function model (2) generated and / or trained according to step 1) on the vehicle, and providing the artificial function model with the collected bus data (5), - Then, the artificial function model (2) obtains and provides the predicted total power demand data (7), The predicted total power demand data represents the predicted total power demand of the vehicle or the predicted power demand of a single current fuse and / or a power-consuming device of the vehicle, - using the acquired bus data (5) to improve the artificial function model (2) during the generation and / or training of the artificial intelligence (3) according to step 1) and / or during the prediction of the energy demand of the vehicle according to step 2), It is characterized in that - the acquired bus data (5) have total power demand data (6) representing the actual total power demand of the vehicle, - wherein the artificial function model (2) provides forecast total power demand data (7), - wherein the total power demand data (6) is compared with the predicted total power demand data (7), - wherein the artificial function model (2) is configured to minimize the deviation (9) between the total power demand data (6) and the predicted total power demand data (7), - wherein, when the total power demand data (6) deviates from the predicted total power demand data (7), the artificial function model (2) minimizes the deviation (9) between the total power demand data (6) and the predicted total power demand data (7).

2. The method (1) according to claim 1, It is characterized in that - providing and / or using said predicted total power demand data on a controller of said vehicle and / or in a development environment.

3. The method (1) according to claim 1 or 2, It is characterized in that -Provide and / or use the vehicle's predicted total electricity demand data on the back end.

4. The method (1) according to claim 3, It is characterized in that -Providing and / or using the vehicle's predicted total electricity demand data at a fixed site remote from the vehicle.

5. The method (1) according to claim 4, It is characterized in that -The fixed site far away from the car is a server.

6. The method (1) according to claim 1 or 2, It is characterized in that The artificial function model (2) is implemented by an artificial neural network or a regression model or an alternative machine learning model.

Images