A method for constructing a multi-physics proxy model of a motor and related devices

Abstract

The present invention relates to the field of model construction technology, and discloses a motor multi-physics field proxy model construction method and related devices. The motor multi-physics field proxy model construction method includes: obtaining on-site operating condition data of a target motor; dividing the sample three-dimensional space into multiple key sampling intervals and multiple non-key sampling intervals based on the data density of the sample three-dimensional space, and setting a first sampling number for the key sampling interval and a second sampling number for the non-key sampling interval; using a Latin hypercube sampling method, sampling the on-site operating condition data in the key sampling interval and the non-key sampling interval according to the first sampling number and the second sampling number, respectively, to obtain a first working condition sample set; based on the first working condition sample set, training an initial proxy model until the average error meets the error requirement, thereby obtaining a motor multi-physics field target proxy model. The present invention can improve sample rationality and improve the accuracy of the proxy model.

Description

Technical Field

[0001] The present invention relates to the technical field of model construction, and in particular to a method for constructing a multi-physics field proxy model of a motor and related devices. Background Art

[0002] Multi-physics simulations require increasing computational complexity as the model complexity increases. In the field of digital twin intelligent operation and maintenance of hydropower and pumped storage unit motors, different methods are needed to construct real-time simulation models for different models, thereby achieving dimensionality reduction or data fitting of the numerical calculation order, and enabling real-time display of the multi-physics fields of each component through visualization. The surrogate model is an efficient mathematical approximation model that has gradually developed into a new computational method because it can significantly improve computational efficiency. The sampling method studies how to construct a fitting function with the least error using the fewest sample points, while being able to account for random errors in the data. The accuracy of the surrogate model depends on the distribution of the sample points.

[0003] Existing sample point collection often relies on a single sampling method. This often results in errors and uncertainties in the sample distribution. The proxy model constructed after the initial sampling may not meet on-site requirements, necessitating secondary sampling. This is often time-consuming and labor-intensive, and the sampling results often contain duplicate points, making deduplication time-consuming and labor-intensive. Furthermore, the actual operating conditions of power plants are often concentrated in a single area. The sample data obtained through uniform sampling is not universal, resulting in low accuracy and a poor fit for actual operating conditions in the trained proxy model. Summary of the Invention

[0004] In view of this, the present invention provides a method and related devices for constructing a motor multi-physics field proxy model to improve sample rationality and improve the accuracy of the proxy model.

[0005] In the first aspect, the present invention provides a method for constructing a multi-physics field proxy model of a motor, which includes: obtaining on-site operating condition data of a target motor; dividing the sample three-dimensional space into multiple key sampling intervals and multiple non-key sampling intervals based on the data density of the sample three-dimensional space, and setting a first sampling number for the key sampling interval and a second sampling number for the non-key sampling interval, wherein the data density of the key sampling interval is greater than the data density of the non-key sampling interval; using the Latin hypercube sampling method, sampling the on-site operating condition data in the key sampling interval and the non-key sampling interval according to the first sampling number and the second sampling number, respectively, to obtain a first condition sample set; performing model training on the initial proxy model based on the first condition sample set until the average error meets the error requirement, thereby obtaining a motor multi-physics field target proxy model.

[0006] In this implementation, the sampling intervals are divided according to the data density in the sample three-dimensional space. The intervals with high density are key sampling intervals, and the intervals with low density are non-key sampling intervals. Considering that the actual operating conditions of the power station are often concentrated in a certain area, that is, the key sampling intervals with high density, different sampling numbers are set for different sampling intervals. In actual application, the number of samples per unit area in the key sampling interval is greater than that in the non-key interval, which optimizes the sample space in the process of constructing the motor temperature field proxy model, making the sample data collected subsequently universal and able to fit the actual operating data situation, thereby improving the accuracy of the sample space point collection and the degree of fit with the field data. Furthermore, the initial proxy model is trained until the average error meets the error requirements, thereby improving the accuracy of the motor multi-physics field target proxy model.

[0007] In an optional embodiment, the on-site operating condition data includes inlet wind speed data, air temperature data and current data, and the sample three-dimensional space is divided into multiple key sampling intervals and multiple non-key sampling intervals based on the data density of the sample three-dimensional space, including: establishing a three-dimensional coordinate system with the inlet wind speed, air temperature and current to obtain the sample three-dimensional space, and dividing the sample three-dimensional space into multiple sample three-dimensional subspaces; determining point elements in the sample three-dimensional subspace based on the inlet wind speed data, air temperature data and current data; selecting a first preset number of sample three-dimensional subspaces with the largest number of point elements as key sampling intervals, and using other sample three-dimensional subspaces as non-key sampling intervals.

[0008] In this implementation, the coordinate system is constructed using the input features of the model, and the resulting sample three-dimensional space can accurately reflect the distribution of the data. By reflecting the spatial data density by the number of feature points, the difficulty of density calculation can be reduced and the efficiency and accuracy of interval division can be improved.

[0009] In an optional embodiment, the on-site operating condition data also includes the stator average temperature, and the first operating condition sample set includes the first operating condition sample characteristics and the first operating condition sample label of each sampling point. The on-site operating condition data is sampled in the key sampling interval and the non-key sampling interval according to the first sampling quantity and the second sampling quantity to obtain the first operating condition sample set, including: sampling in each sample three-dimensional subspace in the key sampling interval according to the first sampling quantity, and sampling in each sample three-dimensional subspace in the non-key sampling interval according to the second sampling quantity to obtain multiple sampling points; obtaining the inlet wind speed data, air temperature data and current data corresponding to each sampling point to obtain the first operating condition sample characteristics; when the sampling point has a corresponding stator average temperature, the stator average temperature is used as the first operating condition sample label corresponding to the sampling point; when the sampling point does not have a corresponding stator average temperature, the inlet wind speed data, air temperature data and current data are simulated based on the stator periodic boundary three-dimensional fluid-thermal coupling model to generate a simulated average temperature, and the simulated average temperature is used as the first operating condition sample label corresponding to the sampling point.

[0010] In this implementation, sampling is performed separately in each sample subspace. Each sample subspace corresponds to a smaller number of samples, which reduces sampling difficulty and ensures a more even distribution of samples. Furthermore, the currently available, accurate stator average temperature is preferentially used as the sample label. When no corresponding stator average temperature exists for a sampling point, simulation is performed using a data simulation model to fill in the sample set, which improves the accuracy of subsequent model training.

[0011] In an optional embodiment, the initial proxy model is trained based on the first working condition sample set until the average error meets the error requirement, and the motor multi-physics field target proxy model is obtained, which includes: performing sample quality evaluation on the first working condition sample set, and when the sample quality does not meet the quality requirement, expanding the sample of the first working condition sample set; training the initial proxy model based on the first working condition sample set, and using the first working condition sample set to judge the average error of the trained initial proxy model, and when the average error does not meet the error requirement, expanding the sample of the first working condition sample set until the average error meets the error requirement, and obtaining the motor multi-physics field target proxy model.

[0012] In this implementation, sample expansion is performed twice, before and after model training. Sample expansion is performed when sample quality does not meet quality requirements; sample expansion is performed when model accuracy is low. This application only performs sample sampling once, and subsequently only performs sample expansion. This can reduce the difficulty of secondary sampling, reduce the number of sample deduplication steps, and improve the efficiency of sample addition. Repeating the quality assessment and model evaluation steps until the requirements are met can further improve model accuracy.

[0013] In an optional embodiment, sample expansion of the first operating condition sample set includes: obtaining a second preset number of target sampling points with the worst quality or the largest error; arranging the target sampling points sequentially along the X-axis direction of the sample three-dimensional space, grouping adjacent target sampling points into target sampling point groups, and calculating the X-axis distance of the target sampling points within the target sampling point group; obtaining a third preset number of target sampling point groups with the largest X-axis distance as new sampling point groups, and setting new sample points between two target sampling points in the new sampling point group; using the inlet wind speed data, air temperature data, and current data corresponding to the new sample points as new operating condition sample features of the new sample points; performing data simulation on the inlet wind speed data, air temperature data, and current data of the new sample points based on the three-dimensional fluid-thermal coupling model of the stator periodic boundary to generate a new average temperature, and using the new average temperature as a new operating condition sample label for the new sampling point.

[0014] In an optional embodiment, setting a new sample point between two target sampling points in the new sampling point group includes: determining a target X-axis coordinate at 1 / 2 of the X-axis direction of the two target sampling points in the new sampling point group; and determining a point with the largest distance from other sampling points and other new sample points on the plane of the target X-axis coordinate as the new sample point.

[0015] In this implementation, adding new sampling points between distant sampling points can ensure the uniformity of the sampling point distribution and avoid duplicate sampling points, thus ensuring the quality of the newly added sampling points. At the same time, this application first determines a coordinate axis and then determines the final sampling point information, which can improve the efficiency of adding sampling points.

[0016] In the second aspect, the present invention provides a motor multi-physics field proxy model construction device, which includes: an acquisition module for acquiring the on-site operating condition data of the target motor; a division module for dividing the sample three-dimensional space into multiple key sampling intervals and multiple non-key sampling intervals based on the data density of the sample three-dimensional space, and setting a first sampling number for the key sampling interval and a second sampling number for the non-key sampling interval, and the data density of the key sampling interval is greater than the data density of the non-key sampling interval; a sampling module for using the Latin hypercube sampling method to sample in the key sampling interval and the non-key sampling interval according to the first sampling number and the second sampling number, respectively, to obtain a first working condition sample set; a training module for performing model training on the initial proxy model based on the first working condition sample set until the average error meets the error requirement, thereby obtaining the motor multi-physics field target proxy model.

[0017] In a third aspect, the present invention provides a computer device comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to execute the motor multi-physics agent model construction method of the first aspect or any corresponding embodiment thereof.

[0018] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the motor multi-physics field proxy model construction method of the first aspect or any corresponding embodiment thereof.

[0019] In a fifth aspect, the present invention provides a computer program product comprising computer instructions for causing a computer to execute the method for constructing a motor multi-physics proxy model according to the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0021] Figure 1 is a flow chart of a method for constructing a multi-physics proxy model of a motor according to an embodiment of the present invention;

[0022] Figure 2 is a first flow chart of another method for constructing a motor multi-physics proxy model according to an embodiment of the present invention;

[0023] Figure 3 is a second flow chart of another method for constructing a motor multi-physics proxy model according to an embodiment of the present invention;

[0024] Figure 4 is a structural block diagram of a device for constructing a multi-physics proxy model of a motor according to an embodiment of the present invention;

[0025] Figure 5 Schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0026] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the scope of protection of the present invention.

[0027] The multi-physics simulation of power equipment requires hardware conditions with a large number of cores, high frequency, and large memory. A single simulation takes a long time. The existing common sampling method for single sampling often has errors and uncertainties in the distribution of samples. The proxy model constructed after the initial sampling may not meet the needs of the site. Therefore, secondary sampling is required, which is often time-consuming and labor-intensive. The sampling results also contain repeated sampling points, and the deduplication work is also time-consuming and labor-intensive. At the same time, the disordered and random sample points in the sampling process will lead to low efficiency. The present application proposes a method for constructing a multi-physics proxy model of a motor. The sampling interval is divided according to the data density in the sample three-dimensional space. The high-density sampling interval is the key sampling interval, and the low-density sampling interval is the non-key sampling interval. Considering that the actual operating conditions of the power station are often concentrated in a certain area, that is, the high-density key sampling interval, different sampling numbers are set for different sampling intervals. In actual application, the number of samples per unit area in the key sampling interval is greater than that in the non-key interval. The sample space in the process of constructing the motor temperature field proxy model is optimized, so that the sample data obtained subsequently is universal and can fit the actual operating data, thereby improving the accuracy of the sample space point collection and the fit with the field data. Furthermore, the initial proxy model is trained until the average error meets the error requirement, thereby improving the accuracy of the motor multi-physics field target proxy model.

[0028] According to an embodiment of the present invention, an embodiment of a method for constructing a multi-physics field proxy model of a motor is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0029] In this embodiment, a method for constructing a multi-physics proxy model of a motor is provided. Figure 1 is a flow chart of a method for constructing a multi-physics proxy model of a motor according to an embodiment of the present invention. It should be noted that if there are substantially the same results, this embodiment does not use Figure 1 The process sequence shown is limited. Figure 1 As shown, the process includes the following steps:

[0030] Step S101: Acquire on-site operating condition data of a target motor.

[0031] Among them, the on-site operating condition data includes inlet wind speed data, air temperature data, stator average temperature data and stator current data.

[0032] Among them, according to different application scenarios, temperature data and current data can be obtained for the stator winding or stator core.

[0033] In a possible implementation, inlet wind speed data V, air temperature data T1, stator winding average temperature data T2, and stator winding current data of the target motor for one year are obtained.

[0034] In another possible implementation, the inlet wind speed data V, air temperature data T1, stator core average temperature data T3, and stator core current data of the target motor for one year are obtained.

[0035] Among them, the on-site operating condition data is usually recorded in a time series format, and each operating condition data includes a corresponding value and a timestamp.

[0036] Step S102 : dividing the sample 3D space into a plurality of key sampling intervals and a plurality of non-key sampling intervals based on the data density of the sample 3D space, setting a first sampling quantity for the key sampling intervals and a second sampling quantity for the non-key sampling intervals.

[0037] A three-dimensional sample space is constructed using inlet wind speed, air temperature, and stator current. The sample space is then divided into multiple key sampling intervals and multiple non-key sampling intervals based on the density of point elements corresponding to field operating condition data. The data density in key sampling intervals is greater than that in non-key sampling intervals.

[0038] Specifically, within the constructed sample 3D space, point features are identified using the inlet wind speed data, air temperature data, and stator current data corresponding to the same timestamp as a 3D coordinate. Areas with a high density of feature points are divided into key sampling intervals, while areas with a low density of feature points are divided into non-key sampling intervals.

[0039] For example, if the X-axis of the sample 3D space represents inlet wind speed, the Y-axis represents air temperature, and the Z-axis represents stator current, the inlet wind speed data at time t1 is v1, the air temperature data is t1, and the stator current data is i1. Then, the coordinates of the feature point corresponding to time t1 are determined to be (v1, t1, i1).

[0040] Furthermore, sampling data is set in key sampling intervals and non-key intervals according to the density of different point elements.

[0041] Specifically, a first unit sampling number is determined for a key sampling interval within a unit space, and the first sampling number is determined based on the spatial volume of the key sampling interval. Similarly, a second unit sampling number is determined for a non-key sampling interval within a unit space, and the second sampling number is determined based on the spatial volume of the non-key sampling interval. The first unit sampling number is greater than the second unit sampling number. That is, the greater the density of feature points, the greater the number of samples in the area.

[0042] Step S103 , using the Latin hypercube sampling method, sampling the on-site operating condition data in the key sampling interval and the non-key sampling interval according to the first sampling quantity and the second sampling quantity, respectively, to obtain a first operating condition sample set.

[0043] Specifically, a Latin hypercube sampling method is used to collect a first sampling number of sampling points from multiple point elements in a key sampling interval, and a second sampling number of sampling points is collected from multiple point elements in a non-key sampling interval.

[0044] In a possible implementation, sampling is performed once on the entire key sampling interval, and sampling is performed once on the entire non-key sampling interval.

[0045] In another possible implementation, the key sampling interval is divided into multiple unit spaces, and the Latin hypercube sampling method is used to sample in the unit space according to the first unit sampling number; similarly, the non-key sampling interval is divided into multiple unit spaces, and the Latin hypercube sampling method is used to sample in the unit space according to the second unit sampling number.

[0046] Furthermore, the stator average temperature data corresponding to the inlet wind speed data, air temperature data and stator current data at each sampling point is obtained, and the inlet wind speed data, air temperature data, stator average temperature data and stator current data at each sampling point are used as the first operating condition sample set.

[0047] Step S104 : performing model training on the initial proxy model based on the first operating condition sample set until the average error meets the error requirement, thereby obtaining a motor multi-physics field target proxy model.

[0048] Specifically, the first working condition sample set is divided into a training set and a validation set, and an initial proxy model of the motor multi-physics field is constructed. The inlet wind speed data, air temperature data, and stator current data of the training set are used as input features, and the stator average temperature data is used as a label feature to train the initial proxy model.

[0049] The inlet wind speed data, air temperature data, and stator current data of the validation set are input into the initial proxy model to obtain the predicted value of the proxy model, and the average error value between the stator average temperature data of the validation set and the predicted value of the proxy model is calculated. When the error value is greater than or equal to the error threshold, the model training of the initial proxy model is repeated. When the error value is less than the error threshold, the trained motor multi-physics field target proxy model is output.

[0050] The method for constructing a multi-physics proxy model of a motor provided in this embodiment is to divide the sampling intervals according to the data density in the sample three-dimensional space, with the high-density intervals being the key sampling intervals and the low-density intervals being the non-key sampling intervals. Considering that the actual working conditions of the power station are often concentrated in a certain area, namely the key sampling intervals with high density, different sampling numbers are set for different sampling intervals. In actual application, the number of samples per unit area in the key sampling intervals is greater than that in the non-key intervals, which optimizes the sample space in the process of constructing the motor temperature field proxy model, so that the sample data obtained by subsequent collection is universal and can fit the actual operating data situation, thereby improving the accuracy of the sample space point collection and the degree of fit with the field data. Furthermore, the initial proxy model is trained until the average error meets the error requirements, thereby improving the accuracy of the motor multi-physics target proxy model.

[0051] This embodiment provides a method for constructing a multi-physics proxy model of a motor. Figure 2 and Figure 3 , Figure 2 is a first flow chart of another method for constructing a motor multi-physics agent model according to an embodiment of the present invention, Figure 3 is a second flow chart of another method for constructing a multi-physics proxy model of a motor according to an embodiment of the present invention. It should be noted that if substantially the same results are achieved, this embodiment does not use Figure 2 The process sequence shown is limited. Figure 2 As shown, the process includes the following steps:

[0052] Step S201: Acquire on-site operating condition data of the target motor.

[0053] For details, please see Figure 1 Step S101 of the illustrated embodiment will not be described in detail here.

[0054] Step S202 : dividing the sample 3D space into a plurality of key sampling intervals and a plurality of non-key sampling intervals based on the data density of the sample 3D space, setting a first sampling quantity for the key sampling intervals and a second sampling quantity for the non-key sampling intervals.

[0055] Specifically, the above step S202 includes:

[0056] Step S2021 : establishing a three-dimensional coordinate system based on the inlet wind speed, air temperature, and current to obtain a sample three-dimensional space, and dividing the sample three-dimensional space into a plurality of sample three-dimensional subspaces.

[0057] Specifically, a three-dimensional coordinate system is established, with inlet wind speed, air temperature, and current as the X-axis, Y-axis, and Z-axis, respectively. A sample 3D space is constructed based on the data ranges of inlet wind speed, air temperature, and current. Inlet wind speed, air temperature, and current are not limited to specific axes and can be set based on actual conditions.

[0058] For example, a three-dimensional coordinate system is constructed with the inlet wind as the X-axis, the air temperature as the Y-axis, and the current as the Z-axis. The numerical ranges of the inlet wind speed data, air temperature data, and stator current data are [X0, X1], [Y0, Y1], and [Z0, Z1], respectively, to construct a three-dimensional sample space.

[0059] Furthermore, the sample three-dimensional space is divided into a plurality of D×D×D sample three-dimensional subspaces of equal size.

[0060] Step S2022 : determining point elements in the sample 3D subspace based on the inlet wind speed data, the air temperature data, and the current data.

[0061] Specifically, in the divided sample three-dimensional subspace, a point element is determined according to the inlet wind speed data, air temperature data and stator current data corresponding to the same timestamp, and the coordinates of the point element are the corresponding three data values.

[0062] For details, please see Figure 1 The example of step S102 of the illustrated embodiment will not be described in detail here.

[0063] Step S2023 : Selecting a first preset number of sample stereo subspaces with the largest number of point elements as key sampling intervals, and using other sample stereo subspaces as non-key sampling intervals.

[0064] Based on the density of point elements corresponding to the on-site operating condition data in the sample three-dimensional space, the sample three-dimensional space is divided into multiple key sampling intervals and multiple non-key sampling intervals. The data density of the key sampling intervals is greater than that of the non-key sampling intervals.

[0065] In this implementation method, the number of point elements is used as a parameter to measure density. The number of point elements in each sample stereo subspace is obtained, and the sample stereo subspaces are sorted according to the number of point elements. The sample stereo subspaces with the largest number of point elements, that is, the first preset number of sample stereo subspaces with the largest number of point elements, are taken as key sampling intervals, and the other sample stereo subspaces are taken as non-key sampling intervals.

[0066] For example, select D2 The sample stereo subspace is used as the key sampling interval.

[0067] Step S2024: setting a first sampling quantity for the key sampling interval and a second sampling quantity for the non-key sampling interval.

[0068] In one possible implementation, a first unit sampling number of a key sampling interval within a sample stereo subspace is determined, and the first sampling number is determined based on the number of sample stereo subspaces contained in the key sampling interval. Similarly, a second unit sampling number of a non-key sampling interval within the sample stereo subspace is determined, and the second sampling number is determined based on the number of sample stereo subspaces contained in the non-key sampling interval. The first unit sampling number is greater than the second unit sampling number.

[0069] In another possible implementation, a first unit sampling number is determined as m1 for the key sampling interval, and the first sampling number is evenly divided into each 3D sample subspace; a second unit sampling number is determined as m2 for the non-key sampling interval, and the second sampling number is evenly divided into each 3D sample subspace. Here, m1 is greater than m2.

[0070] In another possible implementation, the second unit sampling number of the non-important sampling interval is determined to be M:

[0071] .

[0072] Wherein, N is the number of variables. In this implementation, the features include inlet wind speed data, air temperature data, and stator current data. Therefore, N is 3.

[0073] Determine the first unit sampling quantity of the key sampling interval as: .

[0074] Among them, the sampling density of the key sampling interval is: The number of samples in each corresponding sample stereo subspace is .

[0075] in, is the total spatial volume of the key sampling interval, is the spatial volume of the sample stereo subspace.

[0076] in, .

[0077] Determine the number of sample points for initial sampling .

[0078] In this implementation, the coordinate system is constructed using the input features of the model, and the resulting sample three-dimensional space can accurately reflect the distribution of the data. By reflecting the spatial data density by the number of feature points, the difficulty of density calculation can be reduced and the efficiency and accuracy of interval division can be improved.

[0079] In step S203 , a Latin hypercube sampling method is used to perform sampling in a key sampling interval and a non-key sampling interval according to a first sampling quantity and a second sampling quantity, respectively, to obtain a first working condition sample set.

[0080] In this implementation, the key sampling interval is divided into multiple unit spaces, and the Latin hypercube sampling method is used to sample in the unit space according to the first unit sampling number; similarly, the non-key sampling interval is divided into multiple unit spaces, and the Latin hypercube sampling method is used to sample in the unit space according to the second unit sampling number.

[0081] Specifically, the above step S203 includes:

[0082] Step S2031 : Sampling is performed in each sample stereo subspace of the key sampling interval according to a first sampling quantity, and sampling is performed in each sample stereo subspace of the non-key sampling interval according to a second sampling quantity to obtain a plurality of sampling points.

[0083] Among them, the maximum and minimum distance Latin hypercube is used for sampling, and the maximum and minimum criteria are expressed as:

[0084] .

[0085] in, is the distance between two sampling points in the sample stereo subspace.

[0086] in, . t=1 or t=2.

[0087] According to the above-mentioned Latin hypercube sampling method, sampling is performed in the key sampling interval and the non-key sampling interval to obtain multiple sampling points.

[0088] Step S2032: Obtain the first operating condition sample features and the first operating condition sample labels of each sampling point.

[0089] Specifically, the inlet wind speed data, air temperature data, and stator current data corresponding to each sampling point are obtained as the first operating condition sample features of the sampling point.

[0090] If the on-site operating condition data of the target motor contains stator average temperature data corresponding to the inlet wind speed data, air temperature data and stator current data of the sampling point, the stator average temperature data corresponding to the inlet wind speed data, air temperature data and stator current data at each sampling point is obtained as the first operating condition sample label corresponding to the sampling point.

[0091] If the stator average temperature data corresponding to the inlet wind speed data, air temperature data and stator current data of the sampling point does not exist in the field operating condition data of the target motor, the inlet wind speed data, air temperature data and current data are simulated based on the three-dimensional fluid-thermal coupling model of the stator periodic boundary to generate a simulated average temperature, and the simulated average temperature is used as the first operating condition sample label corresponding to the sampling point.

[0092] For example, for each sample point m, based on the periodic boundary three-dimensional fluid-thermal coupling model of the stator winding and the stator core, the numerical simulation of the operating point is realized using commercial software to generate the average temperature of the stator winding and the stator core under different operating conditions, and generate the corresponding data set .

[0093] .

[0094] Among them, P1 is the stator winding heat source, and P2 is the stator core heat source.

[0095] The above method is used to generate the stator average temperature. When simulating the stator winding, the stator winding average temperature is used as the first working condition sample label. When simulating the stator core, the stator core average temperature is used as the first working condition sample label.

[0096] In this implementation, sampling is performed separately in each sample subspace. Each sample subspace corresponds to a smaller number of samples, which reduces sampling difficulty and ensures a more even distribution of samples. Furthermore, the currently available, accurate stator average temperature is preferentially used as the sample label. When no corresponding stator average temperature exists for a sampling point, simulation is performed using a data simulation model to fill in the sample set, which improves the accuracy of subsequent model training.

[0097] Step S204 : performing model training on the initial proxy model based on the first operating condition sample set until the average error meets the error requirement, thereby obtaining a motor multi-physics field target proxy model.

[0098] Specifically, the above step S204 includes:

[0099] Step S2041 : performing sample quality evaluation on the first working condition sample set. When the sample quality does not meet the quality requirement, performing sample expansion on the first working condition sample set.

[0100] Specifically, a sampling evaluation is performed on the first working condition sample set. The sampling result sample set B is:

[0101] .

[0102] The sample quality of the first working condition sample set is evaluated using the following formula:

[0103] .

[0104] in, The smaller the value, the better the uniformity of the sampling results. When the value is greater than the threshold, it means that the sample quality does not meet the quality requirements, and the first working condition sample set is expanded until the sample quality meets the quality requirements.

[0105] Step S2042: training the initial proxy model based on the first working condition sample set.

[0106] Specifically, the above step S2042 specifically includes:

[0107] Step a1: Build a processing agent model.

[0108] Specifically, the sampled Kriging surrogate model is constructed, and the basic function of the initial surrogate model is:

[0109] .

[0110] Where F represents the trend and z is the residual. F is usually modeled by a certain linear function:

[0111] .

[0112] Where f is the basis function and β is the fitting coefficient.

[0113] Step a2: train the initial proxy model.

[0114] The first working condition sample set is divided into a training set and a validation set, and an initial proxy model of the motor multi-physics field is constructed. The inlet wind speed data, air temperature data, and stator current data of the training set are used as input features, and the stator average temperature data is used as the label feature to train the initial proxy model.

[0115] Step S2043: Use the first working condition sample set to perform average error judgment on the trained initial proxy model. When the average error does not meet the error requirement, expand the first working condition sample set until the average error meets the error requirement, thereby obtaining the motor multi-physics field target proxy model.

[0116] For the n sample points of the validation set, the inlet wind speed data, air temperature data and stator current data are input into the initial proxy model to obtain the predicted value of the proxy model , and calculate the stator average temperature data T of the validation set n The predicted value of the surrogate model The average error value.

[0117] Among them, the stator average temperature data T i The predicted value of the surrogate model The error value is:

[0118] .

[0119] Stator average temperature data T n The predicted value of the surrogate model The average error value is:

[0120] .

[0121] When the error value is greater than or equal to the error threshold, the initial proxy model is repeatedly trained. When the error value is less than the error threshold, the trained motor multi-physics target proxy model is output.

[0122] In one implementation, the error threshold is 10%.

[0123] In the above step S2041 and the above step S2043, the method for expanding the sample set of the first working condition is the same, and the sample set is expanded by increasing the sampling points.

[0124] Specifically, the method for expanding the sample set of the first working condition includes:

[0125] Step b1: obtaining a second preset number of target sampling points with the worst quality or the largest error.

[0126] Get the t target sampling points with the worst quality or the largest error.

[0127] Step b2: sequentially arrange the target sampling points along the X-axis direction of the sample stereo space, group adjacent target sampling points into target sampling point groups, and calculate the X-axis distances of the target sampling points in the target sampling point groups.

[0128] Arrange the t target sampling points in order of size along the X-axis direction of the sample three-dimensional space to obtain the sample set X new ={X1,…,X t}. The adjacent target sampling points are taken as the target sampling point group {(X1,X2),…,(X t-1 ,X t)}. Further, the X-axis distance between two target sampling points in the target sampling point group is calculated, and the distance set is obtained as D={d1,…,d t+1}.

[0129] Step b3: obtaining a target sampling point group with a third preset number of the largest X-axis distance as a newly added sampling point group, and setting a new sampling point between two target sampling points in the newly added sampling point group.

[0130] Select the L target sampling point groups with the largest X-axis distance as the new sampling point groups, and add new sampling points within the interval where the new sampling point groups are located. Specifically, place the new sampling point {t+1,…,t+L} between two sampling points in the new sampling point group.

[0131] Specifically, the target X-axis coordinate is determined at 1 / 2 of the X-axis direction of the two target sampling points in the newly added sampling point group, and the point with the largest distance from other sampling points and other newly added sampling points is determined on the plane of the target X-axis coordinate as the newly added sample point.

[0132] Step b4: Use the inlet wind speed data, air temperature data, and current data corresponding to the newly added sample point as the newly added working condition sample features of the newly added sample point.

[0133] Step b5: Based on the three-dimensional fluid-thermal coupling model of the stator periodic boundary, data simulation is performed on the inlet wind speed data, air temperature data and current data of the newly added sample points to generate a new average temperature, and the new average temperature is used as the new working condition sample label of the new sampling point.

[0134] Specifically, the data simulation method of step S2032 is used to generate the newly added average temperature, which will not be described in detail here.

[0135] Since the multi-physics field simulation of power equipment requires hardware conditions with many cores, high frequency and large memory, a single simulation takes a long time. This application achieves targeted optimization and improvement by optimizing the sample space in the process of constructing the motor temperature field proxy model, avoiding the low efficiency problem of disorderly and randomly added sample points, accelerating the improvement rate and efficiency of the proxy model, and reducing the computer hardware storage requirements and the time for calculating disorderly and randomly added sample points during the proxy model training process.

[0136] The motor multi-physics proxy model construction method provided in this embodiment performs sample expansion twice, before and after model training. When the sample quality does not meet the quality requirements, sample expansion is performed; when the model accuracy is low, sample expansion is performed again. This application only performs sample sampling once, and then only performs sample expansion, which can reduce the difficulty of secondary sampling, reduce the number of sample deduplication steps, and improve the efficiency of sample addition.

[0137] Specifically, adding new sampling points between distant sampling points can ensure uniform distribution of the sampling points and avoid duplicate sampling points, thus ensuring the quality of the newly added sampling points. Furthermore, this application first determines a coordinate axis and then determines the final sampling point information, which can improve the efficiency of adding new sampling points. Repeating the quality assessment and model evaluation steps until the requirements are met can further improve the accuracy of the model.

[0138] This embodiment also provides a device for constructing a multi-physics proxy model for a motor, which is used to implement the above-mentioned embodiments and preferred implementations. Details already described will not be repeated here. As used below, the term "module" may refer to a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation using hardware, or a combination of software and hardware, is also possible and contemplated.

[0139] This embodiment provides a device for constructing a multi-physics proxy model of a motor. Figure 4 is a structural block diagram of a device for constructing a multi-physics agent model of a motor according to an embodiment of the present invention. Figure 4 As shown, the motor multi-physics field agent model building device includes:

[0140] The acquisition module 401 is used to acquire the on-site operating condition data of the target motor.

[0141] The division module 402 is used to divide the sample three-dimensional space into multiple key sampling intervals and multiple non-key sampling intervals based on the data density of the sample three-dimensional space, and set a first sampling number for the key sampling interval and a second sampling number for the non-key sampling interval, wherein the data density of the key sampling interval is greater than the data density of the non-key sampling interval.

[0142] The sampling module 403 is configured to perform sampling in a key sampling interval and a non-key sampling interval according to a first sampling quantity and a second sampling quantity using a Latin hypercube sampling method to obtain a first operating condition sample set.

[0143] The training module 404 is used to perform model training on the initial proxy model based on the first operating condition sample set until the average error meets the error requirement, thereby obtaining a motor multi-physics field target proxy model.

[0144] In some optional implementations, the partitioning module 402 includes:

[0145] The division unit is used to establish a three-dimensional coordinate system based on the inlet wind speed, air temperature and current to obtain a sample three-dimensional space, and divide the sample three-dimensional space into a plurality of sample three-dimensional subspaces.

[0146] The determining unit is used to determine point elements in the sample stereo subspace based on the inlet wind speed data, the air temperature data and the current data.

[0147] The allocating unit is configured to select a first preset number of sample stereo subspaces having the largest number of point elements as key sampling intervals, and use other sample stereo subspaces as non-key sampling intervals.

[0148] In some optional implementations, the sampling module 403 includes:

[0149] The sampling unit is used to perform sampling in each sample stereo subspace of the key sampling interval according to the first sampling quantity, and perform sampling in each sample stereo subspace of the non-key sampling interval according to the second sampling quantity, to obtain multiple sampling points.

[0150] The feature selection unit is used to obtain the inlet wind speed data, air temperature data and current data corresponding to each sampling point to obtain the first working condition sample feature.

[0151] The first label selection unit is configured to use the stator average temperature as the first operating condition sample label corresponding to the sampling point when the stator average temperature corresponds to the sampling point.

[0152] The second label selection unit, when there is no corresponding stator average temperature at the sampling point, performs data simulation on the inlet wind speed data, air temperature data and current data based on the stator periodic boundary three-dimensional fluid-thermal coupling model, generates a simulated average temperature, and uses the simulated average temperature as the first working condition sample label corresponding to the sampling point.

[0153] In some optional implementations, the training module 404 includes:

[0154] The expansion unit is used to perform sample quality evaluation on the first working condition sample set, and when the sample quality does not meet the quality requirements, expand the samples of the first working condition sample set.

[0155] A training unit is used to train the initial proxy model based on the first working condition sample set, and use the first working condition sample set to judge the average error of the trained initial proxy model. When the average error does not meet the error requirement, the first working condition sample set is expanded until the average error meets the error requirement, thereby obtaining the motor multi-physics field target proxy model.

[0156] In some optional embodiments, the expansion unit includes:

[0157] The acquisition subunit is configured to acquire a second preset number of target sampling points with the worst quality or the largest error.

[0158] The calculation subunit is used to arrange the target sampling points in sequence along the X-axis direction of the sample three-dimensional space, group adjacent target sampling points into target sampling point groups, and calculate the X-axis distances of the target sampling points in the target sampling point groups.

[0159] The setting subunit is configured to obtain a target sampling point group with a third preset number of the largest X-axis distances as a new sampling point group, and set a new sampling point between two target sampling points in the new sampling point group.

[0160] A new feature subunit is added, which is used to use the inlet wind speed data, air temperature data and current data corresponding to the newly added sample point as the new working condition sample features of the newly added sample point.

[0161] A new label subunit is added to simulate the inlet wind speed data, air temperature data and current data of the newly added sample points based on the three-dimensional fluid-thermal coupling model of the stator periodic boundary, generate a new average temperature, and use the new average temperature as the new working condition sample label for the new sampling point.

[0162] In some optional embodiments, setting the subunit includes:

[0163] The first determining subunit is configured to determine a target X-axis coordinate at a position half of the X-axis direction of two target sampling points in the newly added sampling point group.

[0164] The second determining subunit is configured to determine, on the plane of the target X-axis coordinate, a point with the largest distance from other sampling points and other newly added sample points as a newly added sample point.

[0165] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0166] The motor multi-physics field proxy model construction device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0167] The embodiment of the present invention also provides a computer device having the above Figure 4 The setup for building a multiphysics proxy model of an electric motor is shown.

[0168] See also Figure 5 , Figure 5 is a structural diagram of a computer device provided by an optional embodiment of the present invention, such as Figure 5As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components utilize different buses to communicate with each other and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of a GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories. Equally, multiple computer devices can be connected, and each device provides part of the necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 5 A processor 10 is taken as an example.

[0169] The processor 10 may be a central processing unit, a network processor, or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic, or any combination thereof.

[0170] The memory 20 stores instructions that can be executed by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0171] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created based on the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0172] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0173] The computer device further includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 may be connected via a bus or other means. Figure 5 The bus connection is taken as an example.

[0174] The input device 30 can receive input digital or character information and generate key signal input related to user settings and function control of the computer device. Examples include a touch screen, keypad, mouse, trackpad, touchpad, pointing stick, one or more mouse buttons, trackball, joystick, etc. The output device 40 may include a display device, auxiliary lighting devices (e.g., LEDs), and tactile feedback devices (e.g., vibration motors). Such display devices include, but are not limited to, liquid crystal displays, light emitting diodes, monitors, and plasma displays. In some optional embodiments, the display device may be a touch screen.

[0175] The embodiment of the present invention also provides a computer-readable storage medium. The above-mentioned method according to the embodiment of the present invention can be implemented in hardware, firmware, or implemented as a computer code that can be recorded in a storage medium, or implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and downloaded through a network and will be stored in a local storage medium, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state drive, etc.; further, the storage medium can also include a combination of the above-mentioned types of memory. It can be understood that a computer, a processor, a microprocessor controller or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor or hardware, the method shown in the above embodiment is implemented.

[0176] A portion of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the form in which the computer program instruction exists in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc. Accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium that can be accessed by the computer.

[0177] Although the embodiments of the present invention have been described with reference to the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention. Such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for constructing a multi-physics proxy model of a motor, characterized in that: The method comprises: Obtain on-site operating condition data of the target motor; Dividing the sample three-dimensional space into a plurality of key sampling intervals and a plurality of non-key sampling intervals based on the data density of the sample three-dimensional space, setting a first sampling number for the key sampling intervals and a second sampling number for the non-key sampling intervals, wherein the data density of the key sampling intervals is greater than the data density of the non-key sampling intervals; Using a Latin hypercube sampling method, sampling the on-site operating condition data in the key sampling interval and the non-key sampling interval according to the first sampling quantity and the second sampling quantity, respectively, to obtain a first operating condition sample set; Performing sampling evaluation on the first operating condition sample set and calculating a sample quality value; when the sample quality value is greater than a threshold, performing sample expansion on the first operating condition sample set; a smaller sample quality value indicates better uniformity of the first operating condition sample set; training an initial proxy model based on the first operating condition sample set until an average error meets an error requirement, thereby obtaining a motor multi-physics field target proxy model; The expanding the first operating condition sample set includes: Obtaining a second preset number of target sampling points with the worst quality or the largest error; The target sampling points are sequentially arranged along the X-axis direction of the sample three-dimensional space, adjacent target sampling points are grouped as target sampling point groups, and the X-axis distances of the target sampling points in the target sampling point groups are calculated; Acquire the target sampling point group with the third preset number of the largest X-axis distances as a newly added sampling point group, and set a new sampling point between two target sampling points in the newly added sampling point group; The inlet wind speed data, air temperature data and current data corresponding to the newly added sample point are used as the newly added working condition sample features of the newly added sample point; Based on the three-dimensional fluid-thermal coupling model of the stator periodic boundary, data simulation is performed on the inlet wind speed data, the air temperature data and the current data of the newly added sample point to generate a newly added average temperature, and the newly added average temperature is used as the new operating condition sample label of the newly added sampling point.

2. The method for constructing a multi-physics proxy model of a motor according to claim 1, wherein: The on-site operating condition data includes inlet wind speed data, air temperature data, and current data. The data density of the sample three-dimensional space is used to divide the sample three-dimensional space into multiple key sampling intervals and multiple non-key sampling intervals, including: Establishing a three-dimensional coordinate system based on inlet wind speed, air temperature, and current to obtain the sample three-dimensional space, and dividing the sample three-dimensional space into a plurality of sample three-dimensional subspaces; determining a point element in the sample stereo subspace based on the inlet wind speed data, the air temperature data, and the current data; The first preset number of sample stereo subspaces with the largest number of point elements are selected as the key sampling intervals, and the other sample stereo subspaces are used as the non-key sampling intervals.

3. The method for constructing a multi-physics proxy model of a motor according to claim 2, wherein: The on-site operating condition data further includes an average stator temperature. The first operating condition sample set includes a first operating condition sample feature and a first operating condition sample label of each sampling point. The on-site operating condition data is sampled in the key sampling interval and the non-key sampling interval according to the first sampling quantity and the second sampling quantity to obtain the first operating condition sample set, including: Sampling is performed in each sample stereo subspace of the key sampling interval according to the first sampling quantity, and sampling is performed in each sample stereo subspace of the non-key sampling interval according to the second sampling quantity, to obtain a plurality of sampling points; Acquire the inlet wind speed data, the air temperature data, and the current data corresponding to each sampling point to obtain the first operating condition sample characteristics; When the sampling point has the corresponding stator average temperature, the stator average temperature is used as the first operating condition sample label corresponding to the sampling point; When the stator average temperature corresponding to the sampling point does not exist, data simulation is performed on the inlet wind speed data, the air temperature data, and the current data based on the stator periodic boundary three-dimensional fluid-thermal coupling model to generate a simulated average temperature, and the simulated average temperature is used as the first operating condition sample label corresponding to the sampling point.

4. The method for constructing a multi-physics proxy model of a motor according to claim 1, wherein: The initial proxy model is trained based on the first operating condition sample set until the average error meets the error requirement to obtain the motor multi-physics field target proxy model, including: An initial proxy model is trained based on the first operating condition sample set, and the first operating condition sample set is used to perform average error judgment on the trained initial proxy model. When the average error does not meet the error requirement, the first operating condition sample set is expanded until the average error meets the error requirement, thereby obtaining a motor multi-physics field target proxy model.

5. The method for constructing a multi-physics proxy model of a motor according to claim 4, wherein: The step of setting a new sampling point between two target sampling points in the new sampling point group includes: Determine a target X-axis coordinate at half of the X-axis direction of the two target sampling points in the newly added sampling point group; A point with the largest distance from other sampling points and other newly added sample points is determined on the plane of the target X-axis coordinate as the newly added sample point.

6. A motor multi-physics field proxy model construction device, characterized in that: The device comprises: An acquisition module is used to obtain the on-site operating condition data of the target motor; a dividing module, configured to divide the sample three-dimensional space into a plurality of key sampling intervals and a plurality of non-key sampling intervals based on the data density of the sample three-dimensional space, and set a first sampling number for the key sampling intervals and a second sampling number for the non-key sampling intervals, wherein the data density of the key sampling intervals is greater than the data density of the non-key sampling intervals; The sampling module is used to use the Latin hypercube sampling method to sample the first sampling number and the second sampling number in the key sampling interval and the non-key sampling interval respectively to obtain a first working condition sample set; perform sampling evaluation on the first working condition sample set, calculate the sample quality value, and when the sample quality value is greater than a threshold, perform sample expansion on the first working condition sample set; the smaller the sample quality value, the better the uniformity of the first working condition sample set; the sample expansion of the first working condition sample set includes: obtaining a second preset number of target sampling points with the worst quality or the largest error; sequentially arranging the target sampling points along the X-axis direction of the sample three-dimensional space, and placing adjacent target sampling points at the same time. The sampling points are the target sampling point group, and the X-axis distance of the target sampling points in the target sampling point group is calculated; the target sampling point group with the largest X-axis distance of the third preset number is obtained as the newly added sampling point group, and a new sample point is set between two target sampling points in the newly added sampling point group; the inlet wind speed data, air temperature data and current data corresponding to the newly added sample point are used as the new working condition sample features of the newly added sample point; based on the three-dimensional fluid-thermal coupling model of the stator periodic boundary, the inlet wind speed data, the air temperature data and the current data of the newly added sample point are simulated to generate a new average temperature, and the new average temperature is used as the new working condition sample label of the newly added sampling point; A training module is used to perform model training on the initial proxy model based on the first operating condition sample set until the average error meets the error requirement, thereby obtaining a motor multi-physics field target proxy model.

7. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the motor multi-physics field proxy model construction method according to any one of claims 1 to 5 by executing the computer instructions.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the motor multi-physics field proxy model construction method according to any one of claims 1 to 5.

9. A computer program product, characterized in that The method comprises computer instructions for causing a computer to execute the method for constructing a multi-physics agent model of a motor according to any one of claims 1 to 5.

Images