CN113492641A

CN113492641A - Method for controlling an air conditioning device of a motor vehicle, air conditioning device and motor vehicle

Info

Publication number: CN113492641A
Application number: CN202110279666.5A
Authority: CN
Inventors: P·克里斯特; G·格斯滕博兰特; M·许布勒; J·伊普斯; M·雅各布; A·舒尔茨; M·施蒂克斯
Original assignee: Bayerische Motoren Werke AG
Current assignee: Bayerische Motoren Werke AG
Priority date: 2020-04-03
Filing date: 2021-03-16
Publication date: 2021-10-12
Also published as: DE102020109299B4; DE102020109299A1

Abstract

The invention relates to a method for controlling an air conditioning device of a motor vehicle, a corresponding air conditioning device and a motor vehicle equipped with the air conditioning device. In the method, a machine learning device is provided which is trained for estimating the actual air temperature in the passenger space of the motor vehicle. Data relating to the air conditioning of the passenger space are also acquired, which data give at least one temperature measured by means of a temperature sensor. The data is processed to determine an estimated temperature for the occupant space. The air conditioning device is then automatically controlled according to the estimated temperature, which is regarded as the respective current actual value.

Description

Method for controlling an air conditioning device of a motor vehicle, air conditioning device and motor vehicle

Technical Field

The invention relates to a method for controlling an air conditioning device of a motor vehicle, a corresponding air conditioning device and a motor vehicle equipped with the air conditioning device.

Background

The tempering of the interior of the motor vehicle, which is optimally matched to the respective user or passenger wishes, can contribute significantly to passenger comfort and is therefore desirable. However, a problem here is that a temperature sensor, which is provided, for example, in the center console of the motor vehicle, cannot always reliably measure the actual air temperature at the location of one or more occupants. But the temperature measured by the temperature sensor may be significantly affected, for example, due to heat transfer of the surrounding components. Thus, for example, a center console, which is typically dark in color, may be heated by insolation to a temperature that may be much higher than the air temperature at the occupant location. It is not practical, however, to provide an additional temperature sensor in the air volume of the interior space itself. Furthermore, the air temperature or the temperature sensed by the occupant may be affected by other effects, such as air flow, open side windows, etc.

In order to provide greater comfort, DE10344795a1 describes a vehicle air conditioning system comprising a front air conditioning device for blowing air-conditioned air to the front space of the passenger compartment and a rear air conditioning device for blowing air-conditioned air to the rear space of the passenger compartment. Based on the load value, the respective blow-out conditions of the air-conditioned air are determined by the respective control device. Here, the blowing conditions are determined by means of a nonlinear model representing the relationship between the load factor on which the load value is based and the target blowing conditions. The other blow-out condition is determined by means of a linear model. This should avoid any increase in cost. However, the interior air temperature in the passenger compartment is also determined here by the interior air temperature sensor and therefore cannot always be regarded as correct or reliable.

Disclosure of Invention

The aim of the invention is to enable improved temperature control in a motor vehicle. According to the invention, this object is achieved by the subject matter of the independent claims. Advantageous embodiments and further developments of the invention are given in the dependent claims, the description and the drawings.

The method according to the invention is used for controlling or regulating an air conditioning system of a motor vehicle and comprises a plurality of method steps. In one method step of the method according to the invention, a machine learning device is provided, which is trained with the aid of predefined or provided training data for estimating the actual air temperature in the passenger space of the motor vehicle. The machine learning device may preferably be or comprise a neural network. For generating the trained machine learning device, i.e. for example for generating a correspondingly trained neural network, a training method known per se, for example a method based on back propagation, can be used. The training data used for this purpose may include, for example, the temperature or temperature data measured by means of at least one temperature sensor of the motor vehicle and the associated actual air temperature in the passenger space, i.e., in the position or air volume where at least one passenger of the motor vehicle typically rests, as desired or as prescribed. Additionally, the training data may include other data, values, or signals, such as blower power, carburetor temperature, and the like. Here, the machine learning device may be trained or to be trained for different situations or given conditions, which is also explained in more detail below.

In a further method step of the method according to the invention, data relating to the air conditioning or tempering of the passenger compartment are acquired, which data indicate at least one temperature measured by means of a temperature sensor of the motor vehicle. The data may be encoded or represented digitally or analog. That is, the data may be read from an electronic data memory or retrieved in the form of an analog signal, for example. The acquired data or signals and/or the data or variables derived or calculated therefrom are then supplied as input data, i.e. fed to the machine learning device.

In a further method step of the method according to the invention, these data are processed by the machine learning device to determine an estimated temperature of the passenger space. In other words, the machine learning device outputs at least one temperature value, which corresponds to the actual air temperature in the passenger space, on the basis of or as a function of the respective supplied input data. The temperature estimated by the machine learning device may deviate from the temperature measured by the temperature sensor of the motor vehicle. This may be caused, for example, by different temperatures or different thermal properties of different components and regions of the motor vehicle, as explained above. Thus, in particular, heat radiation or heat conduction from the component surrounding the temperature sensor to the temperature sensor can influence, in particular increase, the temperature measured by the temperature sensor, which does not correspond to or reflect the air temperature actually present in the passenger compartment. In this case, this effect is taken into account automatically by the machine learning device and with a particularly low computational effort (compared to a complete digital simulation of the entire interior) of the motor vehicle device, which optionally includes heat generation arranged outside the interior.

In a further method step of the method according to the invention, the air conditioning device is automatically controlled or regulated as a function of the estimated temperature, which is regarded as the respective current actual value. That is, the determined estimated temperature can be provided or supplied as a respective current actual value of the air conditioning device and as an input value to a control unit or control device of the air conditioning device or, for example, to another program module or control module of the air conditioning device. The machine learning device may be part of such a control device or part of an air conditioning device. However, the machine learning device can also be a separate unit or device, which can then be connected to the air conditioning device via a data connection, for example via the onboard electrical system of the motor vehicle.

The estimated temperature can be determined continuously or regularly, for example at a predetermined frequency, during the operation of the air conditioning system, so that the method according to the invention, i.e. during operation, can be carried out continuously or regularly or at least partially accordingly. The provision of a machine learning device may mean or include, for example, a loading or activation at the start of the method, i.e., for example, at the start of the operation of the air conditioning system or the motor vehicle.

The machine learning device can be trained specifically for the respective vehicle or vehicle type, i.e. its thermal behavior or its thermal behavior can be automatically taken into account or simulated. The machine learning device can thereby determine or estimate the air temperature actually present in the passenger space particularly precisely and reliably with relatively little computational effort. In order to achieve better results with conventional measures by means of analytical models or thermodynamic simulations of the motor vehicle or of the interior, however, it is necessary to have a manageable computational effort which is not available in motor vehicles. Furthermore, the creation of such models or such simulations is accompanied by significant manufacturing costs, which can currently be saved at least in part. By means of the invention, the air temperature actually present in the passenger compartment can be determined more accurately and more reliably than has hitherto been possible by using temperature sensors alone or using sufficiently simplified evaluation models in terms of computational complexity.

The invention thus enables a more accurate and more reliable regulation or maintenance of a specific temperature in the passenger space. This in turn can improve the thermal comfort of the respective occupant of the motor vehicle. The invention also provides the possibility of controlling or regulating the air conditioning device on the basis of more complex variables and/or a plurality of actual temperature values in different regions of the passenger compartment, by means of a corresponding training of the machine learning device, without additional hardware complexity and therefore particularly cost-effectively. For this purpose, the machine learning device can be provided and trained, for example, for outputting a corresponding temperature profile of the passenger compartment, which can, for example, indicate an average temperature in the passenger compartment, temperatures at different locations of the passenger compartment, or a temperature profile within or across the passenger compartment. Likewise, other values, data or signals can be taken into account in a particularly simple manner when training the machine learning device and subsequently also when determining the estimated temperature. The estimated temperature can thus also be determined particularly accurately and reliably in different situations or under different conditions, i.e. particularly close to the respective actual air temperature.

In an advantageous embodiment of the invention, for generating the training data, at least one reference temperature sensor is provided in the passenger space in an air volume region in which an occupant of the motor vehicle is expected to stay when the motor vehicle is in use and/or which is empirically particularly relevant for air conditioning in the passenger space. Such a particularly relevant position can be located, for example, directly in front of the dashboard or at a distance of a few centimeters from the dashboard. By means of the reference temperature sensor, the reference temperatures measured under different conditions, i.e. the temperatures actually present in the passenger space or air volume region, are then correlated with the temperatures measured at the same time by means of the temperature sensor at the time of their measurement. In other words, the training data comprise at least a value pair or a value tuple consisting of a temperature value measured by means of a temperature sensor and at least one reference temperature value. Furthermore, the training data can accordingly comprise or indicate at which position in the passenger space and/or under which conditions the respective reference temperature value or the respective temperature value was measured. The respective condition or situation may be indicated or characterized, for example, by one or more settings or states of the motor vehicle, ambient conditions or environmental conditions (e.g., ambient temperature, insolation, and/or the like). The training data may also comprise further values, data or signals, for example data from a blower control, measurement data from one or more further temperature sensors (for example a temperature sensor for the air flowing from the outlet into the passenger compartment), and/or the like. These training data may be measured or acquired simultaneously with the mentioned temperatures, for example. The correlation of the respective measured or acquired data at a particular point in time with one another can then be carried out, for example, by means of a respective time stamp.

Since the training data need only be generated once, the reference temperature sensor can be optimally positioned for this purpose and removed from the motor vehicle after the training data have been generated, so that the described advantages of the invention can be achieved in the subsequent operation of the motor vehicle, while the reference temperature sensor does not have to be located in the motor vehicle. That is, occupant comfort is therefore not limited by the reference temperature sensor, nor is the manufacturing cost of the motor vehicle significantly increased.

In an advantageous further development of the invention, a plurality of reference temperature sensors are arranged spatially distributed in the passenger space for generating the training data. Then, a reference temperature value is calculated from the temperatures measured by means of some or all of the reference temperature sensors as part of the training data, in particular as an average of the temperatures measured at the particular time in each case. In this way, the temperature or the temperature distribution actually present in the passenger space can be taken into account or represented particularly accurately and at the same time simply. In particular, local temperature fluctuations in the passenger compartment can be taken into account or compensated for, as a result of which ultimately a more precise and more reliable control or regulation of the air conditioning device and accordingly a more precise and more reliable regulation or maintenance of a corresponding target passenger compartment temperature, for example, which is predefined by the user or the passenger, can be achieved. As already indicated, however, the reference temperature value can be not only a single temperature value, but can also indicate, for example, a spatial temperature profile or, for example, a corresponding average temperature in different regions or sections of the passenger space.

In a further advantageous embodiment of the invention, the temperature is measured by means of a temperature sensor, which is integrated for this purpose in an interior component of the motor vehicle, in particular in a center console or in an instrument panel. In other words, the temperature sensor is not located freely in the air volume or in the region that is to be occupied by the occupant, but is integrated in the interior fitting or trim panel of the motor vehicle. This does not limit the comfort of use of the passenger space and enables a particularly inconspicuous and less distracting arrangement of the temperature sensor. This enables, for example, a particularly attractive and ergonomic configuration of the vehicle interior. The invention allows particularly great flexibility in the arrangement of the temperature sensor, since the position-dependent measurement inaccuracies of the temperature sensor are compensated for by the machine learning device.

In a further advantageous embodiment of the invention, at least one control signal is acquired as part of the data relating to the air conditioning of the passenger compartment, which control signal indicates a control performed by an occupant of the motor vehicle. In particular, the control signal can indicate the controlled blower power, the blower orientation and/or the position of a side window or a roof window of the motor vehicle. By means of such an occupant or user adjustment, the temperature behavior of the motor vehicle or the temperature distribution or temperature change in the occupant space can be influenced. By acquiring the respective adjustment signal and processing it to determine the estimated temperature, the respective influence can be taken into account. As a result, a particularly precise and reliable temperature or air conditioning of the passenger space can ultimately be achieved. The control signal can be acquired, for example, via the onboard power supply of the motor vehicle. Finally, the control signal can relate to almost any control, device or function of the motor vehicle, at least insofar as it is considered as part of the training data for generating the machine learning device.

In a further advantageous embodiment of the invention, at least one status signal is acquired as part of the data relating to the air conditioning of the passenger compartment, which at least one status signal indicates a respective current state of the technical device of the motor vehicle and/or a respective current state of the surroundings of the motor vehicle. In particular, the status signal can indicate the respective current operating state, power requirement, power output, rotational speed, current uninterrupted operating duration, ambient temperature and/or insolation. In other words, factors or influences which may influence the temperature or thermal behavior in the passenger compartment and/or the temperature measured by the temperature sensor can be detected and taken into account by means of the status signal. The technical device of the vehicle may be, for example, a motor, a battery, a cooling device, a pump and/or the like. Such technical devices can generate heat during their operation, which heat can also lead to heating of other structures or components of the motor vehicle. The insolation may give the intensity and/or direction of the insolation. In this case, the state signal can be used in particular to obtain and, for determining the estimated temperature and, ultimately, for controlling or regulating the air conditioning device, to take into account states, influences or variables which are not directly predetermined or set in a targeted manner by the respective user or passenger of the motor vehicle, in particular non-dynamically, during operation of the motor vehicle, or which can be predetermined or set. This may ultimately also lead to or contribute to a more precise and more reliable, i.e. improved, tempering or air conditioning of the interior space.

In a further advantageous embodiment of the invention, an artificial neural network is used as the machine learning device, wherein during the training thereof at least one parameter learned by the trained neural network is removed and the neural network or model thus reduced in complexity is retrained. The parameter to be removed can be selected, for example, by determining the influence of the parameter on the estimated temperature determined by the neural network by means of a corresponding analysis. For example, the parameter can be removed when its influence falls below a predefined threshold value. This may be repeated iteratively, for example until a certain compromise between the size or complexity of the neural network and the accuracy or reliability of the respective result (i.e. the estimated temperature) is reached and/or until, for example, a predefined condition or a predefined complexity criterion and/or accuracy criterion is fulfilled. In this way, the computing power required for carrying out the method according to the invention can advantageously be reduced, wherein the actual thermal behavior of the respective motor vehicle is taken into account particularly simply and particularly precisely.

In a further advantageous embodiment of the invention, use is made of a system having at least one LSTM unit (English and jargon: long short-term memory, German: langes)

) An artificial neural network (long-short term memory) is used as a machine learning device. The LSTM unit may here comprise an input gate, a memory gate (Merkstor) and a forgetting gate as well as an output gate. Thus, the time profile or the time variation can be taken into account when determining the estimated temperature. This is particularly advantageous for accurately and reliably determining the estimated temperature, since thermal systems typically develop with some inertia. It has been found that particularly advantageous control or regulation of the air conditioning system can be achieved with such a network or model architecture.

Another aspect of the present invention is an air conditioning apparatus for a motor vehicle. The air conditioning system according to the invention comprises a machine learning device which is trained for the actual air temperature in the passenger compartment of the motor vehicle by means of provided or predefined training data. The air conditioning system furthermore comprises a control unit for automatically controlling or regulating the air conditioning system, wherein the air conditioning system is designed to carry out or carry out at least one variant of the method according to the invention, in particular automatically or partially automatically. That is, the air conditioning device according to the invention may in particular be the air conditioning device mentioned in connection with the method according to the invention. Correspondingly, the air conditioning device according to the invention may have some or all of the characteristics and/or features mentioned in connection with the method according to the invention. In order to carry out the method according to the invention, the air conditioning system or its control unit can have an electronic or computer-readable data memory and a processor unit connected to the data memory. On the data memory, a computer program that can be executed by means of the processor device and represents or encodes a process or method steps of the method according to the invention or corresponding control instructions for the control unit or the air conditioning unit can be stored. If the machine learning device is implemented fully or partially in software, said software part of the machine learning device may also be stored in the data storage. The machine learning means can in principle also be implemented completely or partially in hardware.

Another aspect of the invention is a motor vehicle having at least one temperature sensor and an air conditioning device according to the invention. The temperature sensor can be part of the air conditioning system or be connected to the air conditioning system. The motor vehicle according to the invention can in particular be the motor vehicle mentioned in connection with the air conditioning device according to the invention and/or in connection with the method according to the invention. Correspondingly, the motor vehicle according to the invention may have some or all of the characteristics and/or features mentioned in connection with the air conditioning device according to the invention and/or in connection with the method according to the invention.

Drawings

Further features of the invention can be derived from the claims, the figures and the description of the figures. The features and feature combinations mentioned above in the description and the features and feature combinations mentioned below in the description of the figures and/or shown in the figures only can be applied not only to the respectively given combination but also to other combinations or can be applied individually without departing from the scope of the invention.

In the drawings:

fig. 1 shows an exemplary schematic flowchart of a method for controlling an air conditioning system of a motor vehicle; and

fig. 2 shows a schematic view of a motor vehicle with a corresponding air conditioning device.

Detailed Description

An exemplary flowchart 10 schematically illustrated in fig. 1 serves to illustrate a method for temperature control or air conditioning in a vehicle. The flowchart 10 here includes method steps S1 to S6, which are explained in more detail below with reference to fig. 2. Fig. 2 shows a schematic illustration of a motor vehicle 12 with a passenger compartment 14 for this purpose. The interior device 16 of the motor vehicle 12 is currently located in or adjacent to the passenger compartment 14. The internal device 16 may be, for example, an instrument panel or a center console. The majority of the volume of the passenger compartment 14 is occupied by the passenger space 18. When the motor vehicle 12 is operated or used, one or more occupants can be located in the occupant space 18, i.e., in a corresponding air volume inside the motor vehicle 12.

In addition, the motor vehicle 12 has an onboard electrical system 20, to which different devices are connected, between which signals or data can be exchanged or transmitted via the onboard electrical system 20. In the present case, the air conditioning device 22 is connected to the onboard electrical system 20. The air conditioning unit 22 includes a controller 24, the controller 24 itself including a processor 26 and a data storage 28 connected to the processor. The processor 26 may be or include, for example, a microprocessor, microchip or microcontroller. An operating program for the air conditioning system 22 can be stored on the data memory 28. The operating program can be executed by the processor 26 in order to automatically or partially automatically execute at least one of the method steps S1 to S6. In the present case, the operating program comprises, in particular, a machine learning device, preferably a neural network, which is trained for estimating the actual air temperature in the passenger space 18.

In addition, the motor vehicle 12 has a ventilation device 30 and a temperature sensor 32, which are likewise connected to the on-board electrical system 20. The ventilation device 30 and/or the temperature sensor 32 may be part of the air conditioning device 22. For example, both the ventilation device 30 and the temperature sensor 32 are integrated in the interior 16. The ventilation device 30 may be or comprise an air outlet and/or a blower, for example. The ventilation device 30 and/or the air conditioning device 22 may also comprise, for example, an air conditioning unit for tempering the air that can be supplied to the passenger compartment 18.

The method steps S1 and S2 are preparatory steps which can be carried out once by the manufacturer of the motor vehicle 12 or of the air conditioning system 22 within the scope of its manufacture, for example, by the respective user, in particular before normal operation of the motor vehicle 12. In method step S1, training data for the machine learning device of the air conditioner 22 is first generated. For this purpose, the air temperature actually present in the passenger space 18 is measured as a reference temperature by means of a reference temperature sensor 34 arranged in the passenger space 18. At the same time, the temperature measured by the temperature sensor 32 is acquired. This can be repeated several times under different conditions. In this case, for example, various user settings or user presets of the ventilation device 30 or the air conditioning device 22, for example, can be present and acquired. Furthermore, the different conditions can be provided, for example, by at least one further, here schematically illustrated, different state of the vehicle component 36, which is likewise in each case acquired for generating training data or as part of the training data.

That is, the training data is a data set containing different data or signals. These data or signals may be divided, for example, into operating data and reference data. The operating data here includes data or signals provided or acquired by devices of the motor vehicle 12 (i.e., here, for example, data or signals of the air conditioning device 22, the temperature sensor 32, the vehicle component 36 and/or other sensors, devices or components of the motor vehicle 12). The reference data includes data or signals obtained by means of a reference device, in this case a reference temperature sensor 34, which is temporarily provided in the motor vehicle 12 only for generating the training data. The reference data may be or include data or signals measured or recorded directly by the reference temperature sensor 34. Additionally or alternatively, the reference data may be or include a quantity calculated or derived therefrom, for example a temperature profile or an average temperature over the space in the passenger space 18. The temperature in a specific region of the passenger space 18, for example, at a corresponding location at which the head of the respective occupant is expected to be located during operation of the motor vehicle 12, can likewise be calculated from the reference data or as part of the reference data.

In method step S2, training data of the machine learning device (which is not yet trained or is only pre-trained at this time) is provided as input data. The machine learning device is then trained with the training data. In this case, the complexity of the machine learning device or the model represented or simulated by the machine learning device can be reduced in order to be able to carry out subsequent calculations more efficiently or with less computational effort by means of the machine learning device. Finally, in method step S2, a finally trained machine learning device is thus generated. The trained machine learning device can then be provided for further method steps, for example by loading it into the data memory 28.

After these preliminary steps, method steps S3 to S6 can be carried out repeatedly during operation of the motor vehicle 12, which is indicated here by a circular path.

In a method step S3, the temperature is measured by means of the temperature sensor 32. For example, in parallel with this, in method step S4, at least one control signal and at least one status signal are acquired, which indicate the control performed by the occupant of motor vehicle 12 and the respective current state of the technical device of motor vehicle 12 (here, for example, vehicle component 36). The data or signals measured and acquired in method steps S3 and S4 form operating data of the motor vehicle 12, which are supplied as input data to a machine learning device.

In method step S5, these operating data are processed by means of a machine learning device. Here, the machine learning device estimates the current air temperature actually present in the air volume of the passenger space 18 on the basis of the operating data and outputs this air temperature.

In method step S6, controller 24 controls or regulates air conditioning unit 22 based on the determined estimated air temperature. This may mean or include, for example, the generation and output of corresponding control signals, for example to an air conditioning compressor or condenser, a coolant pump, a blower, a servomotor and/or the like.

That is, in summary, a machine learning model for estimating the interior temperature of the passenger compartment 14 is constructed here in order to achieve an improved control of the vehicle air comfort on the basis thereof. In summary, the described example shows how vehicle air conditioning with intelligent temperature estimation can be achieved.

List of reference numerals

10 flow chart

12 motor vehicle

14 passenger compartment

16 internal device

18 occupant space

20 vehicle electrical system

22 air conditioner

24 controller

26 processor

28 data memory

30 ventilating element

32 temperature sensor

34 reference temperature sensor

36 vehicle component

Claims

1. A method (10) for controlling an air conditioning device (22) of a motor vehicle (12), having the following method steps:

-providing a machine learning device (24, 26, 28) which is trained with the aid of predefined training data for estimating an actual air temperature in a passenger space (18) of the motor vehicle (12),

-acquiring data relating to the air conditioning of the passenger space (18), said data giving at least one temperature measured by means of a temperature sensor (32) of the motor vehicle (12),

-processing said data by said machine learning means (24, 26, 28) to determine an estimated temperature of the occupant space (18), and

-automatically controlling the air conditioning device (22) according to the estimated temperature, considered as the respective current actual value.

2. Method (10) according to claim 1, characterized in that, for generating the training data, in the passenger space (18) at least one reference temperature sensor (34) is provided in the air volume region (18) in which the passenger of the motor vehicle (12) rests as desired when the motor vehicle is in use, and the reference temperatures measured by means of the reference temperature sensor (34) under different conditions are correspondingly associated with the temperatures measured simultaneously by means of the temperature sensor (32).

3. Method (10) according to claim 2, characterized in that for generating the training data a plurality of reference temperature sensors (34) are spatially distributed in the passenger space (18) and that reference temperature values are calculated from the temperatures measured by means of some or all of the plurality of reference temperature sensors (34) as part of the training data, in particular as an average of the temperatures respectively measured at a specific time.

4. The method (10) according to any one of the preceding claims, wherein the temperature is measured by means of a temperature sensor (32), wherein the temperature sensor (32) is integrated in an interior device element (16) of the motor vehicle (12), in particular in a center console (16) or an instrument panel (16).

5. Method (10) according to one of the preceding claims, characterized in that at least one adjustment signal is acquired as part of the data relating to the air conditioning of the passenger space (18), which at least one adjustment signal gives an adjustment made by an occupant of the motor vehicle (12), in particular a blower power, a blower orientation and/or a position of a side window or a roof window of the motor vehicle (12).

6. Method (10) according to one of the preceding claims, characterized in that at least one status signal is acquired as part of the data relating to the air conditioning of the passenger space (18), which status signal indicates a respective current status of a technical device (30, 36) of the motor vehicle (12) and/or a respective current status of the surroundings of the motor vehicle (12), in particular an operating state, a power demand, a power output, a rotational speed, a current uninterrupted operating duration, an ambient temperature and/or insolation.

7. The method (10) according to one of the preceding claims, characterized in that an artificial neural network is used as the machine learning device (24, 26, 28), wherein at least one parameter learned by the trained neural network is removed during its training and the neural network thus reduced in complexity is retrained as a result.

8. The method (10) according to any one of the preceding claims, using an artificial neural network with at least one LSTM unit as a machine learning device (24, 26, 28).

9. An air conditioning device (22) of a motor vehicle (12), comprising a machine learning device (24, 26, 28) which is trained with the aid of predefined training data for estimating an actual air temperature in a passenger space (18) of the motor vehicle (12), and comprising a controller (24) for automatically controlling the air conditioning device (22), wherein the air conditioning device (22) is designed for carrying out the method (10) according to one of the preceding claims.

10. A motor vehicle (12) having a temperature sensor (32) and an air conditioning device (22) according to claim 9.