DE102019131722B3 - Method for controlling or regulating air conditioning of a vehicle - Google Patents

Abstract

The invention relates to a method for controlling or regulating the air conditioning of a vehicle. At least one neuron of an input layer of a first artificial neural network (2) receives at least one first signal (1) specific to the vehicle type as an input signal. At least one neuron of an output layer of the first artificial neural network (2) outputs at least one first signal (3), which is independent of the vehicle type, as an output signal. At least one neuron of an input layer of a second artificial neural network (4) receives the at least one first vehicle-type-independent signal (3) as an input signal. At least one neuron of an output layer of the second artificial neural network (4) emits at least one second vehicle-type-independent signal (5) as an output signal out. At least one neuron of an input layer of a third artificial neural network (6) receives the at least one second vehicle-type-independent signal (5) as an input signal. At least one neuron of an output layer of the third artificial neural network (6) emits at least one second vehicle-type-specific signal (7) as an output signal The air conditioning of the vehicle is controlled or regulated by means of the at least one second signal (7) specific to the vehicle type.

Description

Technical area

The invention relates to a method for controlling or regulating air conditioning of a vehicle according to the preamble of claim 1.

State of the art

• • Ein Verfahren zur Klimatisierung eines Fahrzeugs. Dies ist eine Reihe von Funktionen, die aus Messungen des Fahrzeugzustands Größen zur Beeinflussung des Fahrzeugzustands ableiten. Diese Reihe von Funktionen ist häufig in Form von Modellbildung, Definition von Kennlinien und Filterungen unter zu Hilfenahme von softwarebasierten Regelungsfunktionen, beispielsweise sog. PID-Regler, erstellt.
• • Sensoren bzw. Vorrichtungen zur Messung des Fahrzeugzustands. Diese Sensoren messen beispielsweise die Außentemperatur bzw. Umgebungstemperatur des Fahrzeugs, die Sonneneinstrahlung auf das Fahrzeug, die Fahrzeuginnenraumtemperatur, die Luftfeuchtigkeit innerhalb des Fahrzeugs und/oder die Luftfeuchtigkeit außerhalb des Fahrzeugs.
• • Aktuatoren bzw. Vorrichtungen zur Beeinflussung des Fahrzeugzustands. Diese Aktuatoren beeinflussen den Fahrzeugzustand durch Änderung von beispielsweise der Spannung des Gebläses, der Drehzahl des Gebläses, der Temperatur der Abluft des Gebläses, der relativen Luftmenge, die in den Fußraum des Fahrzeugs strömt, und/oder der relativen Luftmenge, die unter die Frontscheibe strömt.
• • Steuergeräte bzw. Vorrichtungen zur Berechnung der Größen zur Beeinflussung des Fahrzeugzustands. Messungen der Sensoren werden an diese Steuergeräte weitergeleitet. Auf diesem Steuergerät läuft dann das Verfahren zur Klimatisierung des Fahrzeugs und liefert die Stellgrößen für die Aktuatoren. Das Steuergerät sendet dann diese Stellgrößen an die Aktuatoren.

Air conditioning of a vehicle can include the following components:

• • A method for air conditioning a vehicle. This is a series of functions that derive variables from measurements of the vehicle condition to influence the vehicle condition. This series of functions is often created in the form of modeling, definition of characteristics and filtering with the help of software-based control functions, for example so-called PID controllers.
• • Sensors or devices for measuring the vehicle condition. These sensors measure, for example, the outside temperature or ambient temperature of the vehicle, the solar radiation on the vehicle, the vehicle interior temperature, the humidity inside the vehicle and / or the humidity outside the vehicle.
• • Actuators or devices for influencing the vehicle condition. These actuators influence the vehicle condition by changing, for example, the voltage of the blower, the speed of the blower, the temperature of the exhaust air from the blower, the relative amount of air that flows into the footwell of the vehicle and / or the relative amount of air that flows under the windshield .
• • Control units or devices for calculating the variables to influence the vehicle condition. Measurements from the sensors are forwarded to these control units. The process for air conditioning the vehicle then runs on this control unit and supplies the control variables for the actuators. The control unit then sends these manipulated variables to the actuators.

A current parameterization of the air conditioning of a vehicle is extremely complex, since specially converted test vehicles are used for which the parameterization is carried out.
Test drives are carried out with this test vehicle.

These test drives are relatively complex and can take several weeks or months.

During the test drives, measurement data are recorded that describe the climate behavior of the surroundings and the interior of the vehicle.

The measurement data contain, for example, the outside temperature or ambient temperature of the vehicle, the solar radiation on the vehicle, the vehicle interior temperature, the humidity inside the vehicle and / or the humidity outside the vehicle.

The measurement data form the basis for parameterizing the vehicle. After the test drives have been carried out and the measurement data recorded, the air conditioning of a vehicle is parameterized.

Functions for generating the control variables, for example the interior temperature or the voltage of the interior fan, are manually created in the form of modeling, definition of characteristics and filtering with the aid of software-based control functions, for example so-called PID controllers.

Test drives are carried out again during the parameterization of the air conditioning. During these test drives, the behavior of the air conditioning is evaluated and it is checked whether it meets the requirements for the air conditioning, otherwise the parameters are adjusted again.

The process described must be carried out from the beginning for each new vehicle type.

Knowledge transfer for air conditioning of a new vehicle type cannot be saved.

The DE 10 2014 007 303 A1 describes a method for controlling or regulating at least one device that is part of a cooling system of a vehicle, in particular of a commercial vehicle, and / or that influences the energy efficiency of vehicle operation.

A neural operation unit is provided for regulating or controlling the device, the neural operation unit being set up by means of an artificial neural network to determine a control value of the device as an output value as a function of transmitted values of several input variables.

The input variables include at least a fluid temperature of the cooling system and a topography of a route section of the vehicle ahead.

The nodes of the artificial neural network based on a correlation between the values of the input variables and the output values are specified.

The correlation is determined by optimizing a total energy consumption of the vehicle in the route section ahead and / or a driving ability of the vehicle without heating-related power throttling in the route section ahead.

Object of the invention

The invention is based on the object of providing a method for controlling or regulating air conditioning of a vehicle with an improved expenditure of time.

Solution of the task

The invention is achieved by a method according to claim 1.

Advantages of the invention

The method according to the invention for controlling or regulating the air conditioning of a vehicle comprises three artificial neural networks.

At least one neuron of an input layer of a first artificial neural network receives at least one first vehicle-type-specific signal as an input signal.

Depending on the vehicle type, a signal can be saved with different names and in one or more vehicle-type-specific signals. For example, the outside temperature can be stored in two individual signals for a first vehicle manufacturer and / or type and in four individual signals with different names for a second vehicle manufacturer and / or type.

The first vehicle-type-specific signal can include values measured in the vehicle, to which current values and recordings of time courses can be counted. Furthermore, the first vehicle-type-specific signal can contain data that are received by means of what is known as a cloud infrastructure, for example the reading of weather data from an Internet database.

In addition, data can represent the first vehicle-type-specific signals that are received by means of what is known as Car-2-x communication, for example weather data or other data relevant for air conditioning.

At least one neuron of an output layer of the first artificial neural network outputs at least one first signal, which is independent of the vehicle type, as an output signal.

The first artificial neural network obtains information from the first vehicle-type-specific signals which is independent of the vehicle type and which is relevant to the method according to the invention.

In this sense, the first artificial neural network can be referred to as an encoder.

The first artificial neural network comprises at least one hidden layer in addition to the input and output layers.

The first artificial neural network can be specific to the vehicle type and can be learned separately for each vehicle type.

For the learning of the first artificial neural network it is necessary that first vehicle-type-specific signals, which also include the aforementioned cloud and car-2-x data, are recorded for each specific vehicle, for example during a test drive in the vehicle with a suitable program / Tool for recording measurement data, or, alternatively, using a model to generate synthetic data for each vehicle type.

With these first vehicle-type-specific signals, an auto-encoder is learned which can generate compression and restoration of the first vehicle-type-specific signals.

Only the compression part of this autoencoder is retained as the first artificial neural network / encoder.

At least one neuron of an input layer of a second artificial neural network receives the at least one first signal, which is independent of the vehicle type, as an input signal.

At least one neuron of an output layer of the second artificial neural network outputs at least one second vehicle-type-independent signal as an output signal, which was determined by means of matrix multiplication and / or matrix addition and transformation by linear or non-linear activation functions.

The second artificial neural network comprises at least one hidden layer in addition to the input and output layers.

The second artificial neural network can only be learned for a subset of all existing vehicle types.

The subset can also be just a single, first vehicle type.

At least one neuron of an input layer of a third artificial neural network receives the at least one second signal, which is independent of the vehicle type, as an input signal.

At least one neuron of an output layer of the third artificial neural network outputs at least one second vehicle-type-specific signal as an output signal.

The third artificial neural network converts the second vehicle-type-independent signals into second vehicle-type-specific signals.

The second vehicle-type-specific signals are in a name designation of the respective vehicle manufacturer and / or type.

In this sense, the third artificial neural network can be referred to as a decoder.

In addition to the input and output layers, the third artificial neural network comprises at least one hidden layer.

The third artificial neural network can be specific to the vehicle type and learned separately for each vehicle type.

The learning takes place by means of an autoencoder, which can generate a compression and restoration of second vehicle type-specific signals.

Only the restoration part of this autoencoder is retained as a third artificial neural network / decoder.

The first artificial neural network, the second artificial neural network and the third artificial neural network can be self-learning and adapt their initial weights and parameterizations based on a target function.

For a first vehicle, the first artificial neural network, the second artificial neural network and the third artificial neural network can be learned together (end-to-end).

The first artificial neural network is then individually trained for each vehicle different from the first vehicle.

The third artificial neural network is then individually trained for each vehicle different from the first vehicle.

The first artificial neural network, the second artificial neural network and the third artificial neural network are then learned together (end-to-end).

The air conditioning of the vehicle is controlled or regulated by means of the at least one second signal specific to the vehicle type.

The second vehicle type-specific signal can be, for example, the target voltage and / or the target speed of a blower, the target vehicle interior temperature, the target exhaust air temperature and / or the target position of a recirculation flap.

A knowledge transfer from a first vehicle manufacturer and / or type to a second or further vehicle manufacturer and / or type different from the first vehicle manufacturer is possible by means of the method according to the invention.

For a first vehicle manufacturer and / or type, a one-time extensive effort comparable to that described in the prior art has to be carried out. I.e. test drives are carried out.

During the test drives, measurement data are recorded that describe the climate behavior of the surroundings and the interior of the vehicle.

The first difference to the known procedures is in this step.

Functions for generating the controlled variables, defining characteristics and filtering with the aid of software-based control functions are not generated manually, but the first artificial neural network, the second artificial neural network and the third artificial neural network are trained automatically.

This results in an air conditioning parameterized for the vehicle manufacturer and / or type.

This automated training is less time-consuming compared to manual generation of functions.

Figure list

• Fig.: ein schematisiertes Blockschaltbild eines erfindungsgemäßen Verfahrens zur Steuerung oder Regelung einer Klimatisierung eines Fahrzeugs.

It shows:

• FIG.: A schematic block diagram of a method according to the invention for controlling or regulating air conditioning of a vehicle.

The figure shows a method according to the invention for controlling or regulating air conditioning of a vehicle.

At least one first signal specific to the vehicle type 1 forms an input signal for a first artificial neural network 2 .

An output signal of the first artificial neural network 2 is at least a first signal that is independent of the vehicle type 3 .

The at least one first signal, which is independent of the vehicle type 3 forms an input signal for a second artificial neural network 4th .

The number of the first signals independent of the vehicle type 3 which is the output of the first artificial neural network 2 form matches the number of the first signals independent of the vehicle type 3 match the input signal of the second artificial neural network 4th form.

For example, they form two first signals that are independent of the vehicle type 3 the output signal of the first artificial neural network 2 , these two first form signals that are independent of the vehicle type 3 also the input signal of the second artificial neural network 4th .

An output signal of the second artificial neural network 4th is at least one second signal that is independent of the vehicle type 5 .

The at least one second signal, which is independent of the vehicle type 5 forms the input signal for a third artificial neural network 6 .

The number of the second, vehicle-type-independent signals 5 which is the output of the second artificial neural network 4th form matches the number of second signals, which are independent of the vehicle type 5 match the input signal of the third artificial neural network 6 form.

For example, they form four second signals that are independent of the vehicle type 5 the output signal of the second artificial neural network 4th , then these four second vehicle-type-independent signals form 5 also the input signal of the third artificial neural network 6 .

An output signal of the third artificial neural network 6 is at least one second signal specific to the vehicle type 7th .

The air conditioning of the vehicle is controlled or regulated by means of the at least one second signal specific to the vehicle type 7th .

List of reference symbols

1

First vehicle type-specific signal

2

First artificial neural network

3

First signal independent of vehicle type

4th

Second artificial neural network

5

Second vehicle type-independent signal

6th

Third artificial neural network

7th

Second vehicle type-specific signal

Claims (6)

Method for controlling or regulating air conditioning of a vehicle, characterized in that a. at least one neuron of an input layer of a first artificial neural network (2) receives at least one first vehicle-type-specific signal (1) as an input signal; b. at least one neuron of an output layer of the first artificial neural network (2) outputs at least one first vehicle-type-independent signal (3) as an output signal; c. at least one neuron of an input layer of a second artificial neural network (4) which receives at least one first signal (3) that is independent of the vehicle type as an input signal; d. at least one neuron of an output layer of the second artificial neural network (4) outputs at least one second signal (5) that is independent of the vehicle type as an output signal; e. at least one neuron of an input layer of a third artificial neural network (6) which receives at least one second signal (5) which is independent of the vehicle type as an input signal; f. at least one neuron of an output layer of the third artificial neural network (6) outputs at least one second vehicle-type-specific signal (7) as an output signal; G. the air conditioning of the vehicle is controlled or regulated by means of the at least one second signal (7) specific to the vehicle type. Procedure according to Claim 1 , characterized in that the first artificial neural network (2) is learned separately for each vehicle type. Procedure according to Claim 1 , characterized in that the third artificial neural network (6) is learned separately for each vehicle type. Procedure according to Claim 1 , characterized in that the first artificial neural network (2), the second artificial neural network (4) and the third artificial neural network (6) are self-learning. Procedure according to Claim 4 , characterized in that the initial weights and parameterizations of the first artificial neural network (2), the second artificial neural network (4) and the third artificial neural network (6) are adapted on the basis of a target function. Procedure according to Claim 1 , characterized in that the first artificial neural network (2), the second artificial neural network (4) and the third artificial neural network (6) are learned together.

Images