Electronics arrangement for an electric motor

Abstract

The invention relates to an electronics arrangement (100) for an electric motor (102), comprising an inverter unit (1), to which an electrical input direct current (ES) is supplied during operation and which is configured to output an electrical output alternating current (AS) to a winding system of an electric motor (102), which is connected to the electronics arrangement (100), during operation, and comprising a direct-current intensity determination unit (2), which is configured to determine a present electrical current intensity (I-ES) of the input direct current (ES); wherein the direct-current intensity determination unit (2) comprises an artificial intelligence module (2.1) for determining the present electrical current intensity (I-ES) of the input direct current (ES) without using a current sensor, wherein at least one measured value (W-IQ, W-RPM, W-T, W-UD, W-UDC) representing a present quantity of a physical variable of the electronics arrangement (100) or of an electric motor (102) connected to the electronics arrangement (100) is provided to the artificial intelligence module (2.1) during operation, and wherein the artificial intelligence module (2.1) is trained to determine the present electrical current intensity (I-ES) of the input direct current (ES) on the basis of the at least one measured value (W-IQ, W-RPM, W-T, W-UD, W-UDC).

Description

[0001] DESCRIPTION

[0002] Electronic arrangement for an electric motor

[0003] The present invention relates to an electronic arrangement for an electric motor, comprising: an inverter unit, to which an electrical input direct current is supplied during operation and which is configured to deliver an electrical output alternating current to a winding system of an electric motor connected to the electronic arrangement, and a direct current intensity determination unit, which is configured to determine a current electrical current intensity of the input direct current.

[0004] The present invention is based on the object of realizing a relatively cost-effective electronic arrangement of this type for an electric motor.

[0005] This object is achieved by an electronic arrangement for an electric motor having the features of claim 1.

[0006] The electronics assembly according to the invention for an electric motor comprises an inverter unit, which is supplied with an electrical input direct current during operation and which is configured to output an electrical output alternating current to a winding system of an electric motor connected to the electronics assembly. Such inverter units are well known in the art and are also referred to as inverters, inverter units, or rotary inverter units. The inverter unit is preferably configured to output a three-phase output alternating current. In principle, however, the inverter unit can also be configured to output a single-phase or any other multi-phase output alternating current.The inverter unit preferably comprises power electronics with one or more semiconductor switches, so-called transistors, particularly preferably with several so-called bipolar transistors with an insulated gate electrode (IGBT for short).

[0007] The electronic arrangement according to the invention for an electric motor further comprises a direct current intensity determination unit which is configured to determine a current electrical current intensity of the input direct current supplied to the inverter unit during operation.

[0008] According to the invention, the direct current determination unit comprises an artificial intelligence module, abbreviated to "Ki-Module," for determining the current electrical current of the input direct current without the use of a current sensor. The Ki-Module is implemented by suitable programming of a computing unit, for example, a microprocessor, or any number of interacting computing units.

[0009] At least one measured value is provided to the Kl module, which represents a current quantity of a physical variable of the electronic arrangement or of an electric motor connected to the electronic arrangement. Preferably, a plurality of measured values ​​are provided to the Kl module, which represent different physical variables of the electronic arrangement and / or the electric motor connected to the electronic arrangement. The at least one measured value can, for example, represent an intermediate circuit voltage, a temperature of the inverter unit, in particular a temperature of one or more semiconductor switches of the inverter unit, a torque-generating current component of the output alternating current, also called quadrature current or iQ for short, a field-generating voltage component caused by the output alternating current, also called uD for short, and / or a current speed of the electric motor connected to the electronic arrangement.

[0010] The Kl module is trained in a manner known from the prior art to determine the current electrical current of the input direct current based on the at least one provided measured value by applying methods / algorithms known from the prior art and associated with the field of so-called artificial intelligence. The Kl module can, for example, comprise a so-called artificial neural network and / or be configured to implement methods / algorithms of so-called machine learning.

[0011] Because the DC current determination unit comprises a Kl module according to the invention, the current electrical current of the input DC current can be determined with relatively high accuracy in all operating ranges of the electronics assembly or the electric motor connected to the electronics assembly, without the need for an additional current sensor. This enables the realization of a relatively cost-effective electronics assembly for an electric motor.

[0012] In a preferred embodiment, the Kl module comprises a so-called artificial neural network, which enables an effective determination of the current electrical current of the input direct current with relatively little computing power. This enables the use of a relatively inexpensive computing unit to implement the Kl module and thus the realization of a relatively inexpensive electronic arrangement for an electric motor.

[0013] Artificial neural networks are typically constructed in the form of several consecutive layers, with each layer typically comprising several artificial neurons, and the artificial neurons in one layer are each linked to at least one artificial neuron in the next layer, so that each artificial neuron passes its output on to at least one artificial neuron in the next layer. The first layer of the neural network, which is provided with the input parameters, in this case at least one measured value, is also called the input layer. The last layer of the neural network, which provides the output parameters, in this case the current electrical current of the input direct current, is also called the output layer. All layers between the input layer and the output layer are also called hidden layers.

[0014] In a preferred embodiment, the artificial neural network comprises only a single hidden layer. This makes it possible to achieve a sufficiently accurate determination of the current intensity of the input direct current with relatively low computational effort, thus allowing the use of a particularly cost-effective computing unit, such as a microprocessor, to implement the AI ​​module.

[0015] Preferably, each hidden layer of the artificial neural network is implemented as a fully connected layer. In a fully connected layer, each artificial neuron in the respective layer is connected to every single artificial neuron in the preceding layer and to every single artificial neuron in the subsequent layer. This enables a particularly effective determination of the current electrical current of the input direct current with minimal computing power, thus allowing the use of a particularly cost-effective computing unit for the implementation of the AI ​​module.

[0016] To efficiently represent nonlinear effects, the artificial neural network, in a preferred embodiment, has a sigmoid function, also called a swan-neck function, Fermi function, or S-function, as the activation function for at least one artificial neuron. Preferably, at least all artificial neurons of the at least one hidden layer of the artificial neural network each have a sigmoid function as their activation function.

[0017] Since training artificial intelligence typically requires more computing power than executing the trained artificial intelligence, the AI ​​module is preferably implemented as a fully trained model. The AI ​​model is thus trained in advance, for example, in the laboratory, using a relatively powerful computing unit, and the fully trained AI model is then implemented in the computing unit of the AI ​​module according to the invention. This enables the use of a particularly cost-effective computing unit for the implementation of the AI ​​module.

[0018] Preferably, the DC current determination unit is configured to determine the current electrical current of the input DC current at a relatively low repetition frequency in the range of 10 Hz to 1000 Hz. This enables relatively resource-efficient operation of the DC current determination unit.

[0019] An embodiment of the present invention is described below with reference to the accompanying figures. Herein:

[0020] Fig. 1 is a schematic diagram of an electronic arrangement according to the invention for an electric motor, and

[0021] Fig. 2 is a schematic diagram of an artificial intelligence module of the electronics assembly of Fig. 1.

[0022] Fig. 1 shows an electronic arrangement 100 according to the invention which is connected to a direct current source 101 and to an electric motor 102.

[0023] The electronic arrangement 100 comprises an inverter unit 1 which is electrically connected to the direct current source 101 and to the electric motor 102.

[0024] During operation, the inverter unit 1 is supplied with an electrical input direct current ES from the direct current source 101.

[0025] The inverter unit 1 is configured in a manner known from the prior art to provide, during operation, a three-phase electrical output alternating current AS to a winding system 102.1 of the electric motor 102 (not shown in detail).

[0026] The electronics assembly 100 further comprises a direct current determination unit 2 configured to determine a current electrical current I-ES of the input direct current ES. The direct current determination unit 2 comprises an artificial intelligence module (abbreviated to: AI module) 2.1 for determining the current electrical current I-ES of the input direct current ES without the use of a current sensor.

[0027] The Kl module 2.1 includes an artificial neural network 2.2, which is implemented as a pre-trained model.

[0028] The artificial neural network 2.2 is provided with a speed measurement value WR.PM, which represents a current speed of the electric motor 102, a quadrature current measurement value W-IQ, which represents a torque-forming current component of the output alternating current AS, a voltage measurement value W-UD, which represents a field-forming voltage component caused by the output alternating current AS, a voltage measurement value W-UDC, which represents an intermediate circuit voltage applied to the input side of the inverter unit 1, and a temperature measurement value WT, which represents a temperature of a semiconductor switch of the inverter unit 1, as input values.

[0029] The artificial neural network 2.2 comprises - as shown in Fig. 2 - an input layer 2.2.1, a single hidden layer 2.2.2 and an output layer 2.2.3.

[0030] The input layer 2.2.1 comprises five artificial neurons 2.2.1.1, wherein the rotational speed measurement value WR.PM is provided to one of the artificial neurons 2.2.1.1 of the input layer 2.2.1, the quadrature current measurement value W-IQ is provided to one of the artificial neurons 2.2.1.1 of the input layer 2.2.1, the voltage measurement value W-UD is provided to one of the artificial neurons 2.2.1.1 of the input layer 2.2.1, the voltage measurement value W-UDC is provided to one of the artificial neurons 2.2.1.1 of the input layer 2.2.1, and the temperature measurement value WT is provided to one of the artificial neurons 2.2.1.1 of the input layer 2.2.1.

[0031] The hidden layer 2.2.2 comprises a multitude of artificial neurons 2.2.2.1, each of which has a sigmoid function as its activation function.

[0032] The hidden layer 2.2.2 is implemented as a fully connected layer, which means that each artificial neuron 2.2.2.1 of the hidden layer 2.2.2 is connected to each individual artificial neuron 2.2.1.1 of the input layer 2.2.1 as well as to each individual artificial neuron 2.2.3.1 of the output layer 2.2.3.

[0033] The output layer 2.2.3 has a single artificial neuron 2.2.3.1, whose output value represents the current electrical current I-ES of the input direct current ES.

[0034] The artificial neural network 2.2 is trained in a manner known from the prior art to determine the current electrical current I-ES of the input direct current ES based on the speed measurement value WR.PM, the quadrature current measurement value W-IQ, the voltage measurement value W-UD, the voltage measurement value W-UDC and the temperature measurement value WT.

[0035] The direct current detection unit 2 is configured to determine the current electrical current I-ES of the input direct current ES using the Kl module 2.1 with a repetition frequency in the range of 10 Hz to 1000 Hz, preferably with a repetition frequency of 100 Hz. List of reference symbols

[0036] 100 electronic arrangement

[0037] 1 inverter unit

[0038] 2 DC current detection unit

[0039] 2.1 Artificial Intelligence Module

[0040] 2.2 artificial neural network

[0041] 2.2.1 Input layer

[0042] 2.2.1.1 artificial neurons

[0043] 2.2.2 hidden layer

[0044] 2.2.2.1 artificial neurons

[0045] 2.2.3 Output layer

[0046] 2.2.3.1 artificial neuron

[0047] 101 DC source

[0048] 102 Electric motor

[0049] 102.1 Winding system

[0050] AS output alternating current

[0051] ES input DC current

[0052] I-ES electrical current of the input direct current

[0053] W-IQ quadrature current measurement value

[0054] WR.PM speed measurement value

[0055] WT temperature measurement

[0056] W-UD voltage measurement value

[0057] W-UDC voltage measurement value

Claims

P A T E N T A N S P R Ü C H E 1. Electronic arrangement (100) for an electric motor (102), comprising: an inverter unit (1) to which an electrical input direct current (ES) is supplied during operation, and which is designed to output an electrical output alternating current (AS) to a winding system of an electric motor (102) connected to the electronic arrangement (100), and a direct current intensity determination unit (2) which is designed to determine a current electrical current intensity (I-ES) of the input direct current (ES), characterized in that the direct current intensity determination unit (2) comprises an artificial intelligence module (2.1) for the current sensor-free determination of the current electrical current intensity (I-ES) of the input direct current (ES), wherein the artificial intelligence module (2.1) during operation, at least one measured value (W-IQ, W-RPM, WT, W-UD, W-UDC) is provided, which represents a current quantity of a physical variable of the electronic arrangement (100) or of an electric motor (102) connected to the electronic arrangement (100), and wherein the artificial intelligence module (2.1) is trained to determine the current electrical current intensity (I-ES) of the input direct current (ES) based on the at least one measured value (W-IQ, W-RPM, WT, W-UD, W-UDC).

2. Electronic arrangement (100) according to claim 1, wherein the artificial intelligence module (2.1) comprises an artificial neural network (2.2).

3. Electronic arrangement (100) according to claim 2, wherein the artificial neural network (2.2) comprises a single hidden layer (2.2.2).

4. Electronic arrangement (100) according to claim 2 or 3, wherein each hidden layer (2.2.2) of the artificial neural network (2.2) is implemented as a fully connected layer.

5. Electronic arrangement (100) according to one of claims 2 to 4, wherein the artificial neural network (2.1) has a sigmoid function as an activation function.

6. Electronic arrangement (100) according to one of the preceding claims, wherein the artificial intelligence module (2.1) is implemented as a fully trained model.

7. Electronic arrangement (100) according to one of the preceding claims, wherein the direct current intensity determination unit (2) is arranged to determine the electrical current intensity (I-ES) of the input direct current (ES) with a repetition frequency in the range of 10 Hz to 1000 Hz.

Images