Transformer magnetic core loss prediction method and system based on empirical formula and model

Abstract

The invention belongs to the field of transformer magnetic core loss prediction, and provides a transformer magnetic core loss prediction method and system based on an empirical formula and a model in order to solve the problems that transformer magnetic core loss prediction cannot be suitable for different working conditions and a transformer magnetic core loss prediction result is poor in precision generally existing at present. The transformer magnetic core loss prediction method based on the empirical formula and the model comprises the following steps: preliminarily obtaining a Steinmaz equation considering a temperature correction factor through fitting; enabling the non-sinusoidal excitation current waveform to be equivalent to a sinusoidal excitation current waveform for calculation, and continuing to correct the Steinmaz equation considering the temperature correction factor to obtain a corrected empirical formula; and processing the magnetic core loss experience prediction value and the corresponding temperature, exciting current, magnetic flux density and frequency of the exciting current by using a pre-trained neural network model to obtain a final magnetic core loss prediction value. The method can accurately predict the magnetic core loss and is suitable for different working conditions.

Description

Technical Field

[0001] The present invention belongs to the field of transformer core loss prediction, and in particular relates to a transformer core loss prediction method and system based on empirical formulas and models. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Transformer core is an indispensable device in power conversion, and high efficiency and high power density core design has become one of the hottest areas at present. However, the requirements of high efficiency and high power density cause high loss problems in magnetic components.

[0004] For general copper wire losses, more accurate loss results can be obtained through finite element simulation software. However, for some special magnetic materials, such as ferrites and nanocrystals, due to their complex microstructures, the loss types include hysteresis loss, eddy current loss, and stray loss. Each loss is related to many factors. It is difficult to accurately analyze the loss through simulation analysis, and the mechanism modeling is also more difficult. The traditional method is to estimate through empirical formulas, such as the Steinmetz equation and its various modified forms, including MSE, GSE, iGSE and i 2 GSE and other correction models. However, this type of method is difficult to achieve a balance between the scope of application, solution accuracy and practical performance. For example, the Steinmetz equation does not consider the influence of operating temperature on the core loss of the transformer, its own accuracy is poor, and only the sinusoidal excitation wave is considered. When other waveforms are considered to correct the equation, the solution accuracy of the equation will be further reduced. The loss separation method can achieve a more accurate solution by calculating separately according to the core loss type, but in this process, more parameters are required, the solution accuracy is greatly affected by the frequency, and the practicality is low.

[0005] In summary, in the field of transformer core loss prediction, there are common technical problems that transformer core loss prediction cannot be applied to different working conditions and the accuracy of transformer core loss prediction results is poor. Summary of the invention

[0006] In order to solve the above technical problems, the present invention provides a transformer core loss prediction method and system based on empirical formulas and models, which can accurately predict the core loss while expanding the scope of application as much as possible to make it suitable for different working conditions, and the method has strong interpretability.

[0007] In order to achieve the above object, the present invention adopts the following technical solution: A first aspect of the present invention provides a transformer core loss prediction method based on empirical formulas and models.

[0008] In one or more embodiments, a transformer core loss prediction method based on an empirical formula and a model is provided, comprising: Considering the influence of temperature on core loss, the proportionality coefficient in the Steinmetz equation is taken as a function of temperature and represented by a set exponential function. The power exponents of the frequency of the excitation current and the amplitude of the magnetic flux density in the Steinmetz equation are both taken as functions of temperature and represented by a set polynomial function. The Steinmetz equation taking into account the temperature correction factor is preliminarily obtained through fitting. Considering the influence of different excitation waveforms on the core loss, by introducing the equivalent sinusoidal frequency and its calculation formula, the non-sinusoidal excitation current waveform is equivalent to the sinusoidal excitation current waveform for calculation, and then the Steinmetz equation taking into account the temperature correction factor is corrected to obtain the corrected empirical formula; Based on the temperature, excitation current, magnetic flux density and the frequency of the excitation current, the empirical prediction value of the core loss is obtained using the modified empirical formula; The pre-trained improved neural network model is used to process the empirical prediction value of the core loss and its corresponding temperature, excitation current, flux density and frequency of the excitation current to obtain the final prediction value of the core loss.

[0009] As an implementation manner, the exponential function of the proportionality coefficient in the Steinmetz equation is characterized by the e-index.

[0010] As an implementation method, the power exponents of both the frequency of the excitation current and the amplitude of the magnetic flux density are characterized by quadratic polynomial functions.

[0011] As an implementation method, the characterization formula of the equivalent sinusoidal frequency and the modified empirical formula are respectively:

[0012]

[0013] in, is the equivalent sinusoidal frequency of the excitation current; It is the peak-to-peak value of the magnetic flux density in one cycle, that is, the difference between the maximum value and the minimum value; is the rate of change of magnetic flux density with time; is the period of the excitation waveform; , ~ is the coefficient of the equation; is the actual frequency of the excitation current; is the transformer core loss; Indicates temperature; is the magnitude of the magnetic flux density.

[0014] As an implementation method, the neural network model adopts an LSTM-AM (long short-term memory-multi-head attention, long short-term memory neural network-attention mechanism) model.

[0015] A second aspect of the present invention provides a transformer core loss prediction system based on empirical formulas and models.

[0016] In one or more embodiments, a transformer core loss prediction system based on an empirical formula and a model includes: A preliminary correction module is used to consider the influence of temperature on the core loss, taking the proportionality coefficient in the Steinmetz equation as a function of temperature and using a set exponential function to represent it, taking the frequency of the excitation current and the amplitude of the magnetic flux density in the Steinmetz equation as functions of temperature and using a set polynomial function to represent them, and preliminarily obtaining the Steinmetz equation taking into account the temperature correction factor through fitting; An empirical formula determination module is used to consider the influence of different excitation waveforms on the core loss. By introducing the equivalent sinusoidal frequency and its calculation formula, the non-sinusoidal excitation current waveform is equivalent to the sinusoidal excitation current waveform for calculation, and the Steinmetz equation taking into account the temperature correction factor is further corrected to obtain the corrected empirical formula; An empirical prediction value calculation module is used to obtain an empirical prediction value of core loss based on temperature, excitation current, magnetic flux density and frequency of the excitation current using a modified empirical formula; The core loss prediction module uses a pre-trained neural network model to process the empirical prediction value of the core loss and its corresponding temperature, excitation current, flux density and frequency of the excitation current to obtain the final core loss prediction value.

[0017] As an implementation manner, in the preliminary correction module, the exponential function of the proportionality coefficient in the Steinmetz equation is characterized by the e-index.

[0018] As an implementation mode, in the preliminary correction module, the power exponents of both the frequency of the excitation current and the amplitude of the magnetic flux density are characterized by quadratic polynomial functions.

[0019] As an implementation manner, in the empirical formula determination module, the characterization formula of the equivalent sinusoidal frequency and the modified empirical formula are respectively:

[0020]

[0021] in, is the equivalent sinusoidal frequency of the excitation current; It is the peak-to-peak value of the magnetic flux density in one cycle, that is, the difference between the maximum value and the minimum value; is the rate of change of magnetic flux density with time; is the period of the excitation waveform; , ~ is the coefficient of the equation; is the actual frequency of the excitation current; is the transformer core loss; Indicates temperature; is the magnitude of the magnetic flux density.

[0022] As an implementation mode, in the core loss prediction module, the neural network model adopts the LSTM-AM model.

[0023] Compared with the prior art, the present invention has the following beneficial effects: (1) The present invention improves the traditional Steinmetz formula by adding a temperature correction coefficient, so that the core loss can be predicted more accurately at different temperatures. At the same time, by introducing the calculation method of equivalent sinusoidal frequency, the improved empirical formula can be applied to the calculation of core loss under various excitation waveforms.

[0024] (2) The present invention combines the improved empirical formula with the artificial intelligence model, and adopts the temperature correction factor, equivalent sinusoidal frequency and black box model in turn, thereby realizing the hybrid drive of the model and data, expanding the scope of application of the power transformer core loss prediction, and improving the accuracy of the prediction.

[0025] (3) The artificial intelligence model of the present invention is implemented using the LSTM-AM model. The LSTM network is used to capture the hidden relationship between the change pattern of physical quantities such as magnetic flux density over time and the core loss, and the AM network is used to determine the weight between each input feature and the output prediction value, thereby greatly improving the prediction accuracy of the transformer core loss. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0027] Figure 1 It is a flow chart of a transformer core loss prediction method based on an empirical formula and a model according to an embodiment of the present invention; Figure 2 It is a specific flow chart of transformer core loss prediction based on empirical formula and model in an embodiment of the present invention; Figure 3 It is a schematic diagram of the structure of a transformer core loss prediction system based on empirical formulas and models according to an embodiment of the present invention; Figure 4 is a schematic diagram of an electronic device according to an embodiment of the present invention; Figure 5 is a comparison of prediction errors between the polynomial form and the exponential form of an embodiment of the present invention; Figure 6 is a comparison between the exponential form predicted loss and the measured loss of an embodiment of the present invention; Figure 7 is the absolute error of the core loss prediction of the embodiment of the present invention; Figure 8 is the prediction error of the improved empirical formula of the embodiment of the present invention; Fig. 9 LSTM structure diagram of an embodiment of the present invention; Fig.10 is the absolute error of model prediction in the embodiment of the present invention. DETAILED DESCRIPTION

[0028] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0029] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0030] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0031] Figure 1 is a flow chart of a transformer core loss prediction method based on empirical formulas and models in an embodiment of the present invention, such as Figure 1 The transformer core loss prediction method based on the empirical formula and model in the embodiment shown may include: S101, considering the influence of temperature on the core loss, the proportionality coefficient in the Steinmetz equation is taken as a function of temperature and is represented by a set exponential function, and the power exponents of the frequency of the excitation current and the amplitude of the magnetic flux density in the Steinmetz equation are both taken as functions of temperature and are represented by a set polynomial function, and the Steinmetz equation taking into account the temperature correction factor is preliminarily obtained through fitting; S102, considering the influence of different excitation waveforms on the core loss, by introducing the equivalent sinusoidal frequency and its calculation formula, the non-sinusoidal excitation current waveform is equivalent to the sinusoidal excitation current waveform for calculation, and the Steinmetz equation taking into account the temperature correction factor is further corrected to obtain the corrected empirical formula; S103, obtaining an empirical prediction value of the core loss using a modified empirical formula based on the temperature, the excitation current, the magnetic flux density, and the frequency of the excitation current; S104, using a pre-trained neural network model to process the core loss empirical prediction value and its corresponding temperature, excitation current, magnetic flux density and excitation current frequency to obtain a final core loss prediction value.

[0032] This embodiment improves the traditional Steinmetz formula by adding a temperature correction coefficient so that it can more accurately predict the core loss at different temperatures. At the same time, by introducing the calculation method of the equivalent sinusoidal frequency, the improved empirical formula can be applied to the calculation of the core loss under various excitation waveforms. The improved empirical formula is combined with the neural network model, and the temperature correction factor, equivalent sinusoidal frequency and artificial intelligence model are used in turn to realize the hybrid drive of the model and data, thereby expanding the scope of application of the power transformer core loss prediction and improving the accuracy of the prediction.

[0033] In step S101, the general form of the Steinmetz equation is shown in equation (1).

[0034] (1) In the above formula, P is the core loss, f is the actual frequency of the excitation current, is the magnitude of the magnetic flux density, and , , is the coefficient that needs to be fitted with data, and its size varies with the material and temperature. The value range is generally 1-3. The value range is generally 2-3.

[0035] Although the Steinmetz equation is very classic, it is only applicable to the case where the excitation waveform is a sinusoidal wave and does not consider the influence of different temperatures. Therefore, it is difficult to make an accurate prediction of the loss of the transformer core.

[0036] In order to take temperature into account based on the Steinmetz equation T First, consider the coefficients of the equation k is a function of temperature, that is k ( T). Generally, there are two classic forms when using functions to fit data, one is the polynomial form and the other is the exponential form, as shown in formula (2) and formula (3), which are called method 1 and method 2 respectively.

[0037] (2) (3) in, Indicates temperature; is the proportionality coefficient; and are the coefficients that need to be fitted.

[0038] The two forms k ( T ) are respectively substituted into formula (1), then the fitting function becomes: The nonlinear least squares fitting can be performed using Matlab's nlinfit function to determine the coefficients of the fitting function. It can be expressed as (4): (4) In the above formula, y is the dependent variable, x is the independent variable, is a nonlinear function that depends on the parameters, is a random error.

[0039] The nlinfit function estimates the parameters by minimizing the residual sum of squares. , the specific formula is (5).

[0040] (5) In the above formula, represents the actual observed value, Represents the observed values ​​of the model.

[0041] In this embodiment, the exponential function of the proportionality coefficient in the Steinmetz equation is represented by the e exponent.

[0042] The power exponents of both the frequency of the exciting current and the amplitude of the magnetic flux density are characterized by quadratic polynomial functions.

[0043] The loss experimental data of a certain material is used to calculate the relevant fitting coefficients, where the experimental temperature ranges from 25°C to 90°C, the frequency ranges from 50 to 500000Hz, and the excitation waveform is mainly divided into sine wave, triangle wave and square wave, with a duration of one cycle. The following uses method 1 and method 2 to fit the loss generated under the sine wave, and the absolute error of the loss prediction is as follows: Figure 5 As shown in Figure 2, the predicted value and the measured value of method 2 are as follows Figure 6As shown, the fitting coefficient values ​​under the two methods (retain 4 significant figures) are shown in Table 1 and Table 2.

[0044] Table 1 Values ​​of coefficients in method 1

[0045] Table 2 Values ​​of coefficients in method 2

[0046] It can be seen that the error of method 2 is smaller when fitting, and the effect is obviously better than method 1, so method 2 is preferred for fitting. Considering that the frequency and the power of magnetic flux density of the Steinmetz equation are different under different temperature conditions, the power of frequency and magnetic flux density can be regarded as a function of temperature. Assuming that the function can be expanded into a Taylor series, the power of frequency and magnetic flux density can be further changed into a polynomial about temperature.

[0047] The improved fitting formula is shown in formula (6).

[0048] (6) After fitting using formula (6), the absolute error of loss prediction under sinusoidal wave excitation is further reduced, and the maximum absolute error does not exceed 0.8×10 5 The absolute error is Figure 7 The fitting coefficients of the empirical formula (6) are shown in Table 3.

[0049] Table 3 Coefficients of the empirical formula (6)

[0050] It can be seen from the above table that the quadratic coefficient of temperature k 5 , k 8 It is basically equal to 0, which verifies the rationality of fitting using a quadratic polynomial.

[0051] In step S102, in order to solve the loss generated by the core under non-sinusoidal wave excitation, equation (6) is further improved. Some scholars have proposed an improved loss prediction empirical formula Modified Steinmetz-Equation for different waveforms. This formula assumes that the core loss is related to the rate of change of the magnetic flux density. By weighting the rate of change of the magnetic flux density at different time points, the weighted average rate of change of the magnetic flux density is obtained, and it is further considered that the core loss is determined by this weighted average rate of change. In this way, the equivalent sinusoidal wave frequency corresponding to any waveform excitation can be calculated. The specific expression formulas are shown in (7)(8).

[0052] The characterization formula of the equivalent sinusoidal frequency and the modified empirical formula are: (7) (8) (9) in, is the equivalent sinusoidal frequency of the excitation current; It is the peak-to-peak value of the magnetic flux density in one cycle, that is, the difference between the maximum value and the minimum value; is the rate of change of magnetic flux density with time; is the period of the excitation waveform; , ~ is the coefficient of the equation; is the actual frequency of the excitation current; is the transformer core loss; Indicates temperature; is the magnitude of the magnetic flux density.

[0053] By introducing the equivalent sinusoidal frequency, formula (7) can calculate the core loss under any waveform, greatly expanding the scope of application of the empirical formula. In many cases, it is difficult to accurately understand the analytical expression of the excitation waveform, and generally there are only discrete sampling points. Therefore, the formula (7) It can be calculated by difference, while the integral term is calculated by trapezoidal formula, as shown in (10).

[0054] (10) Wherein, N is the total number of magnetic flux density sampling points in one cycle.

[0055] The final form of the improved empirical formula is obtained by combining formulas (7), (9) and (10). When fitting, the equivalent sinusoidal frequency of the excitation waveform is first calculated using formula (10) and formula (7), and then the fitting coefficient in formula (9) is calculated using the nlinfit function in Matlab.

[0056] The accuracy of the improved empirical formula is tested using 3400 sample data including sine wave, triangle wave, trapezoidal wave and other waveforms, and the mean square error (MSE) is calculated. The calculation formula of MSE is shown in (11).

[0057] (11) In the above formula, and are the true value and predicted value of the data respectively.

[0058] The absolute error calculated according to the improved empirical formula is as follows: Figure 8As shown, MSE=8.512×10 8 .

[0059] Depend on Figure 7 and Figure 6 By comparison, it can be seen that the absolute error increases significantly, that is, after the applicable scope of the empirical formula is expanded, it is difficult to ensure the accuracy of the loss solution.

[0060] Considering that the prediction effect of core loss for non-sinusoidal excitation waves using formulas (7), (9), and (10) is poor, a black box model constructed using artificial intelligence methods is used to further reduce the prediction error.

[0061] In this embodiment, the artificial intelligence model adopts the LSTM-AM model.

[0062] like Figure 2 As shown in the figure, the specific method is: temperature, excitation waveform (different waveforms are distinguished by numbers 1, 2, 3), magnetic flux density, frequency, and improved empirical formula loss prediction value are used as the input of the neural network, and the output is the final magnetic core loss prediction value. In order to accurately capture the hidden relationship between the change of magnetic flux density over time and the magnetic core loss, the LSTM model is introduced, and the attention mechanism (AM) is further used to accurately learn the relationship between each feature and the output.

[0063] The LSTM network adds a unit state c on the basis of CNN to save the long-term state, solving the problem of gradient disappearance and explosion during CNN training.

[0064] Assume that the input sequence X=[x 1 ,x 2 ,x 3 ,…,x T ], the recursive hidden layer is sequentially t =1,2,3,…,T calculates the activation values ​​of the input gate, forget gate, output gate and memory unit in turn. t The calculation formulas for the time are shown in (12)-(15).

[0065] The input gate is: (12) In the above formula, is the Sigmoid activation function, is the weight matrix between the input gate and the input vector, is the weight matrix of the input gate and the intermediate state vector at the previous moment, is the input gate and the memory unit at the previous moment The weight matrix of is the bias vector.

[0066] The forget gate is: (13) In the above formula, , , They are the forget gate and , , The weight matrix of is the bias vector of the forget gate.

[0067] The memory units are: (14) In the above formula, and The memory unit and , The weight matrix of is the hyperbolic tangent function, is the bias vector.

[0068] The output gate is: (15) In the above formula, , , They are the output gate and , , The weight matrix of .

[0069] The hidden layer output is: (16) The activation function used in this embodiment is the linear rectification function ReLU. The structure diagram of the LSTM model is as follows: Fig. 9 shown.

[0070] The essence of the attention mechanism is weighted summation. Assuming k indivual d The feature vector of (i=1,2,3,…, k ), integrating this k The information of the feature vectors, and the integrated vector is recorded as , here we take the weighted average, as shown in formula (17).

[0071] (17) In the above formula, For weight, the fundamental purpose of AM is to obtain a reasonable weight.

[0072] The solution of weights is divided into two steps: (1) Design a scoring function g for each Calculate the corresponding score , the scoring is based on The greater the relevance to the object of interest, the The larger the value of .

[0073] (2) The k scores obtained are subjected to the softmax function to obtain the final weights , the specific formula is formula (18).

[0074] (18) The AM adopted in this embodiment is constructed by adding a fully connected layer with the same number of nodes and units in the LSTM layer between the LSTM layer and the fully connected layer with 1 neuron, as shown in formula (19).

[0075] (19) In the above formula, is the output vector of the LSTM layer at the current moment, is the attention weight matrix, is the attention bias vector.

[0076] The importance of the output of each neuron in the LSTM layer to the current network output is different. The output of LSTM can be regarded as the basic encoding of the input sequence. The AM in the network is responsible for automatically learning the attention weights, capturing global features, and measuring the importance of each output of the LSTM layer to the entire network output. That is, the output of the LSTM layer and the network output are used as the input of AM, and then the weight coefficient is learned about the function of the two. The weight obtained through learning can further reflect the correlation between the two.

[0077] The 3400 sets of core loss sample data are randomly divided into 70% training set and 30% test set. For the 30% test set, the absolute error of the core loss prediction is as follows: Fig.10 As shown. The mean square error MSE = 1.441×10 8 ,Compared with only the improved empirical formula, the MSE index decreased by 83.39%, achieving a significant effect.

[0078] Figure 3 Schematic diagram of a transformer core loss prediction system based on empirical formulas and models in an embodiment of the present invention. Figure 1 The transformer core loss prediction method based on empirical formula and model corresponds to Figure 3 As shown, the transformer core loss prediction system based on the empirical formula and model in this embodiment may include: A preliminary correction module 301 is used to consider the influence of temperature on the magnetic core loss, take the proportionality coefficient in the Steinmetz equation as a function of temperature and characterize it with a set exponential function, take the power exponents of the frequency of the excitation current and the amplitude of the magnetic flux density in the Steinmetz equation as functions of temperature and characterize them with a set polynomial function, and preliminarily obtain the Steinmetz equation taking into account the temperature correction factor through fitting; An empirical formula determination module 302 is used to consider the influence of different excitation waveforms on the magnetic core loss, and by introducing an equivalent sinusoidal frequency and its calculation formula, the non-sinusoidal excitation current waveform is equivalent to a sinusoidal excitation current waveform for calculation, and the Steinmetz equation taking into account the temperature correction factor is further corrected to obtain a corrected empirical formula; An empirical prediction value calculation module 303 is used to obtain an empirical prediction value of the core loss based on the temperature, the excitation current, the magnetic flux density and the frequency of the excitation current using a modified empirical formula; The core loss prediction module 304 uses a pre-trained neural network model to process the core loss empirical prediction value and its corresponding temperature, excitation current, magnetic flux density and excitation current frequency to obtain a final core loss prediction value.

[0079] In a specific implementation process, in the preliminary correction module 301, the exponential function of the proportionality coefficient in the Steinmetz equation is characterized by the e index.

[0080] In the preliminary correction module 301, the power exponents of both the frequency of the excitation current and the amplitude of the magnetic flux density are represented by quadratic polynomial functions.

[0081] In the empirical formula determination module 302, the characterization formula of the equivalent sinusoidal frequency and the modified empirical formula are respectively:

[0082]

[0083] in, is the equivalent sinusoidal frequency of the excitation current; It is the peak-to-peak value of the magnetic flux density in one cycle, that is, the difference between the maximum value and the minimum value; is the rate of change of magnetic flux density with time; is the period of the excitation waveform; , ~ is the coefficient of the equation; is the actual frequency of the excitation current; is the transformer core loss; Indicates temperature; is the magnitude of the magnetic flux density.

[0084] In the core loss prediction module 304, the neural network model adopts the LSTM-AM model.

[0085] It should be noted here that Figure 3 The various modules in the transformer core loss prediction system based on empirical formulas and models are Figure 1 The various steps in the transformer core loss prediction method based on empirical formulas and models correspond one to one, and the specific implementation process is the same, which will not be repeated here.

[0086] Reference Figure 4 , a schematic diagram of an electronic device is given. It should be noted that, Figure 4 The electronic device 400 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0087] like Figure 4 As shown, the electronic device 400 includes a central processing unit (CPU) 401, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 402 or the program loaded from the storage part 408 to the random access memory (RAM) 403. In the RAM 403, various programs and data required for system operation are also stored. The central processing unit 401, the ROM 402 and the RAM 403 are connected to each other through a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0088] The following components are connected to the I / O interface 405: an input section 406 including a keyboard, a mouse, etc.; an output section 407 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 408 including a hard disk, etc.; and a communication section 409 including a network interface card such as a local area network (LAN) card, a modem, etc. The communication section 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the I / O interface 405 as needed. A removable medium 411, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 410 as needed, so that a computer program read therefrom is installed into the storage section 408 as needed.

[0089] When the central processing unit 401 in the electronic device of this embodiment executes the program, the following is achieved: Figure 1 The steps in the transformer core loss prediction method based on empirical formulas and models are shown.

[0090] In particular, according to an embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a computer readable medium, the computer program including a computer program for executing Figure 1 In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 409, and / or installed from the removable medium 411. When the computer program is executed by the central processing unit 401, various functions defined in the apparatus of the present application are executed.

[0091] in, Figure 1 The computer program instructions corresponding to the method shown may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0092] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0093] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A transformer core loss prediction method based on empirical formulas and models, characterized in that: include: Considering the influence of temperature on core loss, the proportionality coefficient in the Steinmetz equation is taken as a function of temperature and represented by a set exponential function. The power exponents of the frequency of the excitation current and the amplitude of the magnetic flux density in the Steinmetz equation are both taken as functions of temperature and represented by a set polynomial function. The Steinmetz equation taking into account the temperature correction factor is preliminarily obtained through fitting. Considering the influence of different excitation waveforms on the core loss, by introducing the equivalent sinusoidal frequency and its calculation formula, the non-sinusoidal excitation current waveform is equivalent to the sinusoidal excitation current waveform for calculation, and then the Steinmetz equation taking into account the temperature correction factor is corrected to obtain the corrected empirical formula; Based on the temperature, excitation current, magnetic flux density and the frequency of the excitation current, the empirical prediction value of the core loss is obtained using the modified empirical formula; The pre-trained neural network model is used to process the empirical prediction value of the core loss and its corresponding temperature, excitation current, flux density and frequency of the excitation current to obtain the final prediction value of the core loss.

2. The transformer core loss prediction method based on empirical formula and model according to claim 1, characterized in that: The exponential function of the proportionality coefficient in the Steinmetz equation is represented by the exponent e.

3. The transformer core loss prediction method based on empirical formula and model according to claim 1, characterized in that: The power exponents of both the frequency of the exciting current and the amplitude of the magnetic flux density are characterized by quadratic polynomial functions.

4. The transformer core loss prediction method based on empirical formula and model according to claim 1, characterized in that: The characterization formula of the equivalent sinusoidal frequency and the modified empirical formula are: in, is the equivalent sinusoidal frequency of the excitation current; It is the peak-to-peak value of the magnetic flux density in one cycle, that is, the difference between the maximum value and the minimum value; is the rate of change of magnetic flux density with time; is the period of the excitation waveform; , ~ is the coefficient of the equation; is the actual frequency of the excitation current; is the transformer core loss; Indicates temperature; is the magnitude of the magnetic flux density.

5. The transformer core loss prediction method based on empirical formula and model according to claim 1, characterized in that: The neural network model adopts the LSTM-AM model.

6. A transformer core loss prediction system based on empirical formulas and models, characterized in that: include: A preliminary correction module is used to consider the influence of temperature on the core loss, taking the proportionality coefficient in the Steinmetz equation as a function of temperature and using a set exponential function to represent it, taking the frequency of the excitation current and the amplitude of the magnetic flux density in the Steinmetz equation as functions of temperature and using a set polynomial function to represent them, and preliminarily obtaining the Steinmetz equation taking into account the temperature correction factor through fitting; An empirical formula determination module is used to consider the influence of different excitation waveforms on the core loss. By introducing the equivalent sinusoidal frequency and its calculation formula, the non-sinusoidal excitation current waveform is equivalent to the sinusoidal excitation current waveform for calculation, and the Steinmetz equation taking into account the temperature correction factor is further corrected to obtain the corrected empirical formula; An empirical prediction value calculation module is used to obtain an empirical prediction value of core loss based on temperature, excitation current, magnetic flux density and frequency of the excitation current using a modified empirical formula; The core loss prediction module uses a pre-trained neural network model to process the empirical prediction value of the core loss and its corresponding temperature, excitation current, flux density and frequency of the excitation current to obtain the final core loss prediction value.

7. The transformer core loss prediction system based on empirical formula and model as claimed in claim 6, characterized in that: In the preliminary correction module, the exponential function of the proportionality coefficient in the Steinmetz equation is characterized by the e-exponent.

8. The transformer core loss prediction system based on empirical formula and model as claimed in claim 6, characterized in that: In the preliminary correction module, the power exponents of both the frequency of the excitation current and the amplitude of the magnetic flux density are characterized by quadratic polynomial functions.

9. The transformer core loss prediction system based on empirical formula and model as claimed in claim 6, characterized in that: In the empirical formula determination module, the characterization formula of the equivalent sinusoidal frequency and the modified empirical formula are respectively: in, is the equivalent sinusoidal frequency of the excitation current; It is the peak-to-peak value of the magnetic flux density in one cycle, that is, the difference between the maximum value and the minimum value; is the rate of change of magnetic flux density with time; is the period of the excitation waveform; , ~ is the coefficient of the equation; is the actual frequency of the excitation current; is the transformer core loss; Indicates temperature; is the magnitude of the magnetic flux density.

10. The transformer core loss prediction system based on empirical formula and model according to claim 6, characterized in that: In the core loss prediction module, the neural network model adopts the LSTM-AM model.

Images