Temperature estimation model training method and device, temperature estimation method and device, equipment and medium

Abstract

The invention provides a temperature estimation model training method and device, an estimation method and device, equipment and a medium, and relates to the technical field of motors, and the method comprises the steps: determining the importance degree of a to-be-judged attribute to a decision result based on a neighborhood rough set algorithm; based on the importance degree, performing attribute reduction on the to-be-judged attribute to obtain an input attribute set; and training a nonlinear autoregressive neural network based on the input data in the input attribute set to obtain a rotor temperature estimation model for reducing the calculation amount when the motor rotor temperature is estimated and accurately estimating the motor rotor temperature.

Description

Technical Field

[0001] The present invention relates to the field of motor technology, and in particular to a temperature estimation model training method, estimation method, device, equipment and medium. Background Art

[0002] Motors are widely used in electric vehicles, industrial drives, and household appliances. Under high power density and high speed operating conditions, excessive rotor temperature can lead to degraded motor performance, shortened lifespan, and even failure.

[0003] Currently, rotor temperature estimation primarily relies on multi-physics simulation algorithms. These algorithms use numerical simulations of electromagnetic and thermal fields to predict temperature distribution, but the calculations are complex and time-consuming, making them difficult to meet real-time monitoring requirements. Alternatively, electromagnetic analysis and heat conduction equations are used to predict temperature changes based on the motor's internal heat source and heat dissipation characteristics, but this is computationally intensive. Summary of the Invention

[0004] The problem solved by the present invention is how to reduce the amount of calculation when estimating the temperature of a motor rotor and to accurately estimate the temperature of the motor rotor.

[0005] To solve the above problems, the present invention provides a temperature estimation model training method, estimation method, device, equipment and medium.

[0006] In a first aspect, the present invention provides a temperature estimation model training method, comprising:

[0007] Based on the neighborhood rough set algorithm, determine the importance of the attribute to be judged to the decision result;

[0008] Based on the importance, performing attribute reduction on the attribute to be judged to obtain an input attribute set;

[0009] Based on the input data in the input attribute set, a nonlinear autoregressive neural network is trained to obtain a rotor temperature estimation model.

[0010] Optionally, determining the importance of the attribute to be judged to the decision result based on the neighborhood rough set algorithm includes:

[0011] Based on the acquired motor rotor attribute sample set, a neighborhood decision system including a condition attribute set and a decision attribute set is constructed;

[0012] The neighborhood relationship between the motor rotor attribute samples is determined based on the Euclidean distance, and the importance of each condition attribute to the decision attribute set is calculated respectively through the neighborhood relationship, wherein the condition attribute includes the attribute to be judged, and the decision attribute set represents the set consisting of the decision results.

[0013] Optionally, determining the neighborhood relationship between the motor rotor attribute samples based on the Euclidean distance, and calculating the importance of each condition attribute to the decision attribute set respectively through the neighborhood relationship includes:

[0014] Based on the neighborhood relationship, determining a first lower approximation set of the decision attribute set with respect to the definition attribute set, and obtaining a first number of elements, wherein the definition attribute set is composed of the conditional attributes, the first lower approximation set represents all motor rotor attribute samples that clearly belong to the decision attribute set under the definition attribute set, and the first number of elements includes the number of elements in the first lower approximation set;

[0015] determining a first dependency of the definition attribute set relative to the decision attribute set based on the first number of elements and a total number of samples, wherein the total number of samples includes the number of samples in the motor rotor attribute sample set;

[0016] Based on the neighborhood relationship, determining a second lower approximate set of the decision attribute set with respect to the attribute set to be determined, and obtaining a second number of elements, wherein the attribute set to be determined includes a set consisting of the attribute to be determined and the defined attribute set, the second lower approximate set includes all motor rotor attribute samples that clearly belong to the decision attribute set under the attribute set to be determined, and the second number of elements includes the number of elements in the second lower approximate set;

[0017] determining a second dependency of the attribute set to be determined relative to the decision attribute set according to the second element quantity and the total number of samples;

[0018] The importance of the attribute to be determined to the decision result is determined according to the first dependency and the second dependency.

[0019] Optionally, performing attribute simplification on the attributes to be determined based on the importance to obtain an input attribute set includes:

[0020] Each of the attributes to be determined is arranged in descending order according to the importance, and the top m attributes to be determined in the arrangement result are used as the input attribute set.

[0021] Optionally, the input attributes in the input attribute set include at least one of an absolute value of a rotational speed, a square value of a stator current, a phase voltage, a cooling water temperature, a stator winding temperature, a water flow rate, and a shaft end torque.

[0022] Optionally, the training of a nonlinear autoregressive neural network based on the input data in the input attribute set to obtain a rotor temperature estimation model includes:

[0023] Taking the current moment and the input data of a first preset number of historical moments selected according to a preset time step as external input, an external input training set is constructed;

[0024] Using the output data of the previous moment and a second preset number of historical moments selected according to the preset time step as output feedback historical values, to construct an output feedback training set;

[0025] Training the nonlinear autoregressive neural network using the external input training set and the output feedback training set to obtain a training estimation result;

[0026] Measuring the prediction error of the training estimation result based on a preset loss function;

[0027] The connection weights and biases are adjusted based on the prediction error through a preset weight update algorithm until the prediction error meets a preset error requirement, thereby obtaining the rotor temperature estimation model.

[0028] In a second aspect, the present invention provides a method for estimating the temperature of a motor rotor, comprising:

[0029] Construct the input attribute set;

[0030] The rotor temperature estimation result is obtained by processing the input data in the input attribute set through the rotor temperature estimation model, wherein the input attributes in the input attribute set and the rotor temperature estimation model are obtained using the temperature estimation model training method as described in the first aspect.

[0031] In a second aspect, the present invention provides a temperature estimation model training device, comprising:

[0032] Importance module, used to determine the importance of the attribute to be judged to the decision result based on the neighborhood rough set algorithm;

[0033] An attribute reduction module, configured to reduce the attributes to be determined based on the importance to obtain an input attribute set;

[0034] The temperature estimation training module is used to train a nonlinear autoregressive neural network based on the input data in the input attribute set to obtain a rotor temperature estimation model.

[0035] In a third aspect, the present invention provides an electronic device comprising a memory and a processor;

[0036] The memory is used to store computer programs;

[0037] The processor is used to implement the temperature estimation model training method described in the first aspect or the motor rotor temperature estimation method described in the second aspect when executing the computer program.

[0038] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the temperature estimation model training method as described in the first aspect or the motor rotor temperature estimation method as described in the second aspect is implemented.

[0039] The beneficial effects of the temperature estimation model training method of the present invention are:

[0040] Using a neighborhood rough set algorithm, attributes with significant correlations with temperature changes are identified to distinguish primary from secondary influencing factors. By eliminating low-importance attributes and retaining high-importance attributes, the model input dimension can be reduced without affecting the accuracy of the estimated model. This reduces the amount of data to be processed during training and prevents redundant information from interfering with model parameters, thereby improving the model's convergence speed and generalization capabilities. A nonlinear autoregressive neural network is used to capture the complex nonlinear relationship and time series characteristics between input data and temperature. Because the input data has been simplified to remove redundancy, only primary influencing factors need to be processed during training and use, reducing computational resource consumption. The retained primary influencing factors provide sufficient information to maintain high prediction stability even with a small input dimension. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 Schematic diagram of the flow of a temperature estimation model training method according to an embodiment of the present invention;

[0042] Figure 2 This is a flow chart of a refinement of step S100 of the temperature estimation model training method according to an embodiment of the present invention;

[0043] Figure 3 This is a schematic diagram of a detailed process of step S120 of the temperature estimation model training method according to an embodiment of the present invention;

[0044] Figure 4 This is an example diagram of importance according to an embodiment of the present invention;

[0045] Figure 5 This is a schematic diagram of a detailed process of step S300 of the temperature estimation model training method according to an embodiment of the present invention;

[0046] Figure 6 This is a flow chart of a temperature estimation model training method according to an embodiment of the present invention;

[0047] Figure 7 2 is an exemplary diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0048] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, specific embodiments of the present invention are described in detail below with reference to the accompanying drawings. Although certain embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as being limited to the embodiments described herein. Instead, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the drawings and embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0049] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0050] The term "including" and its variations used in this document are open inclusions, that is, "including but not limited to"; the term "based on" means "based at least in part on"; the term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one other embodiment"; the term "some embodiments" means "at least some embodiments"; the term "optionally" means "optional embodiments". The relevant definitions of other terms will be given in the following description. It should be noted that the concepts of "first", "second", etc. mentioned in the present invention are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0051] It should be noted that the modifications of "one" and "multiple" mentioned in the present invention are illustrative rather than restrictive. Those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0052] The names of the messages or information exchanged between multiple devices in the embodiments of the present invention are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0053] In response to the problems existing in the above-mentioned related technologies, this embodiment provides a temperature estimation model training method, estimation method, device, equipment and medium.

[0054] like Figure 1 and Figure 6 As shown, an embodiment of the present invention provides a temperature estimation model training method, including:

[0055] Step S100: Determine the importance of the attribute to be judged to the decision result based on the neighborhood rough set algorithm.

[0056] With the development of sensor technology and the improvement of big data analysis capabilities, a foundation has been provided for data-driven temperature prediction methods. The rotor temperature can be estimated jointly through the real-time collection of multi-dimensional motor operation data.

[0057] Specifically, the attribute to be judged represents the motor operation data that can be collected in real time. Since the collectible motor operation data has multiple dimensions and there are many categories of factors that affect the motor rotor temperature, it is necessary to judge whether the attribute category of the collected motor operation data is a major influencing factor or a minor influencing factor, that is, to determine the importance of the attribute to be judged to the decision result.

[0058] In one embodiment, the absolute value of the speed represents the speed of the motor rotor during operation, which can directly characterize the working intensity of the motor. Therefore, the absolute value of the speed is a major influencing factor and is of high importance. However, the peak value of the open-circuit electromotive force of the motor has limited correlation with the working intensity of the motor. The temperature of the motor rotor cannot be accurately estimated based on the peak value of the open-circuit electromotive force, and is therefore considered a secondary influencing factor.

[0059] Step S200: Based on the importance, attribute reduction is performed on the attributes to be determined to obtain an input attribute set.

[0060] Using all influencing factors as model input will result in information redundancy, increase the dimensionality of the estimation model, reduce model training time, and affect the accuracy of the model estimation. We simplify the attributes to be determined, eliminating those that are less helpful in estimating temperature and retaining the main factors that are most helpful as the input attribute set.

[0061] In one embodiment, a preset number of attributes to be determined are retained to ensure that the accuracy of the estimation is minimal, thereby greatly reducing the model training time, thereby simplifying the input data required for the temperature estimation model and enabling the temperature estimation model to maintain a high estimation accuracy.

[0062] Step S300 : training a nonlinear autoregressive neural network based on the input data in the input attribute set to obtain a rotor temperature estimation model.

[0063] In this embodiment, the neighborhood rough set algorithm is used to identify attributes that have a significant correlation with temperature changes, which are used to distinguish between major influencing factors and minor influencing factors. By eliminating attributes of low importance and retaining attributes of high importance, the model input dimension can be reduced without affecting the accuracy of the estimation model, the amount of data processing during training is reduced, and the interference of redundant information on model parameters is avoided, thereby improving the model convergence speed and generalization ability. The complex nonlinear relationship and time series characteristics between input data and temperature are captured by a nonlinear autoregressive neural network. Since the input data has been simplified to remove redundancy, only the major influencing factors need to be processed during training and use, reducing computing resource consumption. The retained major influencing factors provide sufficient information to maintain a high prediction stability even with fewer input dimensions.

[0064] Alternatively, as Figure 2 As shown, the neighborhood rough set algorithm is based on which the importance of the attribute to be judged to the decision result is determined by:

[0065] Step S110 : constructing a neighborhood decision system including a condition attribute set and a decision attribute set based on the acquired motor rotor attribute sample set.

[0066] Step S120, determining the neighborhood relationship between the motor rotor attribute samples based on the Euclidean distance, and calculating the importance of each condition attribute to the decision attribute set through the neighborhood relationship, wherein the condition attribute includes the attribute to be judged, and the decision attribute set represents the set consisting of the decision results.

[0067] In one embodiment, a neighborhood decision system is constructed based on neighborhood rough set theory. Neighborhood rough sets are derived from classical rough sets. Classical rough sets can only process discretized numerical data, but discretization inevitably results in the loss of important information, affecting the processing results. Neighborhood rough sets can directly extract features from data attributes without discretization.

[0068] For Δ=R N ×R N →R, define (Ω, Δ) as a metric space, where Δ(x i ,x j ) is the distance function, which means any two samples x i with x j The distance function is calculated by using the Euclidean function. The distance function formula is expressed as:

[0069]

[0070] Where N represents the number of dimensions, R represents the set of real numbers, and a krepresents the kth feature dimension of the motor rotor attribute sample, f(x i ,a k ) represents the motor rotor attribute sample x i In feature a k The quantized value of f(x j ,a k ) represents the motor rotor attribute sample x j In feature a k quantized value of .

[0071] For any motor rotor attribute sample x in the domain U i , its neighborhood is expressed as:

[0072] δ(x i )={x|x∈U,Δ(x,x i )≤δ},

[0073] Among them, δ represents the neighborhood radius, δ≥0, δ(x i ) represents x i The δ neighborhood particles of

[0074] By calculating the distance between the particle and other particles in the domain U, we can determine whether it is a neighboring particle and then get x i Neighborhood particle family.

[0075] The constructed neighborhood decision system is expressed as:

[0076] NDT=(U,A=C∪D,D),

[0077] Where NDT represents the neighborhood decision system, U={x1,x2,…,x n} represents the domain, i.e., the motor rotor attribute sample set, A represents the set of all attributes, where D represents the decision attribute set and C represents the condition attribute set.

[0078] Decision attributes include a set of attributes used for classification or decision-making, representing the category or target outcome to which an object belongs. For example, in rotor temperature monitoring, decision attributes include "rotor temperature exceeds the standard" and "rotor temperature does not exceed the standard." Condition attribute sets include a set of attributes used to describe the characteristics of an object or sample. For example, in rotor temperature monitoring, condition attributes include "current," "voltage," and "rotor speed."

[0079] The importance of the attribute to be judged is determined by the inclusion relationship between the neighborhood of the sample and the decision attribute. For example, when a certain attribute to be judged is i The knowledge contained in the sample x i Neighborhood δ(x i )), we can determine the sample x that belongs to the decision attribute set iThe number of is used as the basis for judging the importance. Among the attributes to be judged, the sample x i The larger the number, the more important the attribute to be judged is for decision making.

[0080] Alternatively, as Figure 3 As shown, the neighborhood relationship between the motor rotor attribute samples is determined based on the Euclidean distance, and the importance of each condition attribute to the decision attribute set is calculated respectively through the neighborhood relationship, including:

[0081] Step S121: Based on the neighborhood relationship, determine the first lower approximate set of the decision attribute set with respect to the definition attribute set, and obtain a first number of elements, wherein the definition attribute set is composed of the conditional attributes, the first lower approximate set represents all the motor rotor attribute samples that clearly belong to the decision attribute set under the definition attribute set, and the first number of elements includes the number of elements in the first lower approximate set.

[0082] for and The upper and lower approximations of X with respect to the defined attribute set B are expressed as:

[0083]

[0084] Among them, δ B (x i ) represents all x i The samples whose distance under the defined attribute set B does not exceed δ, N B X They represent the upper approximate set and lower approximate set of the set X with respect to the defined attribute set.

[0085] Get the first lower approximation set of the decision attribute set with respect to the definition attribute set N B D .

[0086] Step S122 : determining a first dependency of the definition attribute set relative to the decision attribute set based on the first number of elements and the total number of samples, wherein the total number of samples includes the number of samples in the motor rotor attribute sample set.

[0087] The dependence formula is expressed as:

[0088]

[0089] Among them, γ represents the dependence, γ B (D) represents the dependency of the definition attribute set B on the decision attribute D, | N B D | represents the cardinality of the first lower approximation set of the decision attribute set with respect to the definition attribute set, and |U| represents the cardinality of the universe.

[0090] Step S123: Based on the neighborhood relationship, determine a second lower approximate set of the decision attribute set with respect to the attribute set to be judged, and obtain a second number of elements, wherein the attribute set to be judged includes a set consisting of the attribute to be judged and the defined attribute set, the second lower approximate set includes all the motor rotor attribute samples that clearly belong to the decision attribute set under the attribute set to be judged, and the second number of elements includes the number of elements in the second lower approximate set.

[0091] Step S124 : determining a second dependency of the attribute set to be determined with respect to the decision attribute set according to the second element quantity and the total number of samples.

[0092] Add the attributes to be judged to the defined attribute set to obtain the attribute set to be judged. The second dependency is expressed as:

[0093]

[0094] in, N B∪a D It represents the second lower approximation set of the decision attribute set with respect to the attribute set to be judged. Its calculation method is the same as the first lower approximation set. a represents the newly added attribute to be judged, and γ B∪a (D) represents the dependency of the attribute set to be judged B∪a on the decision attribute D, | N B∪a D | represents the cardinality of the second lower approximation set.

[0095] Step S125 : determining the importance of the attribute to be determined to the decision result according to the first dependency and the second dependency.

[0096] Importance is expressed as:

[0097] SIG(a,B,D)=γ B∪a (D)-γ B (D),

[0098] Among them, SIG(a,B,D) represents the importance of the newly added attribute a to be judged, that is, the change in the dependency after the newly added attribute a to be judged is added to the defined attribute set B. When the change is positive, the larger the value, the greater the importance of the attribute a to be judged; when the change is negative, it means that the attribute a to be judged is redundant data.

[0099] Optionally, performing attribute simplification on the attributes to be determined based on the importance to obtain an input attribute set includes:

[0100] Each of the attributes to be determined is arranged in descending order according to the importance, and the top m attributes to be determined in the arrangement result are used as the input attribute set.

[0101] In one embodiment, the value range of m is 5-10. In other embodiments, the value of m is determined by actual needs.

[0102] In one embodiment, the factors affecting the rotor temperature estimation are simplified, and the attribute importance is as follows: Figure 4 As shown in the figure, the first 7 attributes with the greatest importance are selected as the input attribute set to avoid data redundancy.

[0103] Optionally, performing attribute simplification on the attributes to be determined based on the importance to obtain an input attribute set includes:

[0104] The attributes to be judged that are higher than the importance threshold are used as input attributes to construct the input attribute set.

[0105] In this embodiment, the importance of different attributes to be determined varies to a certain extent. Setting an importance threshold can ensure the accuracy of rotor temperature estimation.

[0106] Optionally, the input attributes in the input attribute set include at least one of an absolute value of a rotational speed, a square value of a stator current, a phase voltage, a cooling water temperature, a stator winding temperature, a water flow rate, and a shaft end torque.

[0107] like Figure 4 As shown in the figure, the absolute value of speed Speed ​​Abs, the square value of stator current Stator Current Phase2, the phase voltage Voltage Phase, the cooling water temperature CooltT, the stator winding temperature Stator Winding Temp Average, the water flow rate FlowRate, and the shaft end torque Shaft Torque Abs are of high importance. The attributes ranked after 8th are less important than the 7th. To avoid data redundancy, the first 7 attributes are selected to construct the input attribute set.

[0108] Alternatively, as Figure 5 As shown, the training of a nonlinear autoregressive neural network based on the input data in the input attribute set to obtain a rotor temperature estimation model includes:

[0109] Step S310 , using the current moment and input data of a first preset number of historical moments selected according to a preset time step as external input to construct an external input training set.

[0110] Step S320 , using the output data of the previous moment and a second preset number of historical moments selected according to the preset time step as output feedback historical values ​​to construct an output feedback training set.

[0111] Step S330 , training the nonlinear autoregressive neural network using the external input training set and the output feedback training set to obtain a training estimation result.

[0112] Step S340: measuring the prediction error of the training estimation result based on a preset loss function.

[0113] Step S350 , adjusting the connection weights and biases based on the prediction error through a preset weight update algorithm until the prediction error meets a preset error requirement, thereby obtaining the rotor temperature estimation model.

[0114] In one embodiment, a NARX (Nonlinear Auto-Regressive model with Exogenous Inputs) neural network is used as a nonlinear autoregressive neural network to estimate the rotor temperature.

[0115] The mathematical expression of NARX neural network is expressed as:

[0116]

[0117] Among them, the external input history value represents the external input and its previous n x The input data of the historical moment of the step, the output feedback history value represents the previous n steps of the NARX neural network output y(t) y The output data at the historical moment of the step, e(t) represents the prediction error, and the data step length in the external input historical value and the output feedback historical value is the preset time step length.

[0118] The input layer of the NARX neural network can be expressed as:

[0119]

[0120] It is input from external historical values ​​x(t), x(t-1),…, x(tn x ) and output feedback history values ​​y(t-1),…,y(tn y )constitute.

[0121] The output d of the hth neuron in the hidden layer of the NARX neural network h (t) is expressed as:

[0122]

[0123] in, represents the weight of the input delay i to the hidden layer h, represents the weight of output delay j to hidden layer h, n x represents the number of input data at a historical moment, b h represents the hidden layer bias, f h (·) represents the hidden layer activation function, i=0,1,......,n x,j=0,1,......,n y .

[0124] In one embodiment, the hidden layer activation function includes a Sigmoid function or a tanh function. Sigmoid(f(x)=1 / (1+e -x )) or hyperbolic tangent (f(x)=tanh(x)).

[0125] The output layer of the NARX neural network obtains the final prediction by linearly combining the hidden layer activations:

[0126]

[0127] Among them, N h represents the total number of neurons in the hidden layer, represents the hidden layer d h (t) to the output weight, b o represents the output layer bias, f o (·) represents the output layer activation function.

[0128] In one embodiment, the output layer activation function includes a linear function (f(x)=x).

[0129] The NARX neural network can be viewed as a recursive structure that is unfolded in time and is implemented through back-propagation training with time delay. Minimizing the mean square error of the prediction error is the goal of the loss function, which is expressed as:

[0130]

[0131] in, represents the loss function, T represents the total number of steps, y true (t) represents the true target value at time step t, y pred (t) represents the model prediction value at time step t.

[0132] Update the weights by gradient descent:

[0133]

[0134] Among them, w represents the weight and η represents the learning rate, which is used to control the step size of each parameter update.

[0135] An embodiment of the present invention provides a method for estimating the temperature of a motor rotor, comprising:

[0136] Build the input attribute set.

[0137] The rotor temperature estimation result is obtained by processing the input data in the input attribute set through the rotor temperature estimation model, wherein the input attributes in the input attribute set and the rotor temperature estimation model are obtained using the temperature estimation model training method described above.

[0138] An embodiment of the present invention provides a temperature estimation model training device, comprising:

[0139] Importance module, used to determine the importance of the attribute to be judged to the decision result based on the neighborhood rough set algorithm;

[0140] An attribute reduction module, configured to reduce the attributes to be determined based on the importance to obtain an input attribute set;

[0141] The temperature estimation training module is used to train a nonlinear autoregressive neural network based on the input data in the input attribute set to obtain a rotor temperature estimation model.

[0142] like Figure 7 As shown, an electronic device 700 provided by an embodiment of the present invention includes a memory 710 and a processor 720; the memory 710 is used to store computer programs; the processor 720 is used to implement the motor rotor temperature estimation method or the motor rotor temperature estimation method as described above when executing the computer program.

[0143] In other words, an electronic device 700 includes a memory 710 and a processor 720 coupled to the memory 710; the memory 710 is configured to store a computer program; and the processor 720 is configured to perform the following operations when executing the computer program:

[0144] Based on the neighborhood rough set algorithm, determine the importance of the attribute to be judged to the decision result;

[0145] Based on the importance, performing attribute reduction on the attribute to be judged to obtain an input attribute set;

[0146] The rotor temperature is estimated by using a nonlinear autoregressive neural network on the input data in the input attribute set to obtain a rotor temperature estimation result.

[0147] Alternatively, construct an input attribute set;

[0148] The rotor temperature estimation result is obtained by processing the input data in the input attribute set through the rotor temperature estimation model, wherein the input attributes in the input attribute set and the rotor temperature estimation model are obtained using the temperature estimation model training method described above.

[0149] An embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the motor rotor temperature estimation method or the motor rotor temperature estimation method as described above is implemented.

[0150] In other words, a non-volatile computer-readable storage medium stores a computer program, which, when executed by a processor, causes the processor to perform the following operations:

[0151] Based on the neighborhood rough set algorithm, determine the importance of the attribute to be judged to the decision result;

[0152] Based on the importance, performing attribute reduction on the attribute to be judged to obtain an input attribute set;

[0153] The rotor temperature is estimated by using a nonlinear autoregressive neural network on the input data in the input attribute set to obtain a rotor temperature estimation result.

[0154] Alternatively, construct an input attribute set;

[0155] The rotor temperature estimation result is obtained by processing the input data in the input attribute set through the rotor temperature estimation model, wherein the input attributes in the input attribute set and the rotor temperature estimation model are obtained using the temperature estimation model training method described above.

[0156] An electronic device 700 that can serve as a server or client of the present invention will now be described, which is an example of a hardware device that can be applied to various aspects of the present invention. The electronic device 700 is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device 700 can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0157] The electronic device 700 includes a computing unit that can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) or a computer program loaded from a storage unit into a random access memory (RAM). In the RAM, various programs and data required for device operation can also be stored. The computing unit, ROM, and RAM are connected to each other via a bus. An input / output (I / O) interface is also connected to the bus.

[0158] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM). In this application, the units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiments of the present invention. In addition, the functional units in the various embodiments of the present invention can be integrated into a processing unit, or each unit can exist physically separately, or two or more units can be integrated into a single unit. The above-mentioned integrated units can be implemented in the form of hardware or software functional units.

[0159] Although the present invention is disclosed as above, the protection scope of the present invention is not limited thereto. Those skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention, and these changes and modifications will fall within the protection scope of the present invention.

Claims

1. A temperature estimation model training method, characterized in that: include: Based on the neighborhood rough set algorithm, determine the importance of the attribute to be judged to the decision result; Based on the importance, performing attribute reduction on the attribute to be judged to obtain an input attribute set; Based on the input data in the input attribute set, a nonlinear autoregressive neural network is trained to obtain a rotor temperature estimation model.

2. The temperature estimation model training method according to claim 1, characterized in that: The neighborhood rough set algorithm is based on which the importance of the attribute to be judged to the decision result is determined, which includes: Based on the acquired motor rotor attribute sample set, a neighborhood decision system including a condition attribute set and a decision attribute set is constructed; The neighborhood relationship between the motor rotor attribute samples is determined based on the Euclidean distance, and the importance of each condition attribute to the decision attribute set is calculated respectively through the neighborhood relationship, wherein the condition attribute includes the attribute to be judged, and the decision attribute set represents the set consisting of the decision results.

3. The temperature estimation model training method according to claim 2, characterized in that: Determining the neighborhood relationship between the motor rotor attribute samples based on the Euclidean distance, and calculating the importance of each condition attribute to the decision attribute set respectively through the neighborhood relationship includes: Based on the neighborhood relationship, determining a first lower approximation set of the decision attribute set with respect to the definition attribute set, and obtaining a first number of elements, wherein the definition attribute set is composed of the conditional attributes, the first lower approximation set represents all motor rotor attribute samples that clearly belong to the decision attribute set under the definition attribute set, and the first number of elements includes the number of elements in the first lower approximation set; determining a first dependency of the definition attribute set relative to the decision attribute set based on the first number of elements and a total number of samples, wherein the total number of samples includes the number of samples in the motor rotor attribute sample set; Based on the neighborhood relationship, determining a second lower approximate set of the decision attribute set with respect to the attribute set to be determined, and obtaining a second number of elements, wherein the attribute set to be determined includes a set consisting of the attribute to be determined and the defined attribute set, the second lower approximate set includes all motor rotor attribute samples that clearly belong to the decision attribute set under the attribute set to be determined, and the second number of elements includes the number of elements in the second lower approximate set; determining a second dependency of the attribute set to be determined relative to the decision attribute set according to the second element quantity and the total number of samples; The importance of the attribute to be determined to the decision result is determined according to the first dependency and the second dependency.

4. The temperature estimation model training method according to claim 1, characterized in that: The performing attribute simplification on the attributes to be determined based on the importance to obtain an input attribute set includes: Each of the attributes to be determined is arranged in descending order according to the importance, and the top m attributes to be determined in the arrangement result are used as the input attribute set.

5. The temperature estimation model training method according to claim 1, characterized in that: The input attributes in the input attribute set include at least one of an absolute value of a rotational speed, a square value of a stator current, a phase voltage, a cooling water temperature, a stator winding temperature, a water flow rate, and a shaft end torque.

6. The temperature estimation model training method according to any one of claims 1 to 5, characterized in that: The training of a nonlinear autoregressive neural network based on the input data in the input attribute set to obtain a rotor temperature estimation model includes: Taking the current moment and the input data of a first preset number of historical moments selected according to a preset time step as external input, an external input training set is constructed; Using the output data of the previous moment and a second preset number of historical moments selected according to the preset time step as output feedback historical values, to construct an output feedback training set; Training the nonlinear autoregressive neural network using the external input training set and the output feedback training set to obtain a training estimation result; Measuring the prediction error of the training estimation result based on a preset loss function; The connection weights and biases are adjusted based on the prediction error through a preset weight update algorithm until the prediction error meets a preset error requirement, thereby obtaining the rotor temperature estimation model.

7. A method for estimating the temperature of a motor rotor, characterized in that: include; Construct the input attribute set; The rotor temperature estimation result is obtained by processing the input data in the input attribute set through a rotor temperature estimation model, wherein the input attributes in the input attribute set and the rotor temperature estimation model are obtained using the temperature estimation model training method according to any one of claims 1 to 6.

8. A temperature estimation model training device, characterized in that: include: Importance module, used to determine the importance of the attribute to be judged to the decision result based on the neighborhood rough set algorithm; An attribute reduction module, configured to reduce the attributes to be determined based on the importance to obtain an input attribute set; The temperature estimation training module is used to train a nonlinear autoregressive neural network based on the input data in the input attribute set to obtain a rotor temperature estimation model.

9. An electronic device, characterized in that: including memory and processor; The memory is used to store computer programs; The processor is configured to implement the temperature estimation model training method according to any one of claims 1 to 6 or the motor rotor temperature estimation method according to claim 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that The storage medium stores a computer program, and when the computer program is executed by the processor, the temperature estimation model training method according to any one of claims 1 to 6 or the motor rotor temperature estimation method according to claim 7 is implemented.

Images