Apparatus and computer-implemented method for determining temperature of motor rotor and training hybrid model for determining the temperature

Abstract

A device and computer-implemented method for determining the temperature of the rotor (104) of an electric motor (101), wherein the temperature of the stator (102) of the electric motor (101) is measured, wherein the temperature of the stator (102) in the region of the stator (102) facing the rotor (104) is determined based on the temperature of the stator (102), in particular by means of a model (114), the model being constructed to map the temperature of the stator (102) to the temperature of the stator (102) in the region of the stator (102) facing the rotor (104), wherein a physical model (118) is provided, the physical model modeling a first part of an ordinary differential equation for heat transfer, particularly from the stator (102) to the rotor (104), based on the temperature of the stator (102) in the region of the stator (102) facing the rotor (104), wherein a data-driven model (120) is provided. The data-driven model models the second part of the ordinary differential equation based on the temperature of the stator (102) in the region of the stator (102) facing the rotor (104), wherein the temperature of the stator (102) in the region of the stator (102) facing the rotor (104) is mapped by the physical model (118) to the temperature of the rotor (104) or a first part of the heat transfer to the rotor (104), wherein the temperature of the stator (102) in the region of the stator (102) facing the rotor (104) is mapped by the data-driven model (120) to the temperature of the rotor (104) or a second part of the heat transfer to the rotor (104), and wherein the temperature of the rotor (104) is determined based on the first part and the second part, in particular based on the sum of the first part and the second part. An apparatus and computer-implemented method for training a hybrid model (116) for determining the temperature of the rotor (104) of the motor (101) are also described.

Description

Technical Field

[0001] This invention is based on a device and computer-implemented method for determining the rotor temperature of an electric motor and training a hybrid model for determining that temperature. Summary of the Invention

[0002] A computer-implemented method for determining the rotor temperature of an electric motor specifies: measuring the temperature of the motor stator, wherein, in particular, a model is used to determine the stator temperature in the stator-to-rotor region based on the stator temperature, the model being constructed to map the stator temperature to the stator temperature in the stator-to-rotor region; wherein a physical model is provided, the physical model modeling a first part of an ordinary differential equation for heat transfer, particularly from the stator to the rotor, based on the stator temperature in the stator-to-rotor region; wherein a data-driven model is provided, the data-driven model modeling a second part of the ordinary differential equation based on the stator temperature in the stator-to-rotor region, wherein the stator temperature in the stator-to-rotor region is mapped to a first part of heat transfer to the rotor by the physical model, wherein the stator temperature in the stator-to-rotor region is mapped to a second part of heat transfer to the rotor by the data-driven model; and wherein the rotor temperature is determined based on the first part and the second part, in particular based on the sum of the first part and the second part, or based on a parameter characterizing the temperature of the rotor (104), wherein the parameter is determined based on the sum of the first part and the second part.

[0003] In one example, the physical model is constructed to model the heat transfer from the stator and motor coolant to the rotor, where the first part of the heat transfer is determined based on the cooling temperature of the coolant using the physical model.

[0004] In one example, the data-driven model is constructed to model heat transfer based on at least one input parameter, particularly cooling temperature, motor input voltage, or current in the motor, wherein a second part of the heat transfer is determined using the data-driven model based on at least one input parameter.

[0005] It can be specified that: the stator hot spot temperature is determined by means of a model for determining the stator hot spot temperature based on the stator temperature, wherein the model for determining the stator hot spot temperature includes features, and / or wherein the model configured to map the stator temperature to the stator temperature in the stator-to-rotor region includes features, wherein the physical model is configured to determine the first portion based on these features, and / or wherein the data-driven model is configured to determine the second portion based on these features.

[0006] It can be specified that when the parameter or rotor temperature is greater than or equal to the threshold, the cooling intensity of the coolant on the motor should be increased.

[0007] It can be specified that: for multiple values ​​of the parameter, the probability distribution of the parameter, especially the 99th percentile, is determined; or for multiple values ​​of rotor temperature, the probability distribution of the rotor temperature, especially the 99th percentile, is determined. The quantile of the probability distribution is compared with a threshold. If the quantile of the probability distribution is less than the threshold, it is identified as having no increased risk to component protection. If the quantile of the probability distribution is greater than or equal to the threshold, it is identified as having an increased risk to component protection. And / or when the quantile is greater than or equal to the threshold, the cooling intensity of the coolant on the motor is increased.

[0008] A method for training a computer-implemented hybrid model for determining motor rotor temperature specifies: providing multiple tuples, each containing a stator temperature in a stator-to-rotor region and a reference value for the rotor temperature; providing a physical model that models a first part of an ordinary differential equation for heat transfer, particularly from the stator to the rotor, based on the stator temperature in the stator-to-rotor region; providing a data-driven model that models a second part of the ordinary differential equation based on the stator temperature in the stator-to-rotor region; wherein for each tuple, the stator temperature in the stator-to-rotor region is mapped by the physical model to a rotor temperature or a first part of heat transfer to the rotor, and by the data-driven model to a rotor temperature or a second part of heat transfer to the rotor; wherein the rotor temperature is determined based on the first and second parts, particularly based on the sum of the first and second parts; and wherein the physical model and / or the data-driven model are trained based on the deviation between the rotor temperature determined for the corresponding tuple and the reference value of the rotor temperature in the corresponding tuple.

[0009] A method for training a computer implementation of a hybrid model for determining the rotor temperature of an electric motor specifies that: multiple tuples are provided, each containing reference values ​​for stator temperature and rotor temperature; a physical model is provided that models a first part of an ordinary differential equation for heat transfer, particularly from the stator to the rotor, based on the stator temperature in the stator-to-rotor region; a data-driven model is provided that models a second part of the ordinary differential equation based on the stator temperature in the stator-to-rotor region; wherein for each tuple, the stator temperature in the stator-to-rotor region is determined based on the stator temperature, particularly with the aid of the model; the model is constructed to map the stator temperature to a value in the stator-to-rotor region. The stator temperature in the rotor region is mapped to the rotor temperature or a first part of heat transfer to the rotor via a physical model, and to the rotor temperature or a second part of heat transfer to the rotor via a data-driven model. The rotor temperature is determined based on the first and second parts, particularly based on the sum of the first and second parts, or based on a parameter characterizing the rotor temperature, wherein the parameter is determined based on the sum of the first and second parts. The physical model and / or the data-driven model are trained based on the deviation between the rotor temperature determined for a corresponding tuple in a plurality of tuples and a reference value of the rotor temperature in the corresponding tuple.

[0010] The training method may specify that the physical model is constructed to model the heat transfer from the stator and motor coolant to the rotor, wherein the first part of the heat transfer is determined by means of the physical model based on the cooling temperature of the coolant, and wherein these tuples respectively contain the cooling temperature of the coolant.

[0011] The method for training may specify that: the data-driven model is constructed to model heat transfer based on at least one input parameter, particularly cooling temperature, motor input voltage, or current in the motor, wherein a second part of the heat transfer is determined by means of the data-driven model based on at least one input parameter, and wherein these tuples respectively contain the at least one input parameter.

[0012] The method for training may specify: determining the stator hot spot temperature using a model for determining the stator hot spot temperature based on the stator temperature, wherein the model for determining the stator hot spot temperature includes features, and / or wherein a model configured to map the stator temperature to a stator temperature in the stator-to-rotor region includes features, wherein the physical model is configured to determine a first portion based on these features, and / or wherein the data-driven model is configured to determine a second portion based on these features.

[0013] The training method may specify that the hybrid model includes parameters of a device for determining the motor rotor temperature, wherein the hybrid model is trained using a computing environment set up externally to the device, wherein updated parameters are determined during training, wherein the device is connected to the computing environment externally via, in particular, a wireless communication connection to update the parameters, wherein the parameters in the device are replaced with the updated parameters.

[0014] An apparatus for determining motor rotor temperature or training a hybrid model for determining motor rotor temperature is configured to perform the method for determining temperature or the method for training.

[0015] In one example, the apparatus includes at least one processor and at least one non-volatile memory, wherein the at least one processor is configured to execute instructions, and when the at least one processor executes the instructions, the apparatus performs the method, and wherein the at least one non-volatile memory stores the instructions.

[0016] A computer program may also be provided, which contains instructions executable by a computer, wherein when the computer executes the instructions, the method is run on the computer. Attached Figure Description

[0017] Other advantageous embodiments can be derived from the following description and accompanying drawings. In the drawings: Figure 1 A first schematic diagram of the device is shown. Figure 2 A second schematic diagram of the device is shown. Figure 3 A first flowchart showing the steps of a method for determining the rotor temperature of an electric motor is shown. Figure 4 A second flowchart showing the steps of a first method for training a hybrid model for determining temperature is shown. Figure 5 A third flowchart shows the steps of a second method for training a hybrid model for determining temperature. Detailed Implementation

[0018] Figure 1 The diagram schematically illustrates the temperature characterizing the rotor 104 used to determine the motor 101. Parameters The device 100, wherein the motor has a stator 102 and a rotor 104. In one example, the parameter... It is the temperature of rotor 104. In one example, the parameter It is the temperature characterizing rotor 104. The state. In one example, the parameter Includes the temperature characterizing rotor 104 The state. Parameters can be specified. It is multi-dimensional and includes the operating status of motor 101. Parameters can be specified. Temperature including rotor 104 The temperature of rotor 104 can be specified. Can be derived from parameters Calculated.

[0019] In the example, device 100 includes a temperature measurement device for stator 102. The input terminal. The example includes a temperature sensor 106, which is configured to measure the temperature of the stator 102. .

[0020] The apparatus 100 may optionally include a first device 108 for anomaly identification or reasonableness verification.

[0021] The first device 108 is configured to identify temperature. Abnormalities, or temperature Perform a rationality check.

[0022] The first device 108 is configured to: monitor the temperature of the stator 102 as long as no anomaly is detected and / or during a reasonableness check. If the reading is deemed reasonable, the temperature of stator 102 will be provided. Otherwise, the temperature of stator 102 will not be provided. .

[0023] The apparatus 100 may optionally include a second device 118 for anomaly identification or validity verification.

[0024] The second device 118 is configured to recognize input parameters. Anomalies, or issues with input parameters Perform a rationality check.

[0025] The second device 118 is configured to: input parameters will be processed as long as no anomaly is detected and / or during reasonableness checks. If the input is deemed reasonable, then provide the input parameters. Otherwise, no input parameters are provided. .

[0026] Input parameters For example, input voltage, cooling temperature The current of motor 101.

[0027] The device 100 includes a first model 112. The first model 112 is configured to detect the temperature of the stator 102 as measured by the temperature sensor 106. Determine the hot spot temperature of stator 102. The first model 112 can be an empirical model, a physical model, or a data-driven model. The first model 112 can be used to extract features F, such as model states, and provide these features.

[0028] The device 100 includes a second model 114. The second model 114 is configured to respond to the measured temperature of the stator 102. and input parameters Determine the temperature of the stator 102 in the region of the stator 102 facing the rotor 104. The second model 114 can be an empirical model, a physical model, or a data-driven model. The second model 114 can be used to extract features F, such as model states, and provide these features.

[0029] The first model 112 may have parameters that depend on the temperature of the stator 102 as measured by the temperature sensor 106. And attributed to the measured temperature The hot spot temperature of the stator 102 measured is a pair that has been learned or verified.

[0030] The second model 114 may have parameters that depend on the measured temperatures of the stator 102. And attributed to the measured temperature The temperature of stator 102 measured in the region of stator 102 facing rotor 104 constitutes a pair that has been learned or verified.

[0031] Used to determine hot spot temperature Model 112 and / or used to determine temperature Model 114 contains feature F.

[0032] These features F can be used to determine hotspot temperatures. Model 112 or used to determine temperature The state of model 114, or used to determine hotspot temperatures. Model 112 or used to determine temperature The control parameters of model 114.

[0033] For example, the first model 112 includes a first artificial neural network, wherein features F are provided to the layers of the first artificial neural network at time point t.

[0034] For example, the second model 114 includes a second artificial neural network, wherein features F are provided to the layers of the second artificial neural network at time point t.

[0035] The device 100 includes a hybrid model 116 for mixing ordinary differential equations.

[0036] Mixed ordinary differential equations are, for example: The physical part is f, the data-driven part is g, and the parameters are... To implicitly or explicitly characterize the temperature state of rotor 104, particularly in Kelvin. In the example, the parameter... The change characterizes the heat transfer on rotor 104.

[0037] The mixed ordinary differential equations are defined based on the characteristic F from the first model 112 and / or the second model 114.

[0038] The physical component f is modeled, for example, by a physical model 118 having a parameter α of the physical component f, wherein the physical component f at time point t adjusts the parameter α according to the physical component f. The output is mapped to the physical part f.

[0039] The output of the physics part f in the example depends on the input parameters. Cooling temperature To determine. For example, based on cooling temperature. The difference between the temperature of rotor 104 and the latest calculated temperature is used to determine the cooling temperature. The coolant flows to the heat flow of rotor 104. In the parametric... In the example of the temperature of rotor 104, the difference is calculated. In parameters In the example characterizing the temperature of rotor 104, the difference is determined. ,in Indicates based on parameters The temperature of rotor 104 is determined. In this example, heat flow is achieved by cooling the temperature. The temperature difference between the newly calculated rotor 104 temperature and the actual temperature is determined by multiplying the specific thermal conductivity of the coolant. In this example, the specific thermal conductivity of the coolant is calculated based on parameter α and characteristic F within the range of the physical part f.

[0040] For example, the flow rate of the coolant and the temperature of the coolant calculated according to the following rules Can be used as a feature: .

[0041] It can be specified that the temperature of rotor 104 is estimated by means of the Laplace approximation of parameter α. The uncertainty of parameters Provide a probabilistic description, wherein the parameter It is determined based on the Laplace approximation of the parameter α. This means that the parameter is determined. and for the parameters Uncertainty.

[0042] For example, based on the latest calculated temperature of stator 102 The difference between the temperature of the rotor 104 and the latest calculated temperature is used to determine the heat flow from the stator 102 to the rotor 104. In the parametric... In the example of the temperature of rotor 104, the difference is calculated. In parameters In the example characterizing the temperature of rotor 104, the difference is determined. ,in Indicates based on parameters The determined temperature of rotor 104. In this example, heat flow is determined by the latest calculated temperature of stator 102. The difference between the current rotor 104 temperature and the latest calculated temperature is multiplied by the specific thermal conductivity of the stator 102. In this example, for the stator 102, the specific thermal conductivity of the stator 102 is calculated using the physical parameter α and the characteristic F according to the following rule: Where n is the motor speed.

[0043] Furthermore, in this example, the power loss of rotor 104 is calculated within the range of physical part f. This is based on the characteristic F and the physical parameter α. For example, the power loss is calculated using a family of characteristic curves whose parameters are contained in the physical parameter α. The input parameter of this family of characteristic curves is, in this example, the phase current of the motor. and And the motor speed n.

[0044] Therefore, the temperature difference of rotor 104 calculated through this physical part f in the example is: in The transfer function represents the heat flow from stator 102 to rotor 104, and This represents the heat transfer function from the coolant to the rotor 104.

[0045] The transfer function and This is an illustrative example of the equations in the physics section f. The concept can also be applied to other equations with respect to these parameters.

[0046] The data-driven component g is modeled, for example, by a data-driven model 120 having a parameter β of the data-driven component g, wherein the data-driven component g, at time point t, modulates the parameters according to the parameter β of the data-driven component g. The output is mapped to the data-driven portion g. This output is used to correct the predictions of the physical portion f in a data-driven manner, reflecting effects not reflected in the physical portion f.

[0047] In the example, the output is compared with the estimated change. Add them together. Other combination types and methods can also be used for the output, such as weighted addition.

[0048] Hybrid model 116 pairs of parameters The prediction of the process can be determined, for example, by using numerical methods for solving ordinary differential equations. These methods involve discretizing the ordinary differential equations over time, such as the Euler method.

[0049] The hybrid model 116 is constructed based on the measured temperature provided at time point t. The temperature determined for this purpose Features F and input parameters Determine the parameters obtained by solving the differential equation at time t. .

[0050] The parameters α and β are, for example, based on whether multiple tuples are being learned or validated, and these tuples each contain the measured input parameters. The measured temperature of rotor 104 The measured temperature of stator 102 The measured hot spot temperature of stator 102 and the temperature of stator 102 in the region of stator 102 facing rotor 104, these parameters correspond to each other, especially at time point t.

[0051] For example, the least squares method is used to determine the parameter α. For example, the gradient descent method is used to determine the parameter β. In the parameter... In the example of the temperature of rotor 104, for example, the following parameters α, β are determined, for these parameters, the parameters of the tuple. The deviation from the measured temperature of rotor 104 should be as small as possible.

[0052] In parameters In the example characterizing the temperature of rotor 104, a corresponding method is used, wherein the following parameters α and β are determined, for these parameters, for the corresponding parameters of the tuple. The determined temperature The deviation from the measured temperature of rotor 104 should be as small as possible.

[0053] In the example, device 100 is configured to determine the temperature of the rotor 104 of motor 101 in a method for determining temperature or for training, or to determine the temperature of a hybrid model 116 used for training the temperature of the rotor 104 of motor 101. Training can be specified to take place in a vehicle, and particularly retraining in a computing environment set outside the vehicle, i.e., outside device 100, such as a cloud computing environment. For example, device 100 is configured to update parameters α and / or β in device 100 via a wireless communication connection, i.e., "over-the-air download". For example, tuples in the vehicle are transmitted to a computing environment outside device 100, and parameters α and / or β, i.e., new information, are learned centrally in the computing environment outside device 100. For example, device 100 is set in different vehicles, wherein the computing environment is configured to learn parameters α and / or β based on tuples in different vehicles, and provide the learned parameters α and / or β to each vehicle, i.e., device 100 in the respective vehicle. This enables parameter... A more precise determination.

[0054] Device 100 is configured, for example, based on parameters This more precise state determination optimizes the operation of at least one component or system in the vehicle. In this example, the component or system includes a rotor 104. In this example, the temperature of the rotor 104 is... The temperature of the component or system is affected. For example, device 100 is configured to: be based on a quantized probability p (e.g., p < 0.0001%) or a parameter The quantiles of the probability distribution of the value, or the temperature of rotor 104. The probability distribution quantiles are used to optimize the operating strategy for running the component or system to avoid, for example, exceeding critical temperatures and damage to the component or system.

[0055] In one example, device 100 is constructed such that: when parameter Or the temperature of rotor 104 Greater than the threshold At the same time, this increases the intensity of cooling of the motor by the coolant.

[0056] For example, device 100 is configured to: for parameters Multiple values ​​to determine the parameter For example, the 99th percentile of a probability distribution. That is, for the parameter The confidence interval.

[0057] For example, device 100 is configured to: respond to the temperature of rotor 104 Multiple values ​​are used to determine the temperature of rotor 104. For example, the 99th percentile of the probability distribution. That is, regarding the temperature of the rotor 104 The confidence interval.

[0058] For example, the plurality of values ​​are determined for different time points based on the temperature of the stator 102 as determined at those different time points. For example, the plurality of values ​​are determined for different time points based on input parameters as determined at those different time points. To determine.

[0059] For example, device 100 is configured to: divide the probability distribution into quantiles With threshold A comparison is performed. For example, device 100 is configured such that only the quantiles of the probability distribution are considered. Less than the threshold This is identified as a component protection risk without an increase. For example, device 100 is constructed such that as long as the quantiles of the probability distribution... Greater than or equal to the threshold This is identified as an increased risk to component protection. For example, device 100 is configured such that when the quantile of the probability distribution... Greater than or equal to the threshold At that time, the cooling intensity of the coolant is increased.

[0060] Figure 2 A portion of the device 100 is shown schematically.

[0061] The device 100 includes at least one processor 202 and at least one non-volatile memory 204.

[0062] The at least one processor 202 is configured to execute instructions, and when the at least one processor 202 executes the instructions, the device 100 performs a corresponding method.

[0063] The at least one non-volatile memory 204 stores the instructions.

[0064] It can be specified that the device 100 includes an interface 206 for the temperature sensor 106 and / or for inputting parameters. The interface can be specified as follows: the device 100 includes a temperature sensor 106.

[0065] Figure 3 The diagram shows a computer-implemented method for determining the temperature of rotor 104. The flowchart shows the steps of the method.

[0066] The method for determining the temperature of rotor 104 The method includes step 300.

[0067] In step 300, a physical model 118 and a data-driven model 120 are provided.

[0068] The method for determining the temperature of rotor 104 The method includes step 302.

[0069] In step 302, the temperature of stator 102 is measured. .

[0070] The method for determining the temperature of rotor 104 The method includes step 304.

[0071] In step 304, based on the temperature of stator 102 Determine the temperature of the stator 102 in the region of the stator 102 facing the rotor 104. .

[0072] In the example, the temperature of the stator in the region of the stator 102 facing the rotor 104 is determined using the second model 114. .

[0073] It can be specified that the hot spot temperature of stator 102 is determined using the first model 112. .

[0074] The method for determining the temperature of rotor 104 The method includes step 306.

[0075] In step 306, the temperature of the stator 102 in the region of the stator 102 facing the rotor 104 is measured. The first part of the heat transfer is mapped to the rotor 104 via physical model 118. It can be specified that, using physical model 118, the cooling temperature of the coolant is considered. Determine the first part of heat transfer.

[0076] In step 306, the temperature of the stator 102 in the region of the stator 102 facing the rotor 104 is measured. The second part of the heat transfer is mapped to the rotor 104 via the data-driven model 120. It can be specified that the data-driven model 120 is based on at least one input parameter. Determine the second part of heat transfer.

[0077] In the example, the first and second parts are parameters. The part.

[0078] It can be specified that the first model 112 and / or the second model 114 contain feature F.

[0079] It can be stipulated that the first part is determined based on feature F using physical model 118.

[0080] It can be stipulated that the second part is determined based on feature F using the data-driven model 120.

[0081] The method for determining the temperature of rotor 104 The method includes step 308.

[0082] In step 308, the temperature of rotor 104 is determined based on the first part and the second part. .

[0083] For example, the temperature of rotor 104 is determined based on the sum of the first part and the second part. .

[0084] In one example, the parameter Temperature of rotor 104 That is, the temperature of rotor 104 It is determined based on the solution results of the ordinary differential equations associated with the sum of the first and second parts.

[0085] In one example, the parameter Temperature characterizing rotor 104 That is, the temperature of rotor 104 According to parameters Confirmed, this parameter It is the result of solving the ordinary differential equation related to the sum of the first and second parts.

[0086] The method can specify that: when the parameter Or the temperature of rotor 104 Greater than the threshold At the same time, this increases the intensity of cooling of the motor by the coolant.

[0087] The method can specify: for parameters Multiple values ​​to determine the parameter The probability distribution, such as the normal distribution or the 99th percentile of the student distribution. .

[0088] The method can specify: for the temperature of rotor 104 Multiple values ​​are used to determine the temperature of rotor 104. The probability distribution, such as the normal distribution or the 99th percentile of the student distribution. Quantiles This represents the confidence interval.

[0089] The method can specify: the quantiles of the probability distribution With threshold Compare them.

[0090] The method can specify that: as long as the quantiles of the probability distribution are... Less than the threshold This is identified as a component protection risk without an increase.

[0091] The method can specify that: as long as the quantiles of the probability distribution are... Greater than or equal to the threshold This is identified as an increased risk to component protection.

[0092] The method can specify that: when the quantiles of the probability distribution... Greater than or equal to the threshold When this happens, the cooling intensity of the coolant increases.

[0093] Figure 4 The diagram shows a flowchart of the steps of a first method, particularly a computer-implemented method, for training a hybrid model 116 to determine the temperature of the rotor 104.

[0094] The first method for training the hybrid model 116 includes step 400.

[0095] In step 400, a plurality of tuples are provided, each containing the temperature of the stator 102 in the region of the stator 102 facing the rotor 104. And the reference value for the temperature of rotor 104.

[0096] The first method for training the hybrid model 116 includes step 402.

[0097] In step 402, a physical model 118 and a data-driven model 120 are provided.

[0098] The first method for training the hybrid model 116 includes step 404.

[0099] In step 404, for each tuple, the temperature of the stator 102 in the region of the stator 102 facing the rotor 104 is determined. The temperature of the rotor 104 is mapped to the first part of the heat transfer to the rotor 104 through the physical model 118.

[0100] In step 404, for each tuple, the temperature of the stator 102 in the region of the stator 102 facing the rotor 104 is determined. The data-driven model 120 is mapped to the temperature of rotor 104 or the second part of heat transfer to rotor 104.

[0101] The first method for training the hybrid model 116 includes step 406.

[0102] In step 406, for each tuple, the temperature of rotor 104 is determined based on the first and second portions determined for the corresponding tuple. .

[0103] In the example, the temperature of rotor 104 is determined based on the sum of the corresponding first and second parts. .

[0104] The first method for training the hybrid model 116 includes step 408.

[0105] In step 408, the temperature of rotor 104 is determined based on the temperature of the corresponding element among the plurality of elements. The deviation from the reference value of the temperature of rotor 104 in the corresponding tuple is used to train the physical model 118 and / or the data-driven model 120.

[0106] Figure 5 The diagram shows a hybrid model 116, specifically computer-implemented, for training to determine the temperature of rotor 104. The flowchart of the steps of the second method.

[0107] The second method for training the hybrid model 116 includes step 500.

[0108] In step 500, a plurality of tuples are provided, each containing the temperature of the stator 102. And the reference value for the temperature of rotor 104.

[0109] The second method for training the hybrid model 116 includes step 502.

[0110] In step 502, a physical model 118 and a data-driven model 120 are provided.

[0111] The second method for training the hybrid model 116 includes step 504.

[0112] In step 504, for each tuple, based on the temperature of stator 102... Determine the temperature of the stator 102 in the region of the stator 102 facing the rotor 104. .

[0113] In the example, the temperature of the stator 102 in the region of the stator 102 facing the rotor 104 is determined using the second model 114. .

[0114] The second method for training the hybrid model 116 includes step 506.

[0115] In step 506, the temperature of the stator 102 in the region of the stator 102 facing the rotor 104 is measured. The temperature of the rotor 104 is mapped to the first part of the heat transfer to the rotor 104 through the physical model 118.

[0116] In step 506, the temperature of the stator 102 in the region of the stator 102 facing the rotor 104 is measured. The data-driven model 120 is mapped to the temperature of rotor 104 or the second part of heat transfer to rotor 104.

[0117] The second method for training the hybrid model 116 includes step 508.

[0118] In step 508, the temperature of rotor 104 is determined based on the first part and the second part. .

[0119] In the example, the temperature of rotor 104 is determined based on the sum of the first and second parts. .

[0120] The second method for training the hybrid model 116 includes step 510.

[0121] In step 510, the temperature of rotor 104 is determined based on the corresponding tuple among the plurality of tuples. The deviation from the reference value of the temperature of rotor 104 in the corresponding tuple is used to train the physical model 118 and / or the data-driven model 120.

[0122] During training, it can be specified that: using physical model 118, the cooling temperature of the coolant is considered. The first part of heat transfer is determined. In training, in this case, these tuples, for example, contain the cooling temperature of the coolant. .

[0123] During training, it can be specified that: using a data-driven model 120 based on at least one input parameter The second part of heat transfer is determined. In training, in this case, these tuples, for example, each contain at least one of the input parameters. .

[0124] During training, it can be specified that the hot spot temperature of stator 102 is determined using the first model 112.

[0125] During training, it can be specified that the first model 112 and / or the second model 114 contain features F.

[0126] During training, it can be specified that: physical model 118 determines the first part based on feature F, and / or data-driven model 120 determines the second part based on feature F.

[0127] For example, the least squares method is used to determine the parameter α during training. Alternatively, gradient descent is used to determine the parameter β during training.

[0128] The training method may specify that the hybrid model 116 is trained using a computing environment set up outside the device 100.

[0129] During training, for example, the updated parameters α and / or β are determined.

[0130] For example, in order to update parameters, the device 100 connects to a computing environment outside the device 100 via a communication connection, particularly wirelessly. Parameters α and / or β in the device are replaced with updated parameters, for example, via the communication connection.

Claims

1. A computer-implemented method for determining a temperature of a rotor (104) of an electrical machine (101), characterized in that, (302) The temperature of the stator (102) of the motor (101) is measured, wherein the temperature of the stator (102) in the region of the stator (102) facing the rotor (104) is determined (304) based on the temperature of the stator (102), in particular by means of a model (114), the model being constructed to map the temperature of the stator (102) to the temperature of the stator (102) in the region of the stator (102) facing the rotor (104), wherein (300) a physical model (118) is provided, the physical model modeling the first part of the ordinary differential equation for heat transfer, particularly from the stator (102) to the rotor (104), based on the temperature of the stator (102) in the region of the stator (102) facing the rotor (104), wherein (300) a data-driven model (120) is provided, the data-driven model being based on the temperature of the stator (102) in the region of the stator (102) facing the rotor (104) The temperature of the stator (102) in the region is used to model the second part of the ordinary differential equation, wherein the temperature of the stator (102) in the region of the stator (102) facing the rotor (104) is mapped (306) to a first part of the heat transfer to the rotor (104) by the physical model (118), wherein the temperature of the stator (102) in the region of the stator (102) facing the rotor (104) is mapped (306) to a second part of the heat transfer to the rotor (104) by the data-driven model (120), and wherein the temperature of the rotor (104) is determined (308) according to the first part and the second part, in particular according to the sum of the first part and the second part, or according to a parameter characterizing the temperature of the rotor (104), wherein the parameter is determined according to the sum of the first part and the second part.

2. The method of claim 1, wherein, The physical model (118) is configured to model the heat transfer from the coolant in the stator (102) and motor (101) to the rotor (104), wherein the first part of the heat transfer is determined (306) based on the cooling temperature of the coolant by means of the physical model (118).

3. The method according to any of the preceding claims, characterized in that, The data-driven model (120) is constructed based on at least one input parameter ( The heat transfer is modeled, particularly the cooling temperature, the input voltage of the motor (101), or the current in the motor (101), wherein the heat transfer is modeled using a data-driven model (120) based on at least one input parameter ( (306) Determine the second part of the heat transfer.

4. The method according to any of the preceding claims, characterized in that, The hot spot temperature of the stator (102) is determined (304) by means of a model (112) for determining the hot spot temperature of the stator (102) based on the temperature of the stator (102), wherein the model (112) for determining the hot spot temperature of the stator (102) includes features (F), and / or wherein a model (114) configured to map the temperature of the stator (102) to the temperature of the stator (102) in the region of the stator (102) toward the rotor (104) includes features, wherein the physical model (118) is configured to determine the first portion based on the features, and / or wherein the data-driven model (120) is configured to determine the second portion based on the features.

5. The method according to any of the preceding claims, characterized in that, When the parameter or the temperature of the rotor (104) is greater than or equal to a threshold, the cooling intensity of the coolant on the motor is increased.

6. The method according to any of the preceding claims, characterized in that, For multiple values ​​of the parameter, the probability distribution of the parameter is determined, in particular the 99th percentile; or for multiple values ​​of the temperature of the rotor (104), the probability distribution of the temperature of the rotor (104) is determined, in particular the 99th percentile. The quantile of the probability distribution is compared with a threshold, and if the quantile of the probability distribution is less than the threshold, it is identified as having no increased component protection risk; if the quantile of the probability distribution is greater than or equal to the threshold, it is identified as having an increased component protection risk, and / or when the quantile of the probability distribution is greater than or equal to the threshold, the cooling intensity of the coolant on the motor (101) is increased.

7. A computer-implemented method for training a hybrid model (116) for determining the temperature of the rotor (104) of an electric motor (101), characterized in that, Provides (400) multiple tuples, each containing a reference value for the temperature of the stator (102) in the region of the stator (102) facing the rotor (104), and a reference value for the temperature of the rotor (104), wherein (402) a physical model (118) is provided, which models a first part of the ordinary differential equation for heat transfer, particularly from the stator (102) to the rotor (104), based on the temperature of the stator (102) in the region of the stator (102) facing the rotor (104), wherein (402) a data-driven model (120) is provided, which models a second part of the ordinary differential equation based on the temperature of the stator (102) in the region of the stator (102) facing the rotor (104), wherein for each tuple, a reference value for the temperature of the stator (102) in the region of the stator (102) facing the rotor (104) is provided, wherein for each tuple, a reference value for the temperature of the stator (102) in the region of the stator (102) facing the rotor (104) is provided, and a reference value for the temperature of ... The temperature of the stator (102) in the region of 104 is mapped (404) by the physical model (118) to the temperature of the rotor (104) or a first part of the heat transfer to the rotor (104), and mapped (404) by the data-driven model (120) to the temperature of the rotor (104) or a second part of the heat transfer to the rotor (104); and wherein the temperature of the rotor (104) is determined (406) based on the first part and the second part, in particular based on the sum of the first part and the second part; and wherein the physical model (118) and / or the data-driven model (120) are trained (408) based on the deviation of the temperature of the rotor (104) determined for the corresponding tuple in the plurality of tuples from the reference value of the temperature of the rotor (104) in the corresponding tuple.

8. A computer-implemented method for training a hybrid model (116) for determining the temperature of the rotor (104) of an electric motor (101), characterized in that, Provides (500) multiple tuples, each containing reference values ​​for the temperature of the stator (102) and the temperature of the rotor (104), wherein (502) a physical model (118) is provided, which models the first part of the ordinary differential equation for heat transfer, particularly from the stator (102) to the rotor (104), based on the temperature of the stator (102) in the region of the stator (102) facing the rotor (104), wherein (502) a data-driven model (120) is provided, which is based on the temperature of the stator (102) in the region of the stator (102) facing the rotor (104). The second part of the ordinary differential equation is modeled by the temperature of the stator (102) in the region of the stator (102) facing the rotor (104), wherein for each tuple, in particular by means of a model (114) constructed to map the temperature of the stator (102) to the temperature of the stator (102) in the region of the stator (102) facing the rotor (104), the temperature of the stator (102) in the region of the stator (102) facing the rotor (104) is determined (504) based on the temperature of the stator (102). The temperature of the stator (102) in the region of the stator (102) facing the rotor (104) is mapped (506) by the physical model (118) to the temperature of the rotor (104) or a first portion of the heat transfer to the rotor (104), and mapped (506) by the data-driven model (120) to the temperature of the rotor (104) or a second portion of the heat transfer to the rotor (104), and wherein the temperature of the rotor (104) is determined (508) based on the first portion and the second portion, especially based on The temperature of the rotor (104) is determined based on the sum of the first part and the second part or based on a parameter characterizing the temperature of the rotor (104), wherein the parameter is determined based on the sum of the first part and the second part, and wherein the physical model (118) and / or the data-driven model (120) are trained (510) based on the deviation of the temperature of the rotor (104) determined for the corresponding tuple in a plurality of tuples from a reference value of the temperature of the rotor (104) in the corresponding tuple.

9. The method according to claim 7 or 8, characterized in that, The physical model (118) is configured to model the heat transfer from the coolant of the stator (102) and the motor (101) to the rotor (104), wherein the physical model (118) determines (404; 506) a first portion of the heat transfer based on the cooling temperature of the coolant, and wherein the tuples respectively contain the cooling temperature of the coolant.

10. The method according to any one of claims 7 to 9, characterized in that, The data-driven model (120) is constructed based on at least one input parameter ( The heat transfer is modeled, particularly the cooling temperature, the input voltage of the motor (101), or the current in the motor (101), wherein the heat transfer is modeled using the data-driven model (120) based on at least one input parameter. ) Determine (404; 506) the second part of the heat transfer, and wherein the tuples respectively contain the at least one input parameter ( ).

11. The method according to any one of claims 7 to 10, characterized in that, The hot spot temperature of the stator (102) is determined (204) by means of a model (112) for determining the hot spot temperature of the stator (102) based on the temperature of the stator (102); wherein the model (112) for determining the hot spot temperature of the stator (102) includes features, and / or wherein a model (114) configured to map the temperature of the stator (102) to the temperature of the stator (102) in the region of the stator (102) toward the rotor (104) includes features, wherein the physical model (118) is configured to determine the first portion based on the features, and / or wherein the data-driven model (120) is configured to determine the second portion based on the features.

12. The method according to any one of claims 7 to 11, characterized in that, The hybrid model (116) includes parameters of a device (100) for determining the temperature of the rotor (104) of the motor (101), wherein the hybrid model (116) is trained in training with the aid of a computing environment set up outside the device (100), wherein updated parameters are determined in training, wherein the device (100) is connected to the computing environment outside the device (100) via, in particular, a wireless communication connection in order to update the parameters, wherein the parameters in the device are replaced with the updated parameters.

13. An apparatus (100) for determining the temperature of the rotor (104) of an electric motor (101) or for training a hybrid model (116) for determining the temperature of the rotor (104) of an electric motor (101), characterized in that, The apparatus (100) is configured to perform the method according to any one of claims 1 to 12.

14. The apparatus (100) according to claim 13, characterized in that The device (100) includes at least one processor (202) and at least one non-volatile memory (204), wherein the at least one processor (202) is configured to execute instructions, and when the at least one processor (202) executes the instructions, the device (100) performs the method according to any one of claims 1 to 9, and wherein the at least one non-volatile memory (204) stores the instructions.

15. A computer program, characterized in that, The computer program contains instructions executable by a computer, which, when executed by the computer, run the method according to any one of claims 1 to 12 on the computer.

Images