A core loss prediction method

Abstract

The application discloses a magnetic core loss prediction method in the technical field of electromagnetic material prediction, comprising the following steps: obtaining the environment parameters and one-dimensional magnetic flux density time sequence signal of a to-be-tested magnetic core, performing prediction based on a pre-trained magnetic core loss prediction model, and obtaining the prediction value of the magnetic core loss; the data processing method of the magnetic core loss prediction model comprises the following steps: converting the environment parameters into a tensor form to obtain a parameter feature vector; converting the one-dimensional magnetic flux density time sequence signal into a two-dimensional time-frequency spectrum graph, performing feature extraction on the two-dimensional time-frequency spectrum graph, and obtaining an image feature vector; splicing the parameter feature vector and the image feature vector to obtain a comprehensive feature representation; and performing full connection layer processing on the comprehensive feature representation to obtain the prediction value of the magnetic core loss; wherein the environment parameters comprise continuous variables and discrete variables. The application solves the problem that the prior art is difficult to accurately predict the magnetic core loss under high frequency or non-sinusoidal excitation.

Description

Technical Field

[0001] This invention relates to a method for predicting magnetic core loss, belonging to the field of electromagnetic material prediction technology. Background Technology

[0002] In the field of power electronic equipment, core loss, as a key parameter for evaluating equipment efficiency and performance, has always received widespread attention within the industry. Traditional core loss prediction techniques are mainly based on experimental measurements and theoretical calculations. Experimental measurement methods obtain core loss data under different conditions through actual testing. While the results are accurate, the measurement cycle is long, the cost is high, and it is not applicable to all application scenarios. On the other hand, theoretical calculation methods estimate core loss by establishing mathematical models based on electromagnetic principles. Although this method is lower in cost, the calculation process is complex, and it often fails to achieve high accuracy in predicting core loss under complex operating conditions. In recent years, with the rapid development of big data and artificial intelligence technologies, machine learning algorithms have been introduced into the field of core loss prediction, providing new ideas for solving this problem.

[0003] While machine learning algorithms have demonstrated significant potential in predicting magnetic core losses, existing methods still suffer from several notable drawbacks. Specifically, current prediction models often consider only single-mode data, such as predictions based solely on waveform or material property data. This single-mode data processing approach ignores the fact that core losses are influenced by multiple factors, resulting in limited prediction accuracy. For instance, relying solely on waveform data may fail to fully reflect the loss characteristics of core materials under different frequencies and temperatures; conversely, relying solely on material property data may ignore the impact of waveform variations on losses under actual operating conditions. Therefore, under complex operating conditions such as high-frequency or non-sinusoidal excitation, existing prediction methods often struggle to accurately predict core losses, thus limiting the performance optimization and efficiency improvement of power electronic devices. To address this issue, the industry urgently needs a core loss prediction method that can comprehensively consider multiple factors and achieve high-precision prediction.

[0004] The information disclosed in this background section is intended only to enhance the understanding of the overall background of the invention and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0005] The technical problem to be solved by this invention is that existing technologies have difficulty in accurately predicting core losses under high-frequency or non-sinusoidal excitation.

[0006] To solve the above-mentioned technical problems, the present invention is implemented using the following technical solution.

[0007] This invention provides a method for predicting magnetic core loss, comprising:

[0008] Acquire the environmental parameters and one-dimensional magnetic flux density time series signal of the magnetic core under test;

[0009] Based on the environmental parameters and the one-dimensional magnetic flux density time series signal, a prediction is made using a pre-trained magnetic core loss prediction model to obtain the predicted value of the magnetic core loss.

[0010] The data processing method for the core loss prediction model includes:

[0011] The environmental parameters are converted into tensor form to obtain the parameter feature vector;

[0012] The one-dimensional magnetic flux density time series signal is converted into a two-dimensional time-spectrum image, and features are extracted from the two-dimensional time-spectrum image to obtain an image feature vector.

[0013] By concatenating the parameter feature vector and the image feature vector, a comprehensive feature representation is obtained;

[0014] The comprehensive feature representation is processed by a fully connected layer to obtain the predicted value of the core loss;

[0015] The environmental parameters include continuous variables and discrete variables. The continuous variables include temperature and frequency, and the discrete variables include material properties and waveform type.

[0016] Based on the aforementioned core loss prediction method, before converting the environmental parameters into tensor form, the method further includes standardizing the continuous variables using the quartile method, wherein the expression for standardizing the continuous variables using the quartile method is:

[0017] (1);

[0018] In the formula, This represents the standardized environmental parameters. Indicates the original environmental parameters. This represents the median of the original environmental parameters. Indicates the interquartile range, where, Q1 represents the first quartile of the original environmental parameters, and Q3 represents the third quartile of the original environmental parameters.

[0019] Based on the aforementioned core loss prediction method, before converting the environmental parameters into tensor form, the method further includes standardizing the discrete variables using One-Hot encoding.

[0020] Based on the aforementioned core loss prediction method, the method for standardizing the discrete variables using One-Hot coding includes:

[0021] Determine all types of the material properties and waveform types;

[0022] For each material category and waveform category, create separate binary columns for material category and waveform category;

[0023] When the discrete variable belongs to the determined material category, mark 1 at the corresponding position in the binary column of the material category; otherwise mark 0.

[0024] When the discrete variable belongs to the determined waveform category, it is marked with 1 at the corresponding position in the waveform category binary column; otherwise, it is marked with 0.

[0025] Based on the aforementioned core loss prediction method, the binary representation of the material category is as follows:

[0026] M i × i = [ M 1 M 2 … M i ] = [ 1 0 … 0 0 1 … 0 0 0 … 0 0 0 … 1 ] (2);

[0027] In the formula, Represents the binary column for material categories. Indicates the first category of material properties. This indicates the second category of material properties. The first property representing the material Each category, material category binary column Each line represents the One-Hot code corresponding to a material category;

[0028] The waveform category binary column is represented as follows:

[0029] W j × j = [ W 1 W 2 … W j ] = [ 1 0 … 0 0 1 … 0 0 0 … 0 0 0 … 1 ] (3);

[0030] In the formula, The binary column representing the waveform category The first category indicating the waveform type, The second category representing the waveform type. The first wave representing the waveform type Each category, waveform category binary column Each line represents a One-Hot code corresponding to a waveform category.

[0031] Based on the aforementioned core loss prediction method, the one-dimensional magnetic flux density time series signal is converted into a two-dimensional time-spectrum graph, including:

[0032] Using Morlet wavelet transform, the one-dimensional magnetic flux density time series signal is decomposed into a set of basis functions composed of wavelet contraction, flattening and translation.

[0033] The coefficients of the basis functions are extracted and visualized to generate a two-dimensional time-spectrum graph of size 1×24×24 pixels;

[0034] The principle expression of the Morlet wavelet transform is as follows:

[0035] (4);

[0036] In the formula, This represents a set of wavelet basis functions decomposed from a one-dimensional magnetic flux density time series signal f(t). Indicates the current time Little Wave The scaling factor for the degree of expansion or contraction The shift factor representing the degree of wavelet transform. Indicates the current moment The differential, Represents positive infinity. It represents negative infinity.

[0037] Based on the aforementioned core loss prediction method, a multi-layer convolutional neural network (CNN) is used to extract features from the two-dimensional time-spectrum image to obtain an image feature vector, including:

[0038] A 32×24×24 high-order tensor is generated by capturing the basic texture and edge information in the two-dimensional temporal spectrum using a first convolutional layer containing 32 convolutional kernels of size 3×3, stride of 1 and padding technique.

[0039] The 32×24×24 high-order tensor is first max-pooled using a max-pooling layer with a configuration window size of 2×2 and a stride of 2, and compressed into a 32×12×12 high-order tensor.

[0040] The 32×12×12 high-order tensor is converted into a 64×12×12 high-order tensor by using a second layer of convolution with 64 kernels of size 3×3 and stride of 1.

[0041] Using a max pooling layer with a configuration window size of 2×2 and a stride of 2, the 64×12×12 high-order tensor is subjected to a second max pooling, reducing the dimensionality to a 64×6×6 multidimensional tensor;

[0042] The 64×6×6 multidimensional tensor is transformed into a one-dimensional vector of length 2304 using a flattening operation;

[0043] The one-dimensional vector of length 2304 is mapped to a new 512-dimensional space using a fully connected layer, and the ReLU activation function is introduced for processing to obtain the image feature vector.

[0044] Based on the aforementioned core loss prediction method, the parameter feature vector and the image feature vector are concatenated to obtain a comprehensive feature representation, including:

[0045] The parameter feature vector is received by a fully connected layer and transformed into a five-dimensional vector:

[0046] The five-dimensional vector is transformed into a high-level parametric feature vector through a hidden layer containing 128 nodes;

[0047] By concatenating the high-level parameter feature vector and the image feature vector, a comprehensive feature representation is obtained.

[0048] Based on the aforementioned core loss prediction method, the comprehensive feature representation is processed through a fully connected layer to obtain the predicted value of the core loss, including:

[0049] The comprehensive feature representation is processed through a fully connected layer, and then predicted through another fully connected layer with a single output node to obtain the predicted value of the core loss.

[0050] Based on the aforementioned core loss prediction method, the mean square error function is used as the loss function. The model parameters are adjusted and trained using the Adam optimization algorithm and the loss function. The loss function is expressed as follows:

[0051] (5);

[0052] In the formula, N represents the total number of training samples. Indicates the first The true core loss value of each sample. Indicates the first Predicted core loss values ​​for each sample.

[0053] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0054] 1. This invention innovatively integrates environmental parameters with one-dimensional magnetic flux density time series signals, enabling the core loss prediction model to comprehensively consider more dimensions of information, thereby improving the accuracy and comprehensiveness of prediction. This invention converts the one-dimensional magnetic flux density time series signal into a two-dimensional time-frequency spectrum and obtains image feature vectors through feature extraction. This not only preserves the time-frequency characteristics of the signal but also further simplifies the data through feature extraction. This invention also converts environmental parameters into parameter feature vectors and fuses them with image feature vectors, thereby improving processing efficiency.

[0055] 2. This invention constructs a magnetic core loss prediction model based on multimodal data fusion and multiple convolutional neural networks. The model utilizes fully connected layers to process the comprehensive feature representation, obtaining predicted core loss values. It fully leverages the complementarity of multimodal data to improve prediction accuracy. This invention, through its multimodal data and multiple convolutional neural network-based core loss prediction method, demonstrates high accuracy and strong generalization ability in core loss prediction. Compared to traditional single-data source or empirical formula methods, this invention can more accurately predict core losses under different materials, waveforms, and operating environments.

[0056] 3. This invention provides a novel solution for predicting the performance of magnetic components in power electronic devices. By accurately predicting core losses, it not only solves the problem of inaccurate core loss prediction under high-frequency or non-sinusoidal excitation in existing technologies, but also promotes technological innovation and development in related fields. By integrating multimodal data processing with multiple convolutional neural networks, it provides new ideas and methods for solving similar problems, optimizing the design and operation of power electronic devices, improving equipment efficiency and reliability, and reducing costs. This is of great significance to the development and progress of the power electronics industry. Attached Figure Description

[0057] Figure 1 This is a schematic diagram of the technical route of a magnetic core loss prediction method provided in an embodiment of the present invention;

[0058] Figure 2 This is a schematic diagram of the time-frequency domain signal processing based on wavelet transform and multilayer convolutional neural network provided in an embodiment of the present invention;

[0059] Figure 3 This is a schematic diagram of the environmental parameter processing process of a magnetic core loss prediction method provided in an embodiment of the present invention;

[0060] Figure 4 This is a schematic diagram of the characteristic fusion process of a magnetic core loss prediction method provided in an embodiment of the present invention. Detailed Implementation

[0061] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0062] The term "and / or" simply describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. Additionally, the character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0063] Example 1

[0064] like Figure 1 As shown in the figure, this embodiment introduces a method for predicting magnetic core losses, including:

[0065] Step 1: Obtain the environmental parameters and one-dimensional magnetic flux density time series signal of the magnetic core under test.

[0066] The environmental parameters include continuous variables and discrete variables. The continuous variables include temperature and frequency, and the discrete variables include material properties and waveform type.

[0067] The environmental parameters of the magnetic core under test provide the external conditions under which the core operates, such as temperature and frequency, as well as the material properties of the core itself and the waveform type of the excitation signal.

[0068] The one-dimensional magnetic flux density time series signal of the magnetic core under test reflects the change of magnetic flux density inside the core over time, and is an important input data for predicting core loss.

[0069] Environmental parameters and magnetic flux density signals can jointly determine the operating state and loss of the magnetic core.

[0070] Step 2: Based on the environmental parameters and the one-dimensional magnetic flux density time series signal, make a prediction based on the pre-trained magnetic core loss prediction model to obtain the predicted value of the magnetic core loss.

[0071] The pre-trained core loss prediction model has learned the complex relationship between environmental parameters and magnetic flux density signals and core loss. By using the pre-trained model and combining the current environmental parameters and magnetic flux density signals, the core loss can be predicted quickly and accurately, and the predicted value of core loss can be given quickly.

[0072] The data processing method for the core loss prediction model is as follows: Figure 3 As shown, it includes:

[0073] (1) Convert the environmental parameters into tensor form to obtain parameter feature vectors.

[0074] A tensor is a multidimensional array that can efficiently represent and store multidimensional data. By converting environmental parameters into tensor form, subsequent matrix operations and feature extraction become easier.

[0075] This embodiment converts environmental parameters, including continuous and discrete variables, into tensor forms suitable for core loss prediction, which facilitates subsequent feature extraction and model training.

[0076] (2) Convert the one-dimensional magnetic flux density time series signal into a two-dimensional time spectrum, extract features from the two-dimensional time spectrum, and obtain an image feature vector.

[0077] Time-spectrum graphs can reflect the energy distribution of a signal at different times and frequencies, and are an important tool for time series signal analysis. Feature extraction techniques can extract useful feature information from images.

[0078] This invention converts a one-dimensional time series signal into a two-dimensional time spectrum, which can intuitively show the changes of the one-dimensional magnetic flux density time series signal in the time and frequency domains, and can extract more useful feature information.

[0079] (3) Concatenate the parameter feature vector and the image feature vector to obtain a comprehensive feature representation, such as... Figure 4 As shown.

[0080] The concatenation operation can merge two feature vectors from different sources into a single comprehensive feature representation, thus preserving all the information from both.

[0081] This invention concatenates the parametric feature vector of environmental parameters and the image feature vector of a one-dimensional magnetic flux density time series signal to obtain a comprehensive feature representation, which can make full use of the information from both to make a comprehensive and accurate prediction of the magnetic core under test.

[0082] (4) The comprehensive feature representation is processed by the fully connected layer to obtain the predicted value of the core loss.

[0083] Fully connected layers are a basic structure of neural networks that can perform nonlinear transformations on the input feature vectors to obtain the final prediction results.

[0084] This invention utilizes a fully connected layer to transform the comprehensive feature representation through a weight matrix and an activation function, learning the complex relationship between the comprehensive feature representation and the predicted value of core loss. Through multiple iterations of training, the fully connected layer can continuously adjust the parameters of the weight matrix and activation function, making the output predicted value of core loss more accurate.

[0085] The magnetic core loss prediction method provided by this invention combines environmental parameters and one-dimensional magnetic flux density time series signals. Through preprocessing, feature extraction, and model prediction, it achieves rapid and accurate prediction of magnetic core loss. This method leverages the advantages of deep learning models, enabling it to learn the complex relationship between input features and output predicted values, thus providing a new approach and method for predicting magnetic core loss.

[0086] Example 2

[0087] Based on the same inventive concept as Embodiment 1, this embodiment introduces a method for predicting magnetic core losses, specifically including the following steps:

[0088] Step 1: Obtain the environmental parameters and one-dimensional magnetic flux density time series signal of the magnetic core under test.

[0089] First, collect data on the type of magnetic core material, current waveform and frequency data, experimental temperature and other experimental environment information, and record the time series signal of magnetic flux density under the corresponding experimental environment.

[0090] The collected data is cleaned to remove outliers and missing values, ensuring data quality and consistency, and obtaining environmental parameters and one-dimensional magnetic flux density time series signals for the magnetic core under test.

[0091] The environmental parameters include continuous variables and discrete variables. The continuous variables include temperature and frequency, and the discrete variables include material properties and waveform type.

[0092] Step 2: Based on the environmental parameters and the one-dimensional magnetic flux density time series signal, make a prediction based on the pre-trained magnetic core loss prediction model to obtain the predicted value of the magnetic core loss.

[0093] The core loss prediction model includes:

[0094] The data preprocessing module is used to standardize environmental parameters and convert them into tensor data.

[0095] The wavelet transform module is used to convert a one-dimensional magnetic flux density time series signal into a two-dimensional time spectrum.

[0096] The image processing branch is used to extract features from the two-dimensional time-spectrum image based on a multi-layer convolutional neural network to obtain an image feature vector.

[0097] The environmental parameter processing branch is used to process the tensor data of environmental parameters from the fully connected layer to obtain parameter feature vectors.

[0098] The feature fusion layer is used to concatenate the parameter feature vector and the image feature vector to obtain a comprehensive feature representation, which is then further processed through a series of fully connected layers to obtain the predicted value of the core loss.

[0099] During training, the dataset is divided into training and test sets in a 4:1 ratio. The model parameters are continuously adjusted using the Adam optimization algorithm and the MSE loss function to improve the model's predictive performance.

[0100] The parameters of the Adam optimizer are preset:

[0101] Initial learning rate Set it to 0.001;

[0102] The exponential decay rate used for first-moment estimation The value is set to 0.9;

[0103] The exponential decay rate used for second-moment estimation The value of is set to 0.999. In addition, for numerical stability, a very small positive value is introduced. The value is ;

[0104] Set the number of training parameter iterations to 100, the batch size to 32, and the learning rate scheduling to be set to multiply the learning rate by 0.1 every 20 iteration cycles.

[0105] The mean square error function is used as the loss function. The model parameters are adjusted based on the Adam optimization algorithm and the loss function to train the core loss prediction model, wherein the loss function is expressed as:

[0106] (1);

[0107] In the formula, N represents the total number of training samples. Indicates the first The true core loss value of each sample. Indicates the first Predicted core loss values ​​for each sample.

[0108] (1) Convert the environmental parameters into tensor form to obtain parameter feature vectors.

[0109] Before converting the environmental parameters into tensor form, the method further includes standardizing the continuous variables using the quartile method, wherein the expression for standardizing the continuous variables using the quartile method is:

[0110] (2);

[0111] In the formula, This represents the standardized environmental parameters. Indicates the original environmental parameters. This represents the median of the original environmental parameters. Indicates the interquartile range, where, Q1 represents the first quartile of the original environmental parameters, and Q3 represents the third quartile of the original environmental parameters.

[0112] Before converting the environmental parameters into tensor form, the discrete variables are standardized using One-Hot encoding.

[0113] A method for standardizing the discrete variables using One-Hot coding includes:

[0114] Determine all types of the material properties and waveform types;

[0115] For each material category and waveform category, create separate binary columns for material category and waveform category;

[0116] When the discrete variable belongs to the determined material category, mark 1 at the corresponding position in the binary column of the material category; otherwise mark 0.

[0117] When the discrete variable belongs to the determined waveform category, it is marked with 1 at the corresponding position in the waveform category binary column; otherwise, it is marked with 0.

[0118] The binary representation of the material category is as follows:

[0119] M i × i = [ M 1 M 2 … M i ] = [ 1 0 … 0 0 1 … 0 0 0 … 0 0 0 … 1 ] (3);

[0120] In the formula, Represents the binary column for material categories. Indicates the first category of material properties. This indicates the second category of material properties. The first property representing the material Each category, material category binary column Each line represents the One-Hot code corresponding to a material category;

[0121] The waveform category binary column is represented as follows:

[0122] W j × j = [ W 1 W 2 … W j ] = [ 1 0 … 0 0 1 … 0 0 0 … 0 0 0 … 1 ] (4);

[0123] In the formula, The binary column representing the waveform category The first category indicating the waveform type, The second category representing the waveform type. The first wave representing the waveform type Each category, waveform category binary column Each line represents a One-Hot code corresponding to a waveform category.

[0124] (2) Convert the one-dimensional magnetic flux density time series signal into a two-dimensional time spectrum, extract features from the two-dimensional time spectrum, and obtain an image feature vector.

[0125] The process of converting the one-dimensional magnetic flux density time series signal into a two-dimensional time-spectrum graph includes:

[0126] Using Morlet wavelet transform, the one-dimensional magnetic flux density time series signal is decomposed into a set of basis functions composed of wavelet contraction, flattening and translation.

[0127] The coefficients of the basis functions are extracted and visualized to generate a two-dimensional time-spectrum graph of size 1×24×24 pixels;

[0128] The principle expression of the Morlet wavelet transform is as follows:

[0129] (5);

[0130] In the formula, This represents a set of wavelet basis functions decomposed from a one-dimensional magnetic flux density time series signal f(t). Indicates the current time Little Wave The scaling factor for the degree of expansion or contraction The shift factor representing the degree of wavelet transform. Indicates the current moment The differential, Represents positive infinity. This represents negative infinity. Temporal information can be extracted by scanning the shift factor of the wavelet transform.

[0131] The scaling factor a corresponds to 1024 sampling points within one period, which is a sequence of integers from 1 to 1023. The translation factor b is automatically obtained by the one-dimensional convolution process and depends on the input one-dimensional magnetic flux density time series signal and the length of the wavelet. Wavelet transform can simultaneously preserve information in the time domain and frequency domain, which is convenient for subsequent feature extraction.

[0132] like Figure 2 As shown, a multi-layer convolutional neural network (CNN) is used to extract features from the two-dimensional time-spectral image to obtain an image feature vector, including:

[0133] A 32×24×24 high-order tensor is generated by using a first convolutional layer containing 32 kernels of size 3×3 with a stride of 1 and employing padding techniques to capture the basic texture and edge information in the two-dimensional temporal spectrum.

[0134] The purpose of using padding in this embodiment is to keep the size of the two-dimensional time-frequency spectrum unchanged. Each convolutional kernel in the first layer is responsible for detecting different local features and generating a 32×24×24 high-order tensor. This multi-channel output enhances the understanding of image details by the magnetic core loss prediction model.

[0135] The 32×24×24 high-order tensor is first max-pooled using a max-pooling layer with a configuration window size of 2×2 and a stride of 2, and compressed into a 32×12×12 high-order tensor.

[0136] The main purpose of performing the first max pooling in this embodiment is to reduce the spatial dimension of the two-dimensional time-frequency spectrum while retaining the most representative features. During the first max pooling process, the amount of computation is reduced, and the robustness of the core loss prediction model to small displacement changes in the input data is increased.

[0137] The 32×12×12 high-order tensor is transformed into a 64×12×12 high-order tensor by using a second layer of convolution with 64 kernels of size 3×3 and stride of 1.

[0138] The level of abstraction is further enhanced. As the number of convolutional layers increases, the learned visual patterns become more complex. Through the second convolutional layer, the high-order tensor data is transformed from 32×12×12 to 64×12×12. The increase in the number of channels means that more feature combinations are taken into account, which helps to improve the expressive power of the core loss prediction model.

[0139] Using a max pooling layer with a configuration window size of 2×2 and a stride of 2, the 64×12×12 high-order tensor is subjected to a second max pooling, reducing its dimensionality to a 64×6×6 multidimensional tensor.

[0140] The purpose of performing a second max pooling in this embodiment is to further reduce the spatial resolution of the feature map, enhance the core loss prediction model's ability to identify key features, and mitigate the risk of overfitting.

[0141] The 64×6×6 multidimensional tensor is transformed into a one-dimensional vector of length 2304 using a flattening operation.

[0142] By using the flattening operation, the 64×6×6 multidimensional tensor is transformed into a one-dimensional vector of length 2304, which makes it easier for subsequent fully connected layers to directly process the 64×6×6 multidimensional tensor, making the entire network structure more compact and efficient.

[0143] The one-dimensional vector of length 2304 is mapped to a new 512-dimensional space using a fully connected layer, and the ReLU activation function is introduced for processing to obtain the image feature vector.

[0144] The purpose of introducing the ReLU activation function in this embodiment is to help the core loss prediction model better learn the complex mapping relationship between input data and core loss.

[0145] (3) Concatenate the parameter feature vector and the image feature vector to obtain a comprehensive feature representation, including:

[0146] The parameter feature vector is received by a fully connected layer and transformed into a five-dimensional vector:

[0147] The five-dimensional vector is transformed into a high-level parametric feature vector through a hidden layer containing 128 nodes;

[0148] By concatenating the high-level parameter feature vector and the image feature vector, a comprehensive feature representation is obtained.

[0149] (4) The comprehensive feature representation is processed through a fully connected layer to obtain a predicted value of the core loss, including:

[0150] The comprehensive feature representation is processed through a fully connected layer, and then predicted through another fully connected layer with a single output node to obtain the predicted value of the core loss.

[0151] Example 3

[0152] A computer-readable storage medium storing computer instructions thereon, characterized in that the computer instructions, when executed by a processor, implement the steps of the method described in Embodiment 1 or 2 above.

[0153] Example 4

[0154] The present invention also provides a computer program product, including computer instructions, characterized in that the computer instructions, when executed by a processor, implement the steps of the methods described in Embodiment 1 or 2 above.

[0155] In summary, this invention innovatively integrates environmental parameters with one-dimensional magnetic flux density time series signals, enabling the core loss prediction model to comprehensively consider more dimensions of information, thereby improving the accuracy and comprehensiveness of prediction. This invention converts the one-dimensional magnetic flux density time series signal into a two-dimensional time-frequency spectrum and obtains image feature vectors through feature extraction. This not only preserves the time-frequency characteristics of the signal but also further simplifies the data through feature extraction. Furthermore, this invention converts environmental parameters into parameter feature vectors and fuses them with image feature vectors, improving processing efficiency.

[0156] This invention constructs a magnetic core loss prediction model based on multimodal data fusion and multiple convolutional neural networks. The model utilizes fully connected layers to process the comprehensive feature representation, obtaining predicted core loss values. It fully leverages the complementarity of multimodal data to improve prediction accuracy. This invention, through its multimodal data and multiple convolutional neural network-based core loss prediction method, demonstrates high accuracy and strong generalization ability in core loss prediction. Compared to traditional single-data source or empirical formula methods, this invention can more accurately predict core losses under different materials, waveforms, and operating environments.

[0157] This invention provides a novel solution for predicting the performance of magnetic components in power electronic devices. By accurately predicting core losses, it not only solves the problem of accurately predicting core losses under high-frequency or non-sinusoidal excitation in existing technologies, but also promotes technological innovation and development in related fields. By integrating multimodal data processing with multiple convolutional neural networks, it provides new ideas and methods for solving similar problems, optimizing the design and operation of power electronic devices, improving equipment efficiency and reliability, and reducing costs. This is of great significance to the development and progress of the power electronics industry. Those skilled in the art will understand that embodiments of this invention can be provided as methods, systems, or computer program products. Therefore, this invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0158] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0159] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0160] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0161] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A method for predicting magnetic core loss, characterized in that, include: Acquire the environmental parameters and one-dimensional magnetic flux density time series signal of the magnetic core under test; Based on the environmental parameters and the one-dimensional magnetic flux density time series signal, a prediction is made using a pre-trained magnetic core loss prediction model to obtain the predicted value of the magnetic core loss. The data processing method for the core loss prediction model includes: The environmental parameters are converted into tensor form to obtain the parameter feature vector; Converting the one-dimensional magnetic flux density time series signal into a two-dimensional time-spectrum graph includes: Using Morlet wavelet transform, the one-dimensional magnetic flux density time series signal is decomposed into a set of basis functions composed of wavelet contraction, flattening and translation. The coefficients of the basis functions are extracted and visualized to generate a two-dimensional time-spectrum graph of size 1×24×24 pixels; The principle expression of the Morlet wavelet transform is as follows: （1）； In the formula, This represents a set of wavelet basis functions decomposed from a one-dimensional magnetic flux density time series signal f(t). Indicates the current time Little Wave The scaling factor for the degree of expansion or contraction The shift factor representing the degree of wavelet transform. Indicates the current moment The differential, Represents positive infinity. Represents negative infinity; Feature extraction is performed on the two-dimensional time-spectrum image to obtain an image feature vector, including: A 32×24×24 high-order tensor is generated by capturing the basic texture and edge information in the two-dimensional temporal spectrum using a first convolutional layer containing 32 convolutional kernels of size 3×3, stride of 1 and padding technique. The 32×24×24 high-order tensor is first max-pooled using a max-pooling layer with a configuration window size of 2×2 and a stride of 2, and compressed into a 32×12×12 high-order tensor. The 32×12×12 high-order tensor is converted into a 64×12×12 high-order tensor by using a second layer of convolution with 64 kernels of size 3×3 and stride of 1. Using a max pooling layer with a configuration window size of 2×2 and a stride of 2, the 64×12×12 high-order tensor is subjected to a second max pooling, reducing the dimensionality to a 64×6×6 multidimensional tensor; The 64×6×6 multidimensional tensor is transformed into a one-dimensional vector of length 2304 using a flattening operation; The one-dimensional vector of length 2304 is mapped to a new 512-dimensional space using a fully connected layer, and the ReLU activation function is introduced for processing to obtain the image feature vector; By concatenating the parameter feature vector and the image feature vector, a comprehensive feature representation is obtained, including: The parameter feature vector is received by a fully connected layer and transformed into a five-dimensional vector: The five-dimensional vector is transformed into a high-level parametric feature vector through a hidden layer containing 128 nodes; By concatenating the high-level parameter feature vector and the image feature vector, a comprehensive feature representation is obtained; The comprehensive feature representation is processed by a fully connected layer to obtain a predicted value for the core loss, including: The comprehensive feature representation is processed through a fully connected layer and then predicted through another fully connected layer with a single output node to obtain the predicted value of the core loss. The environmental parameters include continuous variables and discrete variables. The continuous variables include temperature and frequency, and the discrete variables include material properties and waveform type.

2. The core loss prediction method according to claim 1, characterized in that, Before converting the environmental parameters into tensor form, the method further includes standardizing the continuous variables using the quartile method, wherein the expression for standardizing the continuous variables using the quartile method is: （2）； In the formula, This represents the standardized environmental parameters. Indicates the original environmental parameters. This represents the median of the original environmental parameters. Indicates the interquartile range, where, Q1 represents the first quartile of the original environmental parameters, and Q3 represents the third quartile of the original environmental parameters.

3. The method for predicting magnetic core loss according to claim 1, characterized in that, Before converting the environmental parameters into tensor form, the discrete variables are standardized using One-Hot encoding.

4. The core loss prediction method according to claim 3, characterized in that, A method for standardizing the discrete variables using One-Hot coding includes: Determine all types of the material properties and waveform types; For each material category and waveform category, create separate binary columns for material category and waveform category; When the discrete variable belongs to the determined material category, mark 1 at the corresponding position in the binary column of the material category; otherwise mark 0. When the discrete variable belongs to the determined waveform category, it is marked with 1 at the corresponding position in the waveform category binary column; otherwise, it is marked with 0.

5. The core loss prediction method according to claim 4, characterized in that, The binary representation of the material category is as follows: （3）； In the formula, Represents the binary column for material categories. Indicates the first category of material properties. This indicates the second category of material properties. The first property representing the material Each category, material category binary column Each line represents the One-Hot code corresponding to a material category; The waveform category binary column is represented as follows: （4）； In the formula, The binary column representing the waveform category The first category indicating the waveform type, The second category representing the waveform type. The first wave representing the waveform type Each category, waveform category binary column Each line represents a One-Hot code corresponding to a waveform category.

6. The method for predicting magnetic core loss according to claim 1, characterized in that, The mean square error function is used as the loss function. The model parameters are adjusted based on the Adam optimization algorithm and the loss function to train the core loss prediction model, wherein the loss function is expressed as: （5）； In the formula, Let N represent the total loss and the number of training samples. Indicates the first The true core loss value of each sample. Indicates the first Predicted core loss values ​​for each sample.

Images