Transformer mechanical state in-situ sensing method based on space magnetic field modal evolution

Abstract

The invention provides a transformer mechanical state original taste sensing method based on spatial magnetic field modal evolution, which comprises the following steps: extracting a steady-state magnetic field component and a vibration magnetic field component in a transformer, and obtaining a first detection feature and a second detection feature according to the steady-state magnetic field component, diagnosing whether the magnetic circuit state of the transformer core is normal based on the first detection feature and the second detection feature; and diagnosing whether the axial and radial vibration modes of the transformer winding are normal or not according to the vibration magnetic field component. According to the invention, the internal magnetic field space modal characteristics and the vibration frequency characteristics of the transformer are fused, and the external measurement signals of the transformer are strongly correlated with the change of the internal mechanical state, so that the fault types of the winding and the iron core can be effectively distinguished, the fault area can be preliminarily positioned, and the diagnosis accuracy is greatly improved.

Description

Technical Field

[0001] This invention belongs to the field of transformer winding and core fault detection technology, and particularly relates to an in-situ sensing method for transformer mechanical state based on spatial magnetic field mode evolution. Background Technology

[0002] As a core component of the power grid, the mechanical integrity of the transformer's internal windings and core is fundamental to ensuring safe operation. Existing condition monitoring technologies have the following main shortcomings: 1. Indirectness and Lag: Widely used oil chromatography analysis detects faults by detecting gases from the decomposition of insulating oil, which is a typical "post-mortem" diagnosis. It can only be detected after the insulation material has decomposed due to overheating or discharge, and cannot provide early warning for mechanical faults such as loose windings or deformation. 2. Insensitive to mechanical conditions: While electrical tests (such as short-circuit impedance and frequency response analysis) can reflect winding deformation, they usually require a power outage and cannot achieve continuous online monitoring. Although vibration analysis is conducted online, it is susceptible to background noise interference, and the sensors are usually mounted on the tank base, making them insensitive to slight changes in the internal mechanical conditions of the windings. 3. Lack of spatial location capability: Most existing methods provide an overall condition assessment, but it is difficult to accurately locate whether the fault occurs on the high-voltage side, low-voltage side or core, and lacks the ability to spatially distinguish the internal condition of the transformer. These existing technologies are inherently incapable of providing direct, early, and in-situ sensing of the mechanical state of transformer windings and core. Summary of the Invention

[0003] In response to the problems raised in the background technology, this invention proposes a method for original sensing of the mechanical state of a transformer based on the spatial magnetic field mode evolution.

[0004] To achieve this objective, the present invention adopts the following technical solution: A method for sensing the original mechanical state of a transformer based on spatial magnetic field mode evolution includes: Step A: According to the array deployment principle, multiple sensors are deployed inside the transformer. Each sensor collects the time-varying magnetic field signal of the transformer and summarizes it into a raw spatiotemporal dataset. Step B: Extract the steady-state magnetic field component and the vibrational magnetic field component based on the original spatiotemporal dataset; Extracting steady-state magnetic field components includes extracting the total harmonic distortion (THD) of the magnetic field and processing the THD to construct a THD data sequence. Extracting vibration magnetic field components involves constructing a spatial snapshot to reflect the spatial distribution of the vibration magnetic field at any given moment, and constructing a spatiotemporal matrix based on spatial snapshots from multiple moments. Step C: Obtain the first detection feature and the second detection feature based on the total harmonic distortion rate data sequence of the magnetic field, and diagnose whether the magnetic circuit status of the transformer core is normal based on the first detection feature and the second detection feature. When the magnetic circuit of the transformer core is abnormal, the core abnormality warning is triggered, and step D is not executed; When the transformer core magnetic circuit is in normal condition, proceed to step D; Step D: Extract spatial vibration modes from the spatiotemporal matrix based on intrinsic orthogonal decomposition (POD), calculate the modal vibration frequencies and construct vibration profiles based on the spatial vibration modes, and diagnose whether the axial and radial vibration modes of the transformer windings are normal based on the modal vibration frequencies and vibration profiles. Step E: Output the diagnostic results of the transformer core magnetic circuit state and the axial and radial vibration modes of the transformer windings.

[0005] Preferably, in step A, multiple sensors are arranged inside the transformer according to the array layout principle, including: Three vertical measuring lines are set along the height direction on the side of the transformer tank wall. The three vertical measuring lines correspond to the axial projection positions of the high voltage winding, low voltage winding and iron core, respectively. Multiple high-precision triaxial magnetoresistive sensors are arranged at unequal intervals on each vertical measuring line. Among them, the number of high-precision triaxial magnetoresistive sensors installed at the ends and middle of the winding is greater than the number of high-precision triaxial magnetoresistive sensors installed at other locations of the winding.

[0006] Preferably, in step A, each sensor acquires the time-varying magnetic field signal of the transformer and summarizes it into an original spatiotemporal dataset, including: Each sensor performs continuous data acquisition at a sampling frequency higher than the highest vibration frequency of the transformer windings and core, thereby obtaining the time-varying magnetic field signal of each sensor. , This represents the time-varying magnetic field signal acquired by the i-th sensor at time t. The time-varying magnetic field signal includes three orthogonal components of the magnetic field vector. , M represents the total number of sensors. T represents the length of the sampling time window; The time-varying magnetic field signals collected by M sensors at all times within the sampling time window are aggregated into the original spatiotemporal dataset.

[0007] Preferably, the extraction of total harmonic distortion of the magnetic field includes: For each sensor in the original spatiotemporal dataset, all time-varying magnetic field signals within each cycle. Perform a Fourier transform to obtain the fundamental amplitude and the amplitudes of each harmonic. Calculate the total harmonic distortion rate of the magnetic field according to Formula 1. --Formula 1; in, This represents the total harmonic distortion rate of the magnetic field of the i-th sensor within one cycle. This represents the fundamental amplitude of the i-th sensor within one cycle. Let H represent the amplitude of the h-th harmonic of the i-th sensor within one cycle, where H represents the highest harmonic order within one cycle. ; This allows us to obtain the total harmonic distortion rate of the magnetic field for each sensor over multiple consecutive cycles.

[0008] Preferably, in step B, the total harmonic distortion (THD) of the magnetic field is processed to construct a THD data sequence, including: Set the size of the observation window and the base window to obtain N base windows. , Indicates the size of the observation window. Indicates the size of the base window; Each base window contains the total harmonic distortion rate of the magnetic field for multiple consecutive cycles for each sensor; For the total harmonic distortion (THD) of the magnetic field of all sensors within a basic window, the THD of the magnetic field belonging to the instantaneous interference field point within the basic window is removed, and the average of the remaining THD of the magnetic field within the basic window is taken as the representative THD of the basic window. Obtain representative total harmonic distortion (THD) data for all base windows to form a THD data sequence for an observation window. , This represents the total harmonic distortion of the magnetic field in the first basic window. This represents the representative total harmonic distortion rate of the magnetic field in the second fundamental window. This represents the total harmonic distortion rate of the representative magnetic field in the Nth basic window.

[0009] Preferably, step C includes: The first detection feature is calculated according to Formula 2, and the second detection feature is calculated according to Formula 3. --Formula 2; --Formula 3; This represents the mean offset of the first detection feature, namely the total harmonic distortion rate of the magnetic field. This represents the data sequence of total harmonic distortion rate of the magnetic field within the current observation window. The mean, This represents the long-term baseline mean of the total harmonic distortion rate (THD) data sequence of a healthy magnetic field obtained after prior measurements of a transformer in a healthy state. This represents the second detection characteristic, namely the dispersion ratio of the total harmonic distortion rate of the magnetic field. This represents the data sequence of total harmonic distortion rate of the magnetic field within the current observation window. standard deviation This represents the long-term baseline standard deviation of the total harmonic distortion rate (THD) data sequence of a healthy magnetic field obtained after prior measurements of a transformer in a healthy state. Set the mean outlier threshold and the discreteness anomaly threshold ; Determine the primary diagnostic criteria: Determine the second diagnostic criteria Determine the third diagnostic criteria , and This represents the lower limit and upper limit of mean outlier; When the second and / or third diagnostic conditions are met, it indicates that the transformer core magnetic circuit is abnormal, triggering a transformer core abnormality warning, and step D is not executed; If neither the second nor the third diagnostic condition is met, the transformer core magnetic circuit is normal, and step D is executed. If the first diagnostic criterion is met, the test is deemed invalid.

[0010] Preferably, in step B, extracting the vibration magnetic field components includes constructing a spatial snapshot reflecting the spatial distribution of the vibration magnetic field at any given time, and constructing a spatiotemporal matrix based on the spatial snapshots at multiple times, including: The time-varying magnetic field signals of each sensor in the original spatiotemporal dataset Digital bandpass filtering is performed to preserve the core vibration signal of each sensor. , This indicates that a digital bandpass filter is being used to remove time-varying magnetic field signals that are considered noise. Then, the time points are obtained by rearranging time t according to the continuity of time; For any time The vibration signal amplitudes of the same orthogonal component measured by all M sensors are used to form an M-dimensional vector. , , This indicates that after digital bandpass filtering, it is in Normal component of the Mth sensor at time M T denotes matrix transpose, which transforms an M-dimensional vector... Record it as a spatial snapshot; Take n consecutive time points , , ..., Spatial snapshot, Arranged in chronological order as follows A dimensional spacetime matrix X, .

[0011] Preferably, in step D, extracting spatial vibration modes from the spatiotemporal matrix based on the intrinsic orthogonal decomposition (POD) includes: Calculate the covariance matrix C based on the spatiotemporal matrix X according to Formula 4; --Formula 4; Eigenvalue decomposition of the covariance matrix C is performed according to Formula 5; --Formula 5; The j-th POD mode, i.e., the j-th spatial vibration mode, is represented as: A 3D feature vector, where each element corresponds to a sensor's vibration amplitude weight; The j-th eigenvalue represents the j-th POD mode. Contribution to total vibrational energy The larger the value, the higher the corresponding POD mode. The more important, ; All eigenvalues Arrange them in descending order, and take the first k eigenvalues ​​to obtain the k dominant POD modes. .

[0012] Preferably, in step D, calculating the modal vibration frequencies and constructing the vibration profile based on the spatial vibration modes includes: For a dominant POD mode, the modal time coefficient of the dominant POD mode is calculated according to Formula 6; --Formula Six; Indicates the dominant POD mode The modal time coefficients represent the modal time coefficients in the context of the time coefficients. The contribution of the j-th dominant POD mode in the vibration signal at time t is shown to change over time. Indicates the dominant POD mode Transpose of; Perform Fourier transform calculations on the modal time coefficients of a dominant POD mode to obtain the modal vibration frequencies of that dominant POD mode. ; For the dominant POD mode Interpolation is performed and visualized onto a two-dimensional projection of the transformer tank to form a vibration profile.

[0013] Preferably, diagnosing whether the axial and radial vibration modes of the transformer winding are normal based on the modal vibration frequencies and vibration profiles includes: Set axial first diagnostic conditions and axial second diagnostic conditions. The axial first diagnostic condition includes: for a dominant POD mode Its modal vibration frequency Less than the lower limit of the low frequency of the healthy modal vibration frequency reference The second diagnostic condition for the axial direction includes: for a dominant POD mode The phase difference between the amplitude of the high-voltage side of the winding and the amplitude of the low-voltage side of the winding is greater than the first preset phase difference, and the amplitude of the high-voltage side of the winding and the amplitude of the low-voltage side of the winding are out of phase according to the vibration profile. When a dominant POD mode If both the first and second axial diagnostic conditions are met, then the axial vibration mode of the transformer winding is determined to be abnormal, which is manifested as loosening of the axial clamping force of the transformer winding. There is no need to judge the radial vibration mode of the transformer winding. When a dominant POD mode If only the first or second axial diagnostic condition is met, no judgment is made on whether the axial vibration mode of the transformer winding is abnormal, and the radial vibration mode of the transformer winding is judged. When a dominant POD mode If neither the first nor the second axial diagnostic condition is met, the axial vibration mode of the transformer winding is determined to be normal, and the radial vibration mode of the transformer winding is then assessed. Set radial core diagnostic conditions, radial auxiliary first diagnostic conditions, and radial auxiliary second diagnostic conditions. The radial core diagnostic conditions include: for a dominant POD mode... Its modal vibration frequency In the mid-frequency range of the health modal vibration frequency reference Furthermore, the phase difference between the amplitudes of the corresponding windings on the same horizontal plane is less than or equal to the second preset phase difference, and the amplitudes of the windings on the same horizontal plane are in phase according to the vibration profile; the radial auxiliary first diagnostic condition includes: for a dominant POD mode The dominant POD mode does not exist in a transformer winding in a healthy state; the radial auxiliary second diagnostic condition includes: for a dominant POD mode Its eigenvalues Greater than the health characteristic value ; When a dominant POD mode When the radial core diagnostic conditions are met, the radial vibration mode of the transformer winding is directly determined to be abnormal, which is manifested as radial deformation of the transformer winding. When a dominant POD mode If the radial core diagnostic condition is not met but the radial auxiliary first diagnostic condition and the radial auxiliary second diagnostic condition are met, then the radial vibration mode of the transformer winding is determined to be abnormal, which is manifested as radial deformation of the transformer winding. When a dominant POD mode If none of the radial core diagnostic conditions, radial auxiliary first diagnostic conditions, and radial auxiliary second diagnostic conditions are met, then the radial vibration mode of the transformer winding is determined to be normal, and the presence of local high-frequency modes in the transformer winding is further determined. Set up a first diagnostic condition and a second diagnostic condition for high-frequency modes. The first diagnostic condition for high-frequency modes includes: for a dominant POD mode Its modal vibration frequency Greater than the upper limit of the high-frequency reference of the healthy modal vibration frequency The second diagnostic condition for high-frequency modes includes: for a dominant POD mode Its energy concentration coefficient is greater than the healthy energy concentration coefficient; When a dominant POD mode If both the first and second diagnostic conditions for high-frequency modes are met, it is determined that there is a local high-frequency mode in the transformer winding, which manifests as a local mechanical fault in the winding. When a dominant POD mode If the first diagnostic condition for high-frequency mode and / or the second diagnostic condition for high-frequency mode are not met, it is determined that there is no local high-frequency mode in the transformer winding.

[0014] The advantages of this invention over the prior art are: 1. Direct early warning: Breaking through the limitations of indirectness and lag in existing technologies, it directly captures early changes in the mechanical characteristics of windings and iron cores, thereby moving the fault warning point forward and preventing fault escalation.

[0015] 2. Precise diagnosis and localization: By combining dual-branch diagnosis with modal feature fusion, it can accurately distinguish between types such as core faults, axial loosening or radial deformation of windings, and can locate the fault area, thus solving the problem of fuzzy diagnosis in traditional methods.

[0016] 3. Easy to implement under all operating conditions: The external, non-intrusive sensors support 24-hour continuous online monitoring and are adaptable to various operating conditions; the diagnostic logic is based on clear physical principles, requiring no black-box model, and is easy to explain and implement. Attached Figure Description

[0017] Figure 1 This is a flowchart of a method for sensing the original mechanical state of a transformer based on the modal evolution of spatial magnetic field, according to the present invention. Detailed Implementation

[0018] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0019] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or end that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or ends.

[0021] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0022] like Figure 1 As shown, this application proposes a method for sensing the original mechanical state of a transformer based on the modal evolution of a spatial magnetic field, including: Step A: Following the array deployment principle, multiple sensors are deployed inside the transformer. Each sensor collects the time-varying magnetic field signal of the transformer and aggregates it into a raw spatiotemporal dataset, specifically including: Three vertical measuring lines are set along the height direction on the side of the transformer tank wall. The three vertical measuring lines correspond to the axial projection positions of the high voltage winding, low voltage winding and iron core, respectively. Multiple high-precision triaxial magnetoresistive sensors are arranged at unequal intervals on each vertical measuring line. Among them, the number of high-precision triaxial magnetoresistive sensors installed at the ends and middle of the winding is greater than the number of high-precision triaxial magnetoresistive sensors installed at other locations of the winding.

[0023] In this embodiment, by employing a high-precision triaxial magnetoresistive sensor, the three orthogonal components of the magnetic field vector can be measured simultaneously. In this system, multiple sensors are arranged at unequal intervals along each vertical measuring line, with a particular concentration at the ends and middle of the winding. The ends of the winding are areas of concentrated electrodynamic force. Because the magnetic field lines bend significantly at the ends of the winding and spread into a larger space, forming a so-called "edge effect," the magnetic field lines are most densely packed in this area, thus the leakage magnetic flux density B reaches its maximum value. According to the Ampere force formula... When the current I is constant, the radius B at the end is the largest, so the axial electrodynamic force (the force parallel to the winding axis) is also the largest there; this force attempts to compress or stretch the winding. In the central region of the winding height, the magnetic field lines are basically parallel, and their direction is mainly radial (perpendicular to the winding axis). At this point, the angle between the magnetic field direction (radial) and the current direction (axial, i.e., the direction of the winding conductors) is... Approaching 90 degrees, making (i.e., the maximum value). Although the value of B here may be slightly smaller than at the end, it results in a huge radial electrodynamic force (a force perpendicular to the winding cylinder wall) because its direction is the most "effective". This force attempts to expand the high-voltage winding and compress the low-voltage winding.

[0024] Preferably, in step A, each sensor acquires the time-varying magnetic field signal of the transformer and summarizes it into an original spatiotemporal dataset, including: Each sensor performs continuous data acquisition at a sampling frequency higher than the highest vibration frequency of the transformer windings and core, thereby obtaining the time-varying magnetic field signal of each sensor. , This represents the time-varying magnetic field signal acquired by the i-th sensor at time t. The time-varying magnetic field signal includes three orthogonal components of the magnetic field vector. , M represents the total number of sensors. T represents the length of the sampling time window; The time-varying magnetic field signals collected by M sensors at all times within the sampling time window are aggregated into the original spatiotemporal dataset.

[0025] In this embodiment, each sensor performs continuous data acquisition at a sampling frequency higher than the highest vibration frequency of the transformer windings and core. The sampling frequency can be set to 10 kHz, which is significantly higher than the 50 Hz highest vibration frequency of the transformer windings and core. This is based on the Nyquist-Shannon sampling theorem and the characteristics of the physical signal to be measured. The sampling theorem requires that to reproduce a signal without distortion, the sampling frequency must be greater than twice the highest frequency component of the signal. In engineering practice, at least 2.56 times or higher is typically required. Our target signal is not the power frequency, but the mechanical vibration it induces: the excitation source is the power frequency (50 Hz) and its harmonics. The electromagnetic force generated by the current is predominantly at 100 Hz (twice the power frequency). However, what we actually want to measure is the mechanical vibration generated by the windings and core under the action of electromagnetic force. The vibration frequency (natural frequency) of mechanical structures is usually much higher than the excitation frequency. The main frequency components of the mechanical vibration of transformer windings (such as the vibration of the windings, coils, and pressure plates) are typically between several hundred hertz (Hz) and several thousand hertz (kHz). For example, diagnosing axial loosening of windings requires capturing high-frequency vibration modes that may be as high as 800Hz to 2000Hz.

[0026] Therefore, within the same sampling time window, in the time length interval from 0 to T (e.g., 0-10s), each sensor will collect three orthogonal components at different times t (e.g., t1=0.1s, t2=0.2s, etc.) to form a time-varying magnetic field signal at a certain moment. The time-varying magnetic field signals collected by M sensors at all moments within the sampling time window are summarized into an original spatiotemporal dataset for subsequent analysis and judgment.

[0027] Step B: Extract the steady-state magnetic field component and the vibrational magnetic field component based on the original spatiotemporal dataset; Extracting steady-state magnetic field components includes extracting the total harmonic distortion (THD) of the magnetic field and processing the THD to construct a THD data sequence. In this embodiment, the steady-state magnetic field component is mainly determined by the main magnetic flux, and its harmonic content directly reflects the nonlinearity of the core magnetization process and is a sensitive indicator of the core's health status.

[0028] Preferably, the extraction of total harmonic distortion of the magnetic field includes: For each sensor in the original spatiotemporal dataset, all time-varying magnetic field signals within each cycle. Perform a Fourier transform to obtain the fundamental amplitude and the amplitudes of each harmonic. Calculate the total harmonic distortion rate of the magnetic field according to Formula 1. --Formula 1; in, This represents the total harmonic distortion rate of the magnetic field of the i-th sensor within one cycle. This represents the fundamental amplitude of the i-th sensor within one cycle. Let H represent the amplitude of the h-th harmonic of the i-th sensor within one cycle, where H represents the highest harmonic order within one cycle. ; This allows us to obtain the total harmonic distortion rate of the magnetic field for each sensor over multiple consecutive cycles.

[0029] In this embodiment, the essence of Fourier transform is "analyzing the frequency components of a signal over a continuous time period". Therefore, for a single sensor, by analyzing the signal over a continuous time period (one cycle), the total harmonic distortion (THD) of the magnetic field of that single sensor can be obtained. According to the continuity of time, we can obtain the THD of the magnetic field for multiple cycles of a single sensor, and then obtain the THD of the magnetic field for multiple cycles of all sensors. Since all sensors are based on the same sampling time window and sample simultaneously, the THD of the magnetic field for each cycle of each sensor is corresponding in time, that is, each sensor has its own THD within the same cycle.

[0030] The purpose of obtaining the total harmonic distortion rate of the magnetic field is to determine the health status of the transformer core magnetic circuit. The deterioration of the core condition will directly lead to the nonlinear enhancement of its magnetization characteristics, which is manifested as an increase in the value of the total harmonic distortion rate of the magnetic field.

[0031] Preferably, in step B, the total harmonic distortion (THD) of the magnetic field is processed to construct a THD data sequence, including: Set the size of the observation window and the base window to obtain N base windows. , Indicates the size of the observation window. Indicates the size of the base window; Each base window contains the total harmonic distortion rate of the magnetic field for multiple consecutive cycles for each sensor; For the total harmonic distortion (THD) of the magnetic field of all sensors within a basic window, the THD of the magnetic field belonging to the instantaneous interference field point within the basic window is removed, and the average of the remaining THD of the magnetic field within the basic window is taken as the representative THD of the basic window. Obtain representative total harmonic distortion (THD) data for all base windows to form a THD data sequence for an observation window. , This represents the total harmonic distortion of the magnetic field in the first basic window. This represents the representative total harmonic distortion rate of the magnetic field in the second fundamental window. This represents the total harmonic distortion rate of the representative magnetic field in the Nth basic window.

[0032] In this embodiment, to ensure the stability and anti-interference capability of subsequent judgments, the total harmonic distortion (THD) data of the magnetic field needs to be preprocessed, for example, through smoothing filtering using a sliding window. A time length for an observation window is set, such as 5 minutes, and a time length for a base window is set, such as 10 seconds. This results in 30 base windows, each containing multiple periods of THD from all sensors. For each base window, instantaneous interference outliers, such as THD represented by spikes caused by electromagnetic pulses, are removed. The remaining THD is then averaged or averaged to obtain the representative THD of that base window. This results in the representative THD of each base window, forming a THD data sequence for one observation window. .

[0033] Step C: Obtain the first detection feature and the second detection feature based on the total harmonic distortion rate data sequence of the magnetic field, and diagnose whether the magnetic circuit status of the transformer core is normal based on the first detection feature and the second detection feature. When the magnetic circuit of the transformer core is abnormal, the core abnormality warning is triggered, and step D is not executed; When the transformer core magnetic circuit is in normal condition, proceed to step D; Preferably, step C includes: The first detection feature is calculated according to Formula 2, and the second detection feature is calculated according to Formula 3. --Formula 2; --Formula 3; This represents the mean offset of the first detection feature, namely the total harmonic distortion rate of the magnetic field. This represents the data sequence of total harmonic distortion rate of the magnetic field within the current observation window. The mean, This represents the long-term baseline mean of the total harmonic distortion rate (THD) data sequence of a healthy magnetic field obtained after prior measurements of a transformer in a healthy state. This represents the second detection characteristic, namely the dispersion ratio of the total harmonic distortion rate of the magnetic field. This represents the data sequence of total harmonic distortion rate of the magnetic field within the current observation window. standard deviation This represents the long-term baseline standard deviation of the total harmonic distortion rate (THD) data sequence of a healthy magnetic field obtained after prior measurements of a transformer in a healthy state. Set the mean outlier threshold and the discreteness anomaly threshold ; Determine the primary diagnostic criteria: Determine the second diagnostic criteria Determine the third diagnostic criteria , and This represents the lower limit and upper limit of mean outlier; When the second and / or third diagnostic conditions are met, it indicates that the transformer core magnetic circuit is abnormal, triggering a transformer core abnormality warning, and step D is not executed; If neither the second nor the third diagnostic condition is met, the transformer core magnetic circuit is normal, and step D is executed. If the first diagnostic criterion is met, the test is deemed invalid.

[0034] In this embodiment, a total harmonic distortion rate (THD) data sequence of the magnetic field is obtained within an observation window. Then, we calculate the observation window. Mean offset As the first detection feature, the observation window is calculated. Standard deviation As the second detection feature; then a preset mean anomaly threshold is set. and the discreteness anomaly threshold , Generally, 2 is chosen. The value is greater than Because the standard deviation naturally fluctuates more than the mean, we need to collect data from transformers in a healthy state before making a judgment to determine the lower limit of mean anomaly. and the upper limit of mean outliers These two values ​​represent the upper and lower limits of the iron core's health. Next, determine whether the first diagnostic criterion is met: When the first diagnostic condition is met, it means that the data collected by all sensors corresponds to the observation window. If the values ​​in the sequence are far below the health limit, the possible causes are sensor failure, signal circuit failure, or transformer extreme light load / no load (weak signal). In such cases, the detection can be determined to be invalid due to measurement system failure. If the first diagnostic criterion is not met, then it is necessary to determine whether the second diagnostic criterion is met. Or third diagnostic criteria If either or both of these conditions are met, an abnormality in the core magnetic circuit can be diagnosed. This may be due to loose clamps, damaged inter-laminar insulation, or localized overheating, triggering a core abnormality warning. The winding status diagnosis will be paused or marked, indicating that "the winding diagnosis result is unreliable, meaning there is no need to test the winding problem at this time." If neither the second nor the third diagnostic condition is met, it means that the core magnetic circuit is normal and the magnetic field source is reliable, and the winding problem needs to be diagnosed next. In a transformer, the magnetic field of the winding (mainly the leakage magnetic field) is determined by the current flowing through the winding and the permeability of the iron core. A healthy iron core will provide a stable magnetic circuit with good linearity. In this case, the vibration of the winding is mainly determined by its own mechanical state (compression force, deformation).

[0035] A faulty iron core (such as looseness, overheating, or inter-laminar short circuits) will cause changes in the magnetic permeability of the magnetic circuit, resulting in severe magnetic field distortion and generating a large number of additional harmonic magnetic fields. The vibration of the winding is a response to the magnetic field force, and the root cause of the winding vibration is electromagnetic force. (Magnetic field × Current); therefore, if a core fault causes distortion of the magnetic field B (e.g., a sharp increase in the harmonic content THD_B), the waveform of the electromagnetic force acting on the winding will also be distorted. This change in excitation force will directly lead to changes in the spectrum and amplitude of the winding vibration, even if the winding itself is mechanically intact. Therefore, when a problem is found in the core, it will inevitably lead to a problem in the winding as well. Naturally, there is no need to waste time checking the winding. Only when the core is fine should further testing be conducted to rule out problems with the winding.

[0036] Furthermore, extracting the vibration magnetic field components includes constructing a spatial snapshot reflecting the spatial distribution of the vibration magnetic field at any given time, and constructing a spatiotemporal matrix based on the spatial snapshots at multiple times, specifically including: The time-varying magnetic field signals of each sensor in the original spatiotemporal dataset Digital bandpass filtering is performed to preserve the core vibration signal of each sensor. , This indicates that a digital bandpass filter is being used to remove time-varying magnetic field signals that are considered noise. Then, the time points are obtained by rearranging time t according to the continuity of time; In this embodiment, after digital bandpass filtering, the time-varying magnetic field signals of each sensor in the original spatiotemporal dataset may have data from intermediate moments filtered out. Therefore, it is necessary to rearrange the original moments t according to the continuity of time to obtain the rearranged moments. ; For any time The vibration signal amplitudes of the same orthogonal component measured by all M sensors are used to form an M-dimensional vector. , , This indicates that after digital bandpass filtering, it is in Normal component of the Mth sensor at time M T denotes matrix transpose, which transforms an M-dimensional vector... Record it as a spatial snapshot; In this embodiment, at any given time The same component measured by all M sensors (preferably the normal component most sensitive to radial deformation) is selected. The vibration signal amplitudes are used to form an M-dimensional vector. This is a "spatial snapshot," reflecting the spatial distribution of the vibrating magnetic field at that moment. Take n consecutive time points , , ..., Spatial snapshot, Arranged in chronological order as follows A dimensional spacetime matrix X, ; In this embodiment, the time dimension of the spatiotemporal matrix X is... The spatiotemporal matrix X is a 3D matrix where M is the total number of sensors and n is the total number of time points. Each column of the matrix is ​​a spatial snapshot, which is the set of vibration magnetic field amplitudes of all sensors at a certain moment. Simply put, the spatiotemporal matrix X is a complete vibration magnetic field dataset of "space (sensor) × time (sampling point)", which records the vibration signals of all sensors at different times.

[0037] Step D: Extract spatial vibration modes from the spatiotemporal matrix based on intrinsic orthogonal decomposition (POD), calculate the modal vibration frequencies and construct vibration profiles based on the spatial vibration modes, and diagnose whether the axial and radial vibration modes of the transformer windings are normal based on the modal vibration frequencies and vibration profiles. Preferably, in step D, extracting spatial vibration modes from the spatiotemporal matrix based on the intrinsic orthogonal decomposition (POD) includes: Calculate the covariance matrix C based on the spatiotemporal matrix X according to Formula 4; --Formula 4; Eigenvalue decomposition of the covariance matrix C is performed according to Formula 5; --Formula 5; The j-th POD mode, i.e., the j-th spatial vibration mode, is represented as: A 3D feature vector, where each element corresponds to a sensor's vibration amplitude weight; The j-th eigenvalue represents the j-th POD mode. Contribution to total vibrational energy The larger the value, the higher the corresponding POD mode. The more important, ; All eigenvalues Arrange them in descending order, and take the first k eigenvalues ​​to obtain the k dominant POD modes. .

[0038] In this step, we perform intrinsic orthogonal decomposition (POD) on the spatiotemporal matrix X. The core function of POD is to extract a few dominant vibration modes from complex vibration data. These modes are independent of each other and can cover more than 90% of the vibration energy, serving as a simplified representation of the actual mechanical vibration of the winding. The essence of POD decomposition is eigenvalue decomposition of the covariance matrix C. The core function of the covariance matrix C is to describe the vibration correlation between different sensors, such as whether the vibrations of sensor 1 and sensor 2 are synchronized, or whether the vibration amplitudes of sensor 3 and sensor 5 are proportional. The element in the i-th row and j-th column of the matrix... This represents the "covariance of the vibration signals of sensor i and sensor j" (i.e., the correlation strength between the two).

[0039] Furthermore, after obtaining the covariance matrix C using Formula 4, eigenvalue decomposition is performed on the covariance matrix C according to Formula 5. The core logic of Formula 5 is to find a special set of vectors. This makes the covariance matrix C multiplied by After that, the result is equal to Multiply by a constant So this group These are the modes that can most concisely describe the spatial distribution of vibration; each Each element is an M-dimensional vector, and each element corresponds to a vibration amplitude weight of a sensor, for example: (All elements are positive and their values ​​are close), indicating that the spatial form of this mode is "synchronous vibration of all sensor positions" (corresponding to the overall up-and-down vibration of the winding); for example... (Some elements are positive, some are negative), indicating that the spatial form of this mode is "reverse vibration on both sides of the winding" (corresponding to winding twisting); for example... (The middle element is large, and the two side elements are small), which indicates that the spatial form of this mode is "local vibration in the middle of the winding" (corresponding to the twisting of the coil).

[0040] Furthermore, The size represents the corresponding The proportion of this vibration mode in the total vibration energy, that is: The larger the value, the more "dominant" the vibration mode, and the greater its impact on the overall vibration of the winding. The smaller the value, the "weak" the vibration mode, indicating it is a minor vibration or noise. After obtaining all POD modes, we need to filter out the dominant POD mode and extract all eigenvalues. Arrange them in descending order and take the first k eigenvalues. k is usually 2 to 4. The first 2 to 4 modes can capture more than 90% of the energy. The energy of the following modes is extremely low (e.g., less than 1%), which are noise or minor vibrations and can be ignored.

[0041] Preferably, in step D, calculating the modal vibration frequencies and constructing the vibration profile based on the spatial vibration modes includes: For a dominant POD mode, the modal time coefficient of the dominant POD mode is calculated according to Formula 6; --Formula Six; Indicates the dominant POD mode The modal time coefficients represent the modal time coefficients in the context of the time coefficients. The contribution of the j-th dominant POD mode in the vibration signal at time t is shown to change over time. Indicates the dominant POD mode Transpose of; Perform Fourier transform calculations on the modal time coefficients of a dominant POD mode to obtain the modal vibration frequencies of that dominant POD mode. ; For the dominant POD mode Interpolation is performed and visualized onto a two-dimensional projection of the transformer tank to form a vibration profile.

[0042] In this embodiment, after extracting the dominant POD mode, two key parameters need to be calculated (modal frequency). Modal spatial morphology (as described above) is necessary for subsequent comparison with health benchmarks for diagnosis. As shown; and for The interpretation is for each POD mode. A corresponding "natural vibration mode" of the winding, the modal frequency This refers to the vibration frequency of this mode, such as the axial vibration mode frequency of a healthy winding. If the windings become loose and their stiffness decreases, the frequency will drop, for example, to 400Hz; and A curve representing the overall vibration intensity of the winding as a function of time, for example, at a certain moment. A large value indicates that the overall vibration is very strong, while a small value indicates that the overall vibration is weak.

[0043] Furthermore, it will dominate the POD mode. By mapping the element values ​​to the actual position of the sensor on the tank wall, a vibration profile can be drawn, allowing a direct view of whether the vibration is "uniform," "symmetrical," or "locally concentrated" in space.

[0044] Step E: Output the diagnostic results of the transformer core magnetic circuit state and the axial and radial vibration modes of the transformer windings.

[0045] Preferably, diagnosing whether the axial and radial vibration modes of the transformer winding are normal based on the modal vibration frequencies and vibration profiles includes: Set axial first diagnostic conditions and axial second diagnostic conditions. The axial first diagnostic condition includes: for a dominant POD mode Its modal vibration frequency Less than the lower limit of the low frequency of the healthy modal vibration frequency reference The second diagnostic condition for the axial direction includes: for a dominant POD mode The phase difference between the amplitude of the high-voltage side of the winding and the amplitude of the low-voltage side of the winding is greater than the first preset phase difference, and the amplitude of the high-voltage side of the winding and the amplitude of the low-voltage side of the winding are out of phase according to the vibration profile. When a dominant POD mode If both the first and second axial diagnostic conditions are met, then the axial vibration mode of the transformer winding is determined to be abnormal, which is manifested as loosening of the axial clamping force of the transformer winding. There is no need to judge the radial vibration mode of the transformer winding. When a dominant POD mode If only the first or second axial diagnostic condition is met, no judgment is made on whether the axial vibration mode of the transformer winding is abnormal, and the radial vibration mode of the transformer winding is judged. When a dominant POD mode If neither the first nor the second axial diagnostic condition is met, the axial vibration mode of the transformer winding is determined to be normal, and the radial vibration mode of the transformer winding is then assessed. Set radial core diagnostic conditions, radial auxiliary first diagnostic conditions, and radial auxiliary second diagnostic conditions. The radial core diagnostic conditions include: for a dominant POD mode... Its modal vibration frequency In the mid-frequency range of the health modal vibration frequency reference Furthermore, the phase difference between the amplitudes of the corresponding windings on the same horizontal plane is less than or equal to the second preset phase difference, and the amplitudes of the windings on the same horizontal plane are in phase according to the vibration profile; the radial auxiliary first diagnostic condition includes: for a dominant POD mode The dominant POD mode does not exist in a transformer winding in a healthy state; the radial auxiliary second diagnostic condition includes: for a dominant POD mode Its eigenvalues Greater than the health characteristic value ; When a dominant POD mode When the radial core diagnostic conditions are met, the radial vibration mode of the transformer winding is directly determined to be abnormal, which is manifested as radial deformation of the transformer winding. When a dominant POD mode If the radial core diagnostic condition is not met but the radial auxiliary first diagnostic condition and the radial auxiliary second diagnostic condition are met, then the radial vibration mode of the transformer winding is determined to be abnormal, which is manifested as radial deformation of the transformer winding. When a dominant POD mode If none of the radial core diagnostic conditions, radial auxiliary first diagnostic conditions, and radial auxiliary second diagnostic conditions are met, then the radial vibration mode of the transformer winding is determined to be normal, and the presence of local high-frequency modes in the transformer winding is further determined. Set up a first diagnostic condition and a second diagnostic condition for high-frequency modes. The first diagnostic condition for high-frequency modes includes: for a dominant POD mode Its modal vibration frequency Greater than the upper limit of the high-frequency reference of the healthy modal vibration frequency The second diagnostic condition for high-frequency modes includes: for a dominant POD mode Its energy concentration coefficient is greater than the healthy energy concentration coefficient; When a dominant POD mode If both the first and second diagnostic conditions for high-frequency modes are met, it is determined that there is a local high-frequency mode in the transformer winding, which manifests as a local mechanical fault in the winding. When a dominant POD mode If the first diagnostic condition for high-frequency mode and / or the second diagnostic condition for high-frequency mode are not met, it is determined that there is no local high-frequency mode in the transformer winding.

[0046] In this embodiment, a specific example of the above diagnostic process is provided as follows: During the initial commissioning of a transformer or when the transformer is in a healthy state, it is necessary to complete the collection and calibration of three types of reference data: 1. Axial vibration modal reference: Calibration steps: Collect health data continuously for 72 hours (covering no-load, full-load and other operating conditions), and extract the first two dominant modes, Modal1 and Modal2, through POD decomposition; Modal identifier: Modal1 (i.e.) Corresponding largest eigenvalue (With an energy percentage ≥40%, it is the dominant axial vibration mode in a healthy state). Reference frequency: That is, the lower limit of the low frequency of the healthy modal vibration frequency reference; Spatial morphological benchmark: Modal1's (M-dimensional vector) Visualized on the tank wall, the amplitude of all sensors on the high-pressure side is positive (+), and the amplitude of all sensors on the low-pressure side is also positive (+), that is, they are vibrating in phase, and the phase difference between the amplitudes of the high-pressure side and the low-pressure side sensors is ≤30° (if it exceeds this, it is considered an unhealthy state). The aforementioned Modal1 is the dominant mode of axial vibration under healthy conditions. Referring to this dominant mode can be used as a health benchmark to determine whether there are any abnormalities in axial vibration.

[0047] 2. Radial vibration modal reference: Calibration steps: Same as above, extract the second dominant mode Modal2 (i.e. Corresponding to the second largest eigenvalue (Energy percentage ≥ 30%) Reference frequency: For example, 600Hz±20Hz; Spatial morphology benchmark: The absolute difference in amplitude of sensors within the same horizontal plane (e.g., sensors at corresponding heights on the left and right sides of the fuel tank) is ≤10% (i.e., symmetrical), and all sensor amplitudes have the same sign (no alternation of positive and negative).

[0048] The Modal2 mentioned above is the dominant radial vibration mode under healthy conditions. Referring to this dominant mode, it can be used as a health benchmark to determine whether there is any abnormality in radial vibration.

[0049] 3. Local high-frequency modal reference: Calibration steps: Statistically analyze the frequencies of all dominant modes in the health data, and take the maximum value as the "high-frequency threshold," such as the upper limit of the high-frequency reference for the vibration frequency of the health modes. ; High-frequency threshold: ; Energy concentration benchmark: No mode of energy is concentrated in 3 or fewer adjacent sensors (local area), that is, the sum of the amplitudes of the 3 largest adjacent sensors of a certain mode or the sum of the amplitudes of all sensors is ≤20% (if it exceeds this, it is considered energy concentration).

[0050] The above describes a local high-frequency mode in a healthy state. This dominant mode can be used as a health benchmark to determine whether a local high-frequency mode exists.

[0051] 1. In the actual judgment process, the following are the methods to determine whether the axial vibration mode of the transformer winding is normal: Real-time acquisition of a dominant POD mode ; First diagnostic criterion for axial direction: Less than ; The first diagnostic criterion for axial direction is: the phase difference between the high-pressure side and the low-pressure side is greater than 30°, and it is visualized as the amplitude of the high-pressure side being "+" and the low-pressure side being "-", that is, out of phase; Overall judgment: a. If the first and second axial diagnostic conditions are met, the fault is determined to be "loose axial clamping force of the winding". The fault location is the entire axial area on the high-voltage side or the low-voltage side, based on the two opposite sides. b. If only the first or second axial diagnostic condition is met (e.g., frequency 470Hz, but phase difference 25°), then axial looseness is not determined, and proceed to the next step of radial looseness assessment. c. If neither condition is met, the axial mode of the winding is normal, and proceed to the next step of judging the radial mode.

[0052] 2. In the actual judgment process, the following steps are taken to determine whether the radial vibration mode of the transformer winding is normal: Real-time acquisition of a dominant POD mode ; Radial core diagnostic criteria: Furthermore, the absolute difference in amplitude within the same horizontal plane (e.g., sensors at corresponding heights on the left and right sides of the fuel tank) is ≤10%, and all sensor amplitudes have the same sign (no alternation of positive and negative). For example, the amplitude of the left sensor is 0.9, and the amplitude of the right sensor is 0.1. Visually displaying a "bulge" on the left side satisfies the symmetry breaking condition. Radial assist first diagnostic criterion: It belongs to the new mode, and Modal3 appears in the real-time dominant mode that is not present in the healthy state; For example: when healthy, only Modal1 and Modal2 are present, and Modal3 appears in real time; Radial-assisted second diagnostic condition: a sharp increase in amplitude, with the real-time Modal2 eigenvalue being greater than the healthy eigenvalue. 1.3 times (amplitude increased by 30%); For example: health characteristic values ,real time This represents a dramatic increase in amplitude; Overall judgment: a. Meets the radial core diagnostic conditions; directly determine "winding radial deformation" (fault location: such as bulge in the left high voltage winding); b. If the radial core diagnostic conditions are not met, but the radial auxiliary first diagnostic conditions and radial auxiliary second diagnostic conditions are met, the condition is determined to be "winding radial deformation" (fault location: spatial energy concentration side of the new mode). c. If none of the three conditions are met, the radial mode of the winding is normal, and proceed to the next step of judging the local high-frequency mode.

[0053] 3. In the actual judgment process, the following steps are taken to determine whether local high-frequency modes exist in the transformer windings: Real-time acquisition of a dominant POD mode ; First diagnostic criterion for high-frequency modes: ; Second diagnostic condition for high-frequency modes: Real-time dominant POD mode The "energy concentration coefficient" is greater than 20% (healthy energy concentration coefficient). For example: The sum of the amplitudes of the three largest adjacent sensors of Modal3 is 0.8, the sum of the amplitudes of all sensors is 1.0, the energy concentration factor is 80%, which is greater than the healthy energy concentration factor of 20%, and the visual display shows that the energy is concentrated in the upper part of the high voltage winding, which belongs to "spatial energy concentration"; a. If the first diagnostic condition and the second diagnostic condition of high frequency mode are met, the "winding local mechanical fault" is determined. Fault location: the area corresponding to the three adjacent sensors where energy is concentrated, such as loose wire discs or displacement of pads in the upper part of the high voltage winding. b. If only the first or second diagnostic condition of the high-frequency mode is met, for example, the frequency is 1100Hz, but the energy concentration factor is 15%, a local fault is not determined. c. If neither of the two conditions is met, the winding is determined to be in a normal local state.

[0054] The technical principles of the present invention have been described above with reference to specific embodiments. These descriptions are merely for explaining the principles of the invention and should not be construed as limiting the scope of protection of the invention in any way. Based on this explanation, those skilled in the art can readily conceive of other specific embodiments of the invention without inventive effort, and these embodiments will all fall within the scope of protection of the present invention.

Claims

1. A method for sensing the original mechanical state of a transformer based on spatial magnetic field mode evolution, characterized in that: Step A: According to the array deployment principle, multiple sensors are deployed inside the transformer. Each sensor collects the time-varying magnetic field signal of the transformer and summarizes it into a raw spatiotemporal dataset. Step B: Extract the steady-state magnetic field component and the vibrational magnetic field component based on the original spatiotemporal dataset; Extracting steady-state magnetic field components includes extracting the total harmonic distortion (THD) of the magnetic field and processing the THD to construct a THD data sequence. Extracting vibration magnetic field components involves constructing a spatial snapshot to reflect the spatial distribution of the vibration magnetic field at any given moment, and constructing a spatiotemporal matrix based on spatial snapshots from multiple moments. Step C: Obtain the first detection feature and the second detection feature based on the total harmonic distortion rate data sequence of the magnetic field, and diagnose whether the magnetic circuit status of the transformer core is normal based on the first detection feature and the second detection feature. When the magnetic circuit of the transformer core is abnormal, the core abnormality warning is triggered, and step D is not executed; When the transformer core magnetic circuit is in normal condition, proceed to step D; Step D: Extract spatial vibration modes from the spatiotemporal matrix based on intrinsic orthogonal decomposition (POD), calculate the modal vibration frequencies and construct vibration profiles based on the spatial vibration modes, and diagnose whether the axial and radial vibration modes of the transformer windings are normal based on the modal vibration frequencies and vibration profiles. Step E: Output the diagnostic results of the transformer core magnetic circuit state and the axial and radial vibration modes of the transformer windings.

2. The method for original sensing of transformer mechanical state based on spatial magnetic field mode evolution according to claim 1, characterized in that: In step A, multiple sensors are deployed inside the transformer according to the array deployment principle, including: Three vertical measuring lines are set along the height direction on the side of the transformer tank wall. The three vertical measuring lines correspond to the axial projection positions of the high voltage winding, low voltage winding and iron core, respectively. Multiple high-precision triaxial magnetoresistive sensors are arranged at unequal intervals on each vertical measuring line. Among them, the number of high-precision triaxial magnetoresistive sensors installed at the ends and middle of the winding is greater than the number of high-precision triaxial magnetoresistive sensors installed at other locations of the winding.

3. The method for original sensing of transformer mechanical state based on spatial magnetic field mode evolution according to claim 1, characterized in that: In step A, each sensor acquires the time-varying magnetic field signal of the transformer and aggregates it into a raw spatiotemporal dataset, including: Each sensor performs continuous data acquisition at a sampling frequency higher than the highest vibration frequency of the transformer windings and core, thereby obtaining the time-varying magnetic field signal of each sensor. , This represents the time-varying magnetic field signal acquired by the i-th sensor at time t. The time-varying magnetic field signal includes three orthogonal components of the magnetic field vector. , M represents the total number of sensors. T represents the length of the sampling time window; The time-varying magnetic field signals collected by M sensors at all times within the sampling time window are aggregated into the original spatiotemporal dataset.

4. The method for original sensing of transformer mechanical state based on spatial magnetic field mode evolution according to claim 3, characterized in that: Extracting the total harmonic distortion rate of the magnetic field, including: For each sensor in the original spatiotemporal dataset, all time-varying magnetic field signals within each cycle. Perform a Fourier transform to obtain the fundamental amplitude and the amplitudes of each harmonic. Calculate the total harmonic distortion rate of the magnetic field according to Formula 1. --Formula 1; in, This represents the total harmonic distortion rate of the magnetic field of the i-th sensor within one cycle. This represents the fundamental amplitude of the i-th sensor within one cycle. Let H represent the amplitude of the h-th harmonic of the i-th sensor within one cycle, where H represents the highest harmonic order within one cycle. ; This allows us to obtain the total harmonic distortion rate of the magnetic field for each sensor over multiple consecutive cycles.

5. The method for original sensing of transformer mechanical state based on spatial magnetic field mode evolution according to claim 4, characterized in that: In step B, the total harmonic distortion (THD) of the magnetic field is processed to construct a THD data sequence, including: Set the size of the observation window and the base window to obtain N base windows. , Indicates the size of the observation window. Indicates the size of the base window; Each base window contains the total harmonic distortion rate of the magnetic field for multiple consecutive cycles for each sensor; For the total harmonic distortion (THD) of the magnetic field of all sensors within a basic window, the THD of the magnetic field belonging to the instantaneous interference field point within the basic window is removed, and the average of the remaining THD of the magnetic field within the basic window is taken as the representative THD of the basic window. Obtain representative total harmonic distortion (THD) data for all base windows to form a THD data sequence for an observation window. , This represents the total harmonic distortion of the magnetic field in the first basic window. This represents the representative total harmonic distortion rate of the magnetic field in the second fundamental window. This represents the total harmonic distortion rate of the representative magnetic field in the Nth basic window.

6. The method for original sensing of transformer mechanical state based on spatial magnetic field mode evolution according to claim 5, characterized in that: Step C includes: The first detection feature is calculated according to Formula 2, and the second detection feature is calculated according to Formula 3. --Formula 2; --Formula 3; This represents the mean offset of the first detection feature, namely the total harmonic distortion rate of the magnetic field. This represents the data sequence of total harmonic distortion rate of the magnetic field within the current observation window. The mean, This represents the long-term baseline mean of the total harmonic distortion rate (THD) data sequence of a healthy magnetic field obtained after prior measurements of a transformer in a healthy state. This represents the second detection characteristic, namely the dispersion ratio of the total harmonic distortion rate of the magnetic field. This represents the data sequence of total harmonic distortion rate of the magnetic field within the current observation window. standard deviation This represents the long-term baseline standard deviation of the total harmonic distortion rate (THD) data sequence of a healthy magnetic field obtained after prior measurements of a transformer in a healthy state. Set the mean outlier threshold and the discreteness anomaly threshold ; Determine the primary diagnostic criteria: Determine the second diagnostic criteria Determine the third diagnostic criteria , and This represents the lower limit and upper limit of mean outlier; When the second and / or third diagnostic conditions are met, it indicates that the transformer core magnetic circuit is abnormal, triggering a transformer core abnormality warning, and step D is not executed; If neither the second nor the third diagnostic condition is met, the transformer core magnetic circuit is normal, and step D is executed. If the first diagnostic criterion is met, the test is deemed invalid.

7. The method for original sensing of transformer mechanical state based on spatial magnetic field mode evolution according to claim 3, characterized in that: In step B, extracting the vibration magnetic field components includes constructing a spatial snapshot reflecting the spatial distribution of the vibration magnetic field at any given time, and constructing a spatiotemporal matrix based on the spatial snapshots at multiple times, including: The time-varying magnetic field signals of each sensor in the original spatiotemporal dataset Digital bandpass filtering is performed to preserve the core vibration signal of each sensor. , This indicates that a digital bandpass filter is being used to remove time-varying magnetic field signals that are considered noise. Then, the time points are obtained by rearranging time t according to the continuity of time; For any time The vibration signal amplitudes of the same orthogonal component measured by all M sensors are used to form an M-dimensional vector. , , This indicates that after digital bandpass filtering, it is in Normal component of the Mth sensor at time M T denotes matrix transpose, which transforms an M-dimensional vector... Record it as a spatial snapshot; Take n consecutive time points , , ..., Spatial snapshot, Arranged in chronological order as follows A dimensional spacetime matrix X, .

8. The method for original sensing of transformer mechanical state based on spatial magnetic field mode evolution according to claim 7, characterized in that: In step D, spatial vibration modes are extracted from the spatiotemporal matrix based on the intrinsic orthogonal decomposition (POD), including: Calculate the covariance matrix C based on the spatiotemporal matrix X according to Formula 4; --Formula 4; Eigenvalue decomposition of the covariance matrix C is performed according to Formula 5; --Formula 5; The j-th POD mode, i.e., the j-th spatial vibration mode, is represented as: A 3D feature vector, where each element corresponds to a sensor's vibration amplitude weight; The j-th eigenvalue represents the j-th POD mode. Contribution to total vibrational energy The larger the value, the higher the corresponding POD mode. The more important, ; All eigenvalues Arrange them in descending order, and take the first k eigenvalues ​​to obtain the k dominant POD modes. .

9. The method for original sensing of transformer mechanical state based on spatial magnetic field mode evolution according to claim 8, characterized in that: In step D, the modal vibration frequencies are calculated based on the spatial vibration modes, and a vibration profile is constructed, including: For a dominant POD mode, the modal time coefficient of the dominant POD mode is calculated according to Formula 6; --Formula Six; Indicates the dominant POD mode The modal time coefficients represent the modal time coefficients in the context of the time coefficients. The contribution of the j-th dominant POD mode in the vibration signal at time t is shown to change over time. Indicates the dominant POD mode Transpose of; Perform Fourier transform calculations on the modal time coefficients of a dominant POD mode to obtain the modal vibration frequencies of that dominant POD mode. ; For the dominant POD mode Interpolation is performed and visualized onto a two-dimensional projection of the transformer tank to form a vibration profile.

10. The method for original sensing of transformer mechanical state based on spatial magnetic field mode evolution according to claim 9, characterized in that: Diagnosing the axial and radial vibration modes of transformer windings based on modal vibration frequencies and vibration profiles includes: Set axial first diagnostic conditions and axial second diagnostic conditions. The axial first diagnostic condition includes: for a dominant POD mode Its modal vibration frequency Less than the lower limit of the low frequency of the healthy modal vibration frequency reference The second diagnostic condition for the axial direction includes: for a dominant POD mode The phase difference between the amplitude of the high-voltage side of the winding and the amplitude of the low-voltage side of the winding is greater than the first preset phase difference, and the amplitude of the high-voltage side of the winding and the amplitude of the low-voltage side of the winding are out of phase according to the vibration profile. When a dominant POD mode If both the first and second axial diagnostic conditions are met, then the axial vibration mode of the transformer winding is determined to be abnormal, which is manifested as loosening of the axial clamping force of the transformer winding. There is no need to judge the radial vibration mode of the transformer winding. When a dominant POD mode If only the first or second axial diagnostic condition is met, no judgment is made on whether the axial vibration mode of the transformer winding is abnormal, and the radial vibration mode of the transformer winding is judged. When a dominant POD mode If neither the first nor the second axial diagnostic condition is met, the axial vibration mode of the transformer winding is determined to be normal, and the radial vibration mode of the transformer winding is then assessed. Set radial core diagnostic conditions, radial auxiliary first diagnostic conditions, and radial auxiliary second diagnostic conditions. The radial core diagnostic conditions include: for a dominant POD mode... Its modal vibration frequency In the mid-frequency range of the health modal vibration frequency reference Furthermore, the phase difference between the amplitudes of the corresponding windings on the same horizontal plane is less than or equal to the second preset phase difference, and the amplitudes of the windings on the same horizontal plane are in phase according to the vibration profile; the radial auxiliary first diagnostic condition includes: for a dominant POD mode The dominant POD mode does not exist in a transformer winding in a healthy state; the radial auxiliary second diagnostic condition includes: for a dominant POD mode Its eigenvalues Greater than the health characteristic value ; When a dominant POD mode When the radial core diagnostic conditions are met, the radial vibration mode of the transformer winding is directly determined to be abnormal, which is manifested as radial deformation of the transformer winding. When a dominant POD mode If the radial core diagnostic condition is not met but the radial auxiliary first diagnostic condition and the radial auxiliary second diagnostic condition are met, then the radial vibration mode of the transformer winding is determined to be abnormal, which is manifested as radial deformation of the transformer winding. When a dominant POD mode If none of the radial core diagnostic conditions, radial auxiliary first diagnostic conditions, and radial auxiliary second diagnostic conditions are met, then the radial vibration mode of the transformer winding is determined to be normal, and the presence of local high-frequency modes in the transformer winding is further determined. Set up a first diagnostic condition and a second diagnostic condition for high-frequency modes. The first diagnostic condition for high-frequency modes includes: for a dominant POD mode Its modal vibration frequency Greater than the upper limit of the high-frequency reference of the healthy modal vibration frequency The second diagnostic condition for high-frequency modes includes: for a dominant POD mode Its energy concentration coefficient is greater than the healthy energy concentration coefficient; When a dominant POD mode If both the first and second diagnostic conditions for high-frequency modes are met, it is determined that there is a local high-frequency mode in the transformer winding, which manifests as a local mechanical fault in the winding. When a dominant POD mode If the first diagnostic condition for high-frequency mode and / or the second diagnostic condition for high-frequency mode are not met, it is determined that there is no local high-frequency mode in the transformer winding.

Images