Machine learning based intelligent recognition method for magnetizing inrush current

Abstract

The present application relates to the technical field of transformer, particularly to a kind of intelligent recognition method of excitation inrush current based on machine learning, the present application first constructs a kind of simulation model of high impedance transformer; the model is optimized to make it can output the waveform diagram of excitation inrush current in accordance with the specified input parameter experimental result;Again, this is used as the basis for simulation experiment, greatly reduces the difficulty and time cost of artificial experiment;The data of simulation experiment are processed, analyzed and obtained feature vector;The relationship between the feature vectors of different parameters is found by deep learning to feature vector;Finally, the multi-modal data set is fused at decision level, each parameter obtains decision result separately and then fused together to output the final waveform result according to the gain of positive and negative correlation;Realize the prediction and identification of excitation inrush current;As a basis for the suppression of excitation inrush current.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of transformers, and particularly relates to a magnetizing inrush current intelligent identification method based on machine learning. BACKGROUND

[0002] Magnetizing inrush current refers to a sudden large current phenomenon in an electric circuit, especially when circuit parameters do not match or a load suddenly changes in a power system. The causes of magnetizing inrush current mainly include the following aspects:

[0003] Transformer core saturation:

[0004] When the voltage of a transformer is restored after no-load closing or external fault removal, the maximum magnetic flux far exceeds the saturation magnetic flux of the transformer due to the influence of the remanence and the fact that the magnetic flux in the core cannot suddenly change, which leads to a sharp increase in magnetizing current. The magnetization characteristics of the transformer are closely related, and the more saturated the core is, the greater the magnetizing current required to generate a certain magnetic flux.

[0005] Circuit parameter mismatch:

[0006] When the inductive elements (such as transformers, coils, etc.) in the circuit do not match the circuit parameters of the entire system, a high peak voltage is easily generated when the load suddenly changes, which in turn causes a momentary surge in current.

[0007] System voltage variation:

[0008] When the system voltage changes, the magnetizing voltage is affected, thereby generating magnetizing inrush current. This change may be caused by factors such as load mutation, unstable power grid, etc.

[0009] Phase angle at the moment of closing:

[0010] The phase angle of the voltage at the moment of closing also affects the size of the magnetizing inrush current. In the most unfavorable closing moment, when the voltage instantaneous value is zero, the maximum magnetic flux density in the core can reach 2Φm, and at this time the saturation of the core will be very serious, and the value of the magnetizing current will increase significantly.

[0011] Conventional magnetizing inrush current detection methods include the second harmonic braking method, the waveform recognition method, and the voltage braking method. For example, a Chinese patent application with the publication number CN105021920A discloses a main transformer magnetizing inrush current multi-feature identification method, which mainly uses the second harmonic braking method to identify magnetizing inrush current. The above methods all identify the magnetizing inrush current after it occurs and then act, and cannot predict the occurrence of magnetizing inrush current in advance. Therefore, the present application proposes a magnetizing inrush current intelligent identification method based on machine learning. SUMMARY

[0012] In view of the above background problems, the present application is proposed.

[0013] Therefore, the problem to be solved by the present application is how to construct a magnetizing inrush current prediction model according to experimental data.

[0014] To solve the above technical problems, the present application provides the following technical solutions:

[0015] The first aspect of the present application provides a magnetizing inrush current intelligent identification method based on machine learning, comprising the following specific steps:

[0016] S1. Constructing a simulation model in simulation software according to the geometric and electrical parameters of the transformer;

[0017] S2. Data collection: collecting electrical data when the transformer is in normal operation, electrical data when the transformer is in no-load closing, and electrical data when the voltage is restored after external fault removal;

[0018] S3. Data preprocessing: cleaning, denoising and standardizing the collected raw data;

[0019] S4. Using the preprocessed data as input to perform simulation experiments in the simulation model, and comparing the results obtained from the simulation experiments with the actual experiments, optimizing the parameters of the simulation model until the fitting degree of the simulation results and the actual experiment results is greater than 0.9;

[0020] S5. Using the adjusted simulation model to perform multiple simulation experiments, obtaining experimental results under different parameters, and constructing a simulation data set R based on the experimental results;

[0021] S6. Extracting features from the simulation data set R, establishing a multi-modal feature vector of magnetizing inrush current, opening angle, closing angle, core saturation, time, temperature and residual magnetism size; and mapping the multi-modal feature vector to the same feature space to construct a multi-modal data set;

[0022] S7. Deep learning of the multi-modal data set to learn the potential correlation between different modal data sets;

[0023] S8. Decision-level fusion of the multi-modal data set to construct a magnetizing inrush current prediction model, and verifying the model;

[0024] S9. Using the prediction model to identify the magnetizing inrush current of the transformer in operation.

[0025] The geometric and electrical parameters of the transformer in step S1 include the model, core material, core structure, winding structure, size, basic material quantity of the transformer; and also include rated current, nominal frequency, voltage ratio, current ratio, efficiency and loss, insulation performance and temperature rise performance;

[0026] The core material includes density, magnetic permeability and saturation magnetic induction strength;

[0027] The winding structure also includes the number of turns, coil material and winding resistance;

[0028] And set boundary conditions according to the actual nameplate parameters of the transformer; Specifically, it includes magnetic field boundary, current boundary and voltage boundary.

[0029] The electrical data collected in step S2 during normal operation of the transformer includes operating voltage, operating current;

[0030] The electrical data when the transformer is closed under no-load includes opening angle, residual magnetism, closing angle, and magnetic bias.

[0031] The electrical data when the voltage is restored after external fault removal includes opening angle, residual magnetism, closing angle, and magnetic bias.

[0032] The data preprocessing steps in step S3 include data consistency check and deduplication, outlier processing, and normalization processing.

[0033] In step S4, the fitting degree T satisfies the following formula:

[0034]

[0035] Where y a is the actual value, y p is the predicted value, is the mean of the actual value; and T 2 is greater than 0.9. Establish the feature vector data set Q1 when the transformer is closed under no-load:

[0036] Q1=[θ d ,θ c ,S,t,T,B r ]

[0037] In the formula, θ d is the opening angle, θ c is the closing angle; S is the core saturation, t is the time, T is the temperature;

[0038] Where the opening angle θ d and the closing angle θ c are extracted from the current waveform; the time t is obtained by the built-in timer, and the temperature T is obtained by the temperature sensor; the calculation formula of the core saturation S is:

[0039]

[0040] Where S is the core saturation; B is the current magnetic flux density; B sat is the saturation magnetic flux density of the material, which can be obtained by consulting the core material and database data;

[0041] B=μH

[0042] μ = μ0μ r

[0043] Where μ0 is the vacuum permeability; μ r is the relative permeability of the material; H is the magnetic field strength; which is calculated by the formula:

[0044]

[0045] Where N is the number of turns of the coil; I is the current through the coil, L is the average length of the magnetic circuit;

[0046] B r is the residual magnetism; B r = μ0(μ r -1)H c ; Where H c is the coercive force.

[0047] The data set Q2 is established when the voltage recovers after the transformer external fault is removed:

[0048] The feature vector data set Q1 is established when the transformer is closed under no load:

[0049] Q2 = [θ d , θ c , S, t, T, B r ]

[0050] In the formula, θ d is the closing angle, θ c is the closing angle; S is the core saturation, t is the time, T is the temperature;

[0051] Where the closing angle θ d , the closing angle θ c is extracted by the current waveform; the time t is obtained by the built-in timer, the temperature T is obtained by the temperature sensor; the calculation formula of the core saturation S is:

[0052]

[0053] Where S is the core saturation; B is the current magnetic flux density; B sat is the saturation magnetic flux density of the material which can be obtained by consulting the core material and database data;

[0054] B = μH

[0055] μ = μ0μ r

[0056] Where μ0 is the vacuum permeability; μ r is the relative permeability of the material; H is the magnetic field strength; which is calculated by the formula:

[0057]

[0058] Where N is the number of turns of the coil, I is the current through the coil, and L is the average length of the magnetic circuit.

[0059] B r Is the residual magnetism; B r = mu0(mu r -1)H c ; Where H c Is the coercive force.

[0060] The convolutional neural network (CNN) is used to learn the established data sets Q1 and Q2 respectively, and the correlation between different modal data sets is determined.

[0061] The loss function is expressed by mean square error (MSE) as follows:

[0062]

[0063] Where n is the total number of samples, yi is the actual value of the i-th sample, Is the predicted value of the i-th sample.

[0064] The second aspect of the application provides a computer device, comprising a memory and a processor, the memory stores a computer program, and the processor implements the steps of the above-mentioned machine learning-based field current intelligent identification method when executing the computer program.

[0065] The third aspect of the application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of the above-mentioned machine learning-based field current intelligent identification method.

[0066] The application has the following advantages:

[0067] The application first constructs a simulation model of a high-impedance transformer, optimizes the model to output a waveform diagram of the field current that meets the experimental results according to the specified input parameters, and then performs simulation experiments based on the model, which greatly reduces the difficulty and time cost of manual experiments. BRIEF DESCRIPTION OF DRAWINGS

[0068] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without any creative effort.

[0069] Figure 1 A 3D simulation model of a medium-high impedance transformer for a machine learning-based intelligent excitation inrush identification method.

[0070] Figure 2 A 2D simulation model of a medium-high impedance transformer for a machine learning-based intelligent excitation inrush identification method.

[0071] Figure 3 A flowchart of a machine learning-based intelligent excitation inrush identification method. DETAILED DESCRIPTION

[0072] In order to make the above-mentioned purposes, features and advantages of the present application more apparent and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0073] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application. However, the present application can be practiced in other ways that are not the same as the description, and those skilled in the art can make similar extensions without departing from the spirit of the present application, so the present application is not limited to the specific embodiments disclosed below.

[0074] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an independent or alternative embodiment that excludes other embodiments.

[0075] Embodiment 1

[0076] The first aspect of the present application provides a machine learning-based intelligent excitation inrush identification method, comprising the following specific steps:

[0077] S1, constructing a simulation model in a simulation software according to the geometric and electrical parameters of the transformer;

[0078] The geometric and electrical parameters of the transformer in step S1 include the model, core material, core structure, winding structure, size and basic material quantity of the transformer; and also include rated current, nominal frequency, voltage ratio, current ratio, efficiency and loss, insulation performance and temperature rise performance;

[0079] The core material includes density, magnetic permeability and saturation magnetic induction strength.

[0080] The winding structure further includes the number of turns, coil material and winding resistance;

[0081] And according to the actual nameplate parameters of the transformer, the boundary conditions are set, including magnetic field boundary, current boundary and voltage boundary.

[0082] Specifically, the scheme is specifically introduced by taking a high-voltage built-in high-impedance transformer as an example:

[0083] First, the high-voltage built-in transformer includes a core, a high-voltage winding, a medium-voltage winding, a voltage regulating winding and a low-voltage winding; by setting a reactor inside the transformer oil tank, the impedance is improved by increasing the reactance of the transformer inlet; the key of this method is to effectively shield the leakage magnetic field generated by the reactor to reduce the stray loss generated in the structural parts and prevent local overheating; the hysteresis model of the high-voltage built-in high-impedance transformer is:

[0084]

[0085] In the formula: B is the magnetic induction intensity; H is the magnetic field intensity; B s is the saturation magnetic induction intensity; H c is the coercive force, which represents the magnetic field intensity required for the material to recover to a non-magnetic state after removing the external magnetic field. An adjustment model is constructed by taking K as the adjustment coefficient, which is as follows:

[0086] B'(H) = KB(H)

[0087] Wherein, B'(H) is the adjusted hysteresis model, and K is the adjustment coefficient;

[0088]

[0089] k1≤K≤k2;

[0090] In the above formula, H0 is a material characteristic constant, which is used to adjust the relationship between the magnetic induction intensity and the magnetic field intensity in the equation; generally, the material characteristic constant H0 of different types of materials such as silicon steel and ferrite is not the same, and the value of H0 can be obtained directly by selecting and implementing the data matched with the transformer material through querying the database or textbooks. It can also be calibrated in an experimental measurement manner; for example, the relationship between the magnetic induction intensity B and the magnetic field intensity H under different magnetic field intensities can be recorded by measuring the hysteresis loop of the material, and the relationship between the magnetic induction intensity B and the magnetic field intensity H under different temperatures, different frequencies and different material uniformity is collected synchronously through a contrast experiment; a specific magnetic field intensity value H1 is extracted from the experimental data as a preliminary estimate of the value of H0; the experimental data is fitted by using a nonlinear regression method to calculate the magnetic field intensity value H2 that best fits the experimental data.

[0091] For the instance that the material characteristic constant H0 cannot be obtained by querying the database or textbook, it is verified only by experiment; then

[0092]

[0093] k1≤K≤k2.

[0094] S2, data collection: collecting electrical data of the transformer in normal operation, electrical data of the transformer in no-load closing, and electrical data of the voltage recovery after external fault removal;

[0095] The electrical data of the transformer in normal operation collected in step S2 includes operating voltage and operating current;

[0096] The electrical data of the transformer in no-load closing includes closing angle, residual magnetism, closing angle, and magnetic bias;

[0097] The electrical data of the voltage recovery after external fault removal includes closing angle, residual magnetism, closing angle, and magnetic bias.

[0098] The operating voltage, operating current, and the relationship between the phase difference between voltage and current and time t and temperature T can be analyzed by collecting the electrical data of the transformer in normal operation; and the correlation between the closing angle, residual magnetism, closing angle, and magnetic bias and other parameters can be further explored by analyzing the electrical data in the two special states of transformer no-load closing and external fault removal.

[0099] S3, data preprocessing: cleaning, denoising, and standardizing the collected original data;

[0100] The steps of data preprocessing in step S3 include data consistency check and deduplication, abnormal value processing, and normalization processing.

[0101] S4, using the preprocessed data as input to perform simulation experiments in the simulation model, comparing the results obtained from the simulation experiments and actual experiments, optimizing the parameters of the simulation model until the fitting degree of the simulation experiment results and the actual experiment results is greater than 0.9;

[0102] The fitting degree T in step S4 satisfies the following formula:

[0103]

[0104] Wherein, y a is the actual value, y p is the predicted value, is the mean value of the actual value; and T 2 is greater than 0.9.

[0105] S5, using the adjusted simulation model to perform multiple simulation experiments, obtaining experimental results under different parameters, and constructing a simulation data set R based on the experimental results;

[0106] S6, feature extraction is performed on the simulation data set R, a multi-modal feature vector of the magnetizing inrush current and the opening angle, the closing angle, the core saturation, the time, the temperature and the residual magnetism is established; and the multi-modal feature vector is mapped to the same feature space to construct a multi-modal data set;

[0107] A feature vector data set Q1 of the transformer no-load closing is established:

[0108] Q1=[θ d ,θ c ,S,t,T,B sat ] d ,θ c ,S,t,T,B r ]

[0109] In the formula, θ d is the opening angle, θ c is the closing angle; S is the core saturation, t is the time, and T is the temperature;

[0110] Wherein the opening angle θ d , the closing angle θ c is extracted by the current waveform; the time t is obtained by the built-in timer, and the temperature T is obtained by the temperature sensor; the calculation formula of the core saturation S is:

[0111]

[0112] Wherein, S is the core saturation; B is the current magnetic flux density; B sat is the saturation magnetic flux density of the material, which can be obtained by consulting the core material and database data;

[0113] B=μH

[0114] μ=μ0μ r

[0115] Wherein, the core material: μ0 is the vacuum permeability; μ r is the relative permeability of the material; H is the magnetic field strength; and the formula is calculated as:

[0116]

[0117] Wherein N is the number of turns of the coil; I is the current through the coil, and L is the average length of the magnetic circuit;

[0118] B r is the residual magnetism; B r =μ0(μ r -1)H c ; wherein H c is the coercive force.

[0119] The data set Q2 when the voltage is restored after the transformer external fault is cleared is established:

[0120] The feature vector data set Q1 when the transformer is closed under no-load is established:

[0121] Q2 = [θ d , θ c , S, t, T, B r ]

[0122] In the formula, θ d is the closing angle, θ c is the closing angle; S is the core saturation, t is the time, T is the temperature;

[0123] Wherein the closing angle θ d , the closing angle θ c is extracted by the current waveform; the time t is obtained by the built-in timer, and the temperature T is obtained by the temperature sensor; the calculation formula of the core saturation S is:

[0124]

[0125] Wherein, S is the core saturation; B is the current magnetic flux density; B sat is the saturation magnetic flux density of the material, which can be obtained by consulting the core material and database data;

[0126] B = μH

[0127] μ = μ0μ r

[0128] Wherein, the core material: μ0 is the vacuum permeability; μ r is the relative permeability of the material; H is the magnetic field strength; calculated by the formula:

[0129]

[0130] Wherein N is the number of turns; I is the current through the coil, and L is the average length of the magnetic circuit;

[0131] B r is the residual magnetism; B r = μ0(μ r -1)H c ; wherein H c is the coercive force.

[0132] S7, deep learning is performed on the multi-modal data set to learn the potential correlation between different modal data sets; a convolutional neural network CNN is used to learn the established data sets Q1 and Q2 respectively to determine the correlation between different modal data sets;

[0133] The loss function is represented by mean square error MSE:

[0134]

[0135] wherein n is the total number of samples, yi is the actual value of the i-th sample, is the predicted value of the i-th sample.

[0136] S8, decision-level fusion is performed on the multi-modal data set to construct a prediction model of the magnetizing inrush current, and the model is verified;

[0137] S9, the prediction model is used to identify the magnetizing inrush current during transformer operation.

[0138] In this embodiment, first, a high-impedance transformer model of various different core structures is established through simulation software, and then a hysteresis model is constructed for each high-impedance transformer model to describe the hysteresis characteristics of the core. Then, the parameters of the actual transformer are input into the established hysteresis model, and the saturation value B s of magnetic induction and the coercive force H c value are determined according to the experimental data, which are substituted into the model to output the image of the hysteresis loop and compared with the hysteresis loop obtained by actual experiment, so as to verify the model.

[0139] Then, the magnetic flux density and magnetization intensity in the core are calculated based on the hysteresis model and the solving algorithm; the magnetic flux density B = μH, μ is the magnetic permeability of the core material, and H is the magnetic field intensity; the simulation model is run to obtain the distribution of the magnetic flux density and magnetization intensity in the core; the change of the magnetic flux density with time is calculated considering the hysteresis characteristics of the core, and the state quantity (saturation value B s of magnetic induction and the coercive force H c value) in the hysteresis model is updated and the adjustment coefficient K is continuously optimized, so that the fitting degree of the hysteresis loop image output by the model and the hysteresis loop obtained by actual experiment of the transformer is greater than 0.9, and a high-impedance transformer simulation model B'(H) considering the hysteresis characteristics is obtained.

[0140] The model is used as the basis for simulation experiment to explore the influence of each parameter on the magnetizing inrush phenomenon; specifically, the electrical data of the transformer during normal operation, the electrical data of the transformer during no-load closing, and the electrical data during voltage recovery after external fault removal are collected; including the closing angle, residual magnetism, closing angle, and magnetizing data; then the original data is cleaned, denoised and standardized by data preprocessing, and each data is unified to the same dimension; the preprocessed data is used as input to perform simulation experiment in the simulation model, and the simulation experiment result is compared with the actual experiment result, the parameters of the simulation model are optimized until the fitting degree of the simulation experiment result and the actual experiment result is greater than 0.9; and the final simulation model is obtained.

[0141] A multi-variable simulation experiment is performed using the adjusted simulation model and a simulation data set R is obtained; feature extraction is performed based on the simulation data set R; it is known that during the closing process, the opening angle, residual magnetism, closing angle and magnetic bias data all affect the waveform and amplitude of the field current surge; the above-mentioned parameters are used as feature quantities to establish different feature vector data sets for two states of transformer no-load closing and voltage recovery after transformer external fault removal; and are respectively expressed as:

[0142] Q1 = [θ d , θ c , S, t, T, B r ]

[0143] Q2 = [θ d , θ c , S, t, T, B r ]

[0144] In the formula, θ d is the opening angle, θ c is the closing angle; S is the core saturation, t is the time, and T is the temperature.

[0145] Wherein the opening angle θ d and the closing angle θ c are extracted from the current waveform; the time t is obtained by a built-in timer, and the temperature T is obtained by a temperature sensor; the calculation formula of the core saturation S is:

[0146]

[0147] Wherein, S is the core saturation; B is the current magnetic flux density; B sat is the saturation magnetic flux density of the material, which can be obtained by consulting the core material and database data;

[0148] B = μH

[0149] μ = μ0μ r

[0150] Wherein, the core material: μ0 is the vacuum permeability; μ r is the relative permeability of the material; H is the magnetic field strength; and the formula is calculated as:

[0151]

[0152] Wherein N is the number of turns of the coil; I is the current passing through the coil, L is the average length of the magnetic circuit; B r is the residual magnetism; B r = μ0(μ r -1)H c ; wherein H c is the coercive force.

[0153] The established data sets Q1 and Q2 are learned by using a convolutional neural network (CNN) to determine the correlation between different modal data sets; the value range of each parameter is determined according to several common excitation inrush waveforms; for example, if the closing angle is on the rising edge (0-90°) of the sine wave of the voltage waveform, the waveform of the excitation inrush current will be steeper and the amplitude will be higher; in addition, the angle position at the time of opening will affect the residual magnetism, if the magnetic flux is in a saturated state at the time of opening, the excitation inrush current will be larger and the waveform will be steeper at the time of closing; based on this, the user can divide different precisions for each excitation inrush waveform with typical characteristics according to the needs, and select a specific value range for each parameter; finally, the multi-modal data set is fused at the decision level; the decision result of each parameter is obtained first, and then the decision results of multiple parameters are fused to obtain the final result and output.

[0154] For example, the opening angle includes four states (A1, A2, A3, A4), the closing angle includes four states (B1, B2, B3, B4); the core saturation degree includes three states (C1, C2, C3), the time is divided into multiple levels with 30s as a gradient, the temperature is divided into multiple levels according to 100 degrees as a gradient, the residual magnetism includes four states (D1, D2, D3, D4); and the corresponding common excitation inrush current waveforms are a, b, c, d four waveforms; according to the foregoing simulation, it is known that the opening angle A1 A3 and the closing angle B1 B3 output a waveform; the opening angle A2 A4 and the closing angle B2 B4 output b waveform; and as time goes on, the temperature and the residual magnetism will be affected, which are negatively correlated, so they will have a negative gain on the waveform; and the core saturation degree and the residual magnetism will have a positive gain on the waveform; therefore, after each parameter obtains a decision result, the final waveform result is output according to the positive and negative correlation gains.

[0155] Embodiment 2

[0156] The embodiment provides a computer device, including a memory and a processor, the memory stores a computer program, and the processor implements the steps of the excitation inrush current intelligent identification method based on machine learning in embodiment 1 when executing the computer program.

[0157] The computer device can be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected by a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is configured to perform wired or wireless communication with an external terminal. The wireless communication can be achieved by WIFI, an operator network, NFC (Near Field Communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0158] Embodiment 3

[0159] The embodiment provides a computer readable storage medium, and a computer program is stored on the computer readable storage medium. The computer program is executed by a processor to implement the steps of the excitation inrush current intelligent identification method based on machine learning in the embodiment 1.

[0160] The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0161] In summary, the excitation inrush current intelligent identification method based on machine learning can quickly predict the waveform of the excitation inrush current based on the simulation model through a small number of sensor measurement parameters, and provides a research basis for subsequent suppression of the excitation inrush current.

[0162] It should be noted that the above examples are only used to illustrate the technical solutions of the present application but not limit the present application. Although the present application is described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalently replaced, without departing from the spirit and scope of the technical solutions of the present application, which should be covered in the scope of the claims of the present application.

Claims

1. A method for intelligent identification of excitation inrush current based on machine learning, characterized by: The specific steps include: S1. Construct a simulation model in simulation software according to the geometric and electrical parameters of the transformer; S2. Data collection: Collect electrical data during normal operation of the transformer, electrical data when the transformer is switched on without load, and electrical data when the voltage is restored after the external fault is removed; S3. Data preprocessing: cleaning, denoising and standardizing the collected raw data; S4. Conducting a simulation experiment in a simulation model using the preprocessed data as input, comparing the results of the simulation experiment with those of the actual experiment, and optimizing the parameters of the simulation model until the degree of fit between the simulation experiment results and the actual experiment results is greater than 0.9; S5. Use the adjusted simulation model to conduct multiple simulation experiments to obtain experimental results under different parameters, and use them to construct a simulation data set R; S6. Extract features from the simulation data set R to establish a multimodal feature vector of the magnetizing inrush current and the opening angle, closing angle, core saturation, time, temperature, and residual magnetism; and map the multimodal feature vectors into the same feature space to construct a multimodal data set. S7. Perform deep learning on multimodal datasets to learn the potential correlation between different modal datasets; S8. Perform decision-level fusion on the multimodal data set to construct a prediction model for excitation inrush current and verify the model; S9. Use the prediction model to identify the magnetizing inrush current during transformer operation; The data preprocessing steps in step S3 include data consistency check and deduplication, outlier processing, and normalization processing; The fitting degree T in step S4 2 Satisfies the following formula: ； Among them, y a is the actual value, y p is the predicted value, is the mean of the actual values; and satisfies T 2 Greater than 0.9; In step S2, electrical data of the transformer during normal operation is collected, including operating voltage and operating current; The electrical data of the transformer when it is closed under no-load conditions include opening angle, residual magnetism, closing angle, and bias magnetism; Electrical data when voltage is restored after external fault removal: opening angle, residual magnetism, closing angle, and bias magnetism; Establish the characteristic vector data set Q1 when the transformer is closed under no-load conditions: ； Where, is the opening angle, is the closing angle; S is the core saturation, t is the time, and T is the temperature; The gate angle , closing angle Through current waveform extraction; time t is obtained by the built-in timer, and temperature T is obtained by the temperature sensor; the calculation formula of S core saturation is: ； Among them, S is the core saturation; B is the current magnetic flux density; B sat The saturation flux density of the material can be obtained by consulting the core material and database data; ； Among them, the core material: μ0 is the vacuum permeability; μ r is the relative magnetic permeability of the material; H is the magnetic field strength; calculated by the formula: ； Where N is the number of turns in the coil; I is the current passing through the coil, and L is the average length of the magnetic circuit; B r is the residual magnetism; B r =μ0(μ r -1)H c ; Among them H c is the coercive force; Create data set Q2 when the voltage recovers after the transformer external fault is removed: ； Where, is the opening angle, is the closing angle; S is the core saturation, t is the time, and T is the temperature; The gate angle , closing angle Through current waveform extraction; time t is obtained by the built-in timer, and temperature T is obtained by the temperature sensor; the calculation formula of S core saturation is: ； Among them, S is the core saturation; B is the current magnetic flux density; B sat The saturation flux density of the material can be obtained by consulting the core material and database data; ； Among them, the core material: μ0 is the vacuum permeability; μ r is the relative magnetic permeability of the material; H is the magnetic field strength; calculated by the formula: ； Where N is the number of turns in the coil; I is the current passing through the coil, and L is the average length of the magnetic circuit; B r is the residual magnetism; B r =μ0(μ r -1)H c ; Among them H c is the coercive force; The convolutional neural network (CNN) is used to learn the established datasets Q1 and Q2 respectively to determine the correlation between datasets of different modalities. The loss function is expressed using mean square error MSE: ； Where n is the total number of samples, yi is the actual value of the i-th sample, is the predicted value of the i-th sample; specifically: According to the needs, different precisions are divided for each waveform of the excitation inrush current with typical characteristics, and a specific value range is selected for each parameter. Finally, the multimodal data set is fused at the decision level. The decision result is first obtained through each parameter separately, and then the decision results of multiple parameters are fused to obtain the final result and output.

2. The method for intelligent identification of excitation inrush current based on machine learning according to claim 1, characterized in that: The geometric and electrical parameters of the transformer in step S1 include the transformer model, core material, core structure, winding structure, size and basic material quantity; also include rated current, nominal frequency, voltage ratio, current ratio, efficiency and loss, insulation performance and temperature rise performance; The core material includes density, magnetic permeability, and saturation magnetic induction intensity; The winding structure also includes the number of turns, coil material and winding resistance; Boundary conditions are set according to actual nameplate parameters of the transformer, and the boundary conditions include magnetic field boundary, current boundary and voltage boundary.

3. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the intelligent identification method of excitation inrush current based on machine learning according to any one of claims 1 to 2 are implemented.

4. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the intelligent identification method of excitation inrush current based on machine learning according to any one of claims 1 to 2 are implemented.

Images