A performance evaluation method, device and equipment of a current transformer and a medium

Abstract

The application provides a performance evaluation method, device and equipment of a current transformer and a medium, and relates to the field of data processing. In the method, the current transformer is in a saturation edge excitation state through composite work excitation, and a magnetostrictive micro-vibration response signal and a guided wave propagation response signal are collected. Resonance modal variation characteristics, guided wave path propagation characteristics and local scattering characteristics are extracted to construct a magnetoacoustic modal feature vector sequence. The feature vector sequence is input into a magnetoacoustic-performance mapping model to obtain a ratio difference, a phase angle error, a saturation characteristic offset and a harmonic error prediction value, and to output a degradation category probability distribution. Based on the change trend of each performance prediction value over time and in combination with the degradation category probability distribution, a performance attenuation curve and a degradation category evolution track are generated, so that dynamic evaluation and degradation state judgment of the performance of the current transformer are realized. Implementation of the technical solution facilitates improvement of the performance evaluation accuracy of the current transformer.

Description

Technical Field

[0001] This application relates to the technical field of data processing, specifically to a method, apparatus, equipment, and medium for evaluating the performance of a current transformer. Background Technology

[0002] Against the backdrop of ever-increasing demands for distributed, digital, and highly reliable power grid operation, distribution systems are increasingly burdened with denser terminal connections and more complex load structures. The operating environment, including high-power power electronic devices, distributed renewable energy interfaces, and highly sensitive metering equipment, makes power quality fluctuations, high-frequency harmonic enrichment, and sudden electromagnetic disturbances commonplace. In this context, current transformers, as a crucial link in the metering and protection chain, directly impact the metering accuracy, protection reliability, and user-side energy efficiency management precision of the distribution system.

[0003] In existing power distribution systems, economical current transformers are widely deployed to reduce the overall cost of metering and protection equipment. However, these current transformers commonly suffer from hidden structural defects such as wide manufacturing tolerances, uneven winding processes, and microcracks or localized stress concentrations in the core laminations. While these defects may not cause significant deviations in electromagnetic performance initially, during long-term service, the continuous effects of high-temperature and humid environments, the cumulative impact of complex electromagnetic interference, and the periodic loading of nonlinear harmonic excitation lead to the slow migration of polarization channels within the insulating medium, gradual stress relaxation in the core, and accompanying localized demagnetization. This causes the internal structure to gradually deviate from its initial state, resulting in potential degradation that is difficult to observe directly. Traditional evaluation methods typically rely on steady-state electromagnetic indicators such as ratio difference and phase angle error, which can only detect degradation when the transformer performance has already significantly deteriorated. They cannot provide reliable judgments on early hidden degradation at the structural level and lack joint characterization methods that can reveal changes in time-domain and frequency-domain coupling behavior behind electromagnetic performance degradation. Consequently, the accuracy of performance evaluation throughout the entire lifespan of current transformers is low.

[0004] Therefore, there is an urgent need for a performance evaluation method, device, equipment, and medium for current transformers. Summary of the Invention

[0005] This application provides a method, apparatus, device, and medium for evaluating the performance of current transformers, which facilitates improving the accuracy of performance evaluation throughout the entire lifespan of current transformers.

[0006] The first aspect of this application provides a method for evaluating the performance of a current transformer, the method comprising:

[0007] By applying a composite working excitation to the current transformer through a programmable current source, the current transformer is made to be in an excitation state with saturation edge and non-periodic component under excitation conditions.

[0008] Under the excitation state of the current transformer, the magnetostrictive micro-vibration response signal and the guided wave propagation response signal excited by the composite working excitation are obtained;

[0009] Signal processing is performed on the magnetostrictive micro-vibration response signal and the guided wave propagation response signal, and features are extracted. Based on the extracted features, a magnetoacoustic mode feature vector sequence is constructed.

[0010] The magnetoacoustic mode feature vector sequence is input into a pre-constructed magnetoacoustic-performance mapping model with electromagnetic performance calibration results to obtain the ratio difference prediction value, phase angle error prediction value, saturation characteristic offset prediction value and harmonic error prediction value corresponding to the current transformer, and to obtain the degradation category probability distribution corresponding to the current transformer.

[0011] The changing trends of the ratio difference prediction value, the phase angle error prediction value, the saturation characteristic offset prediction value, and the harmonic error prediction value at different times are determined. Combined with the degradation category probability distribution, the performance degradation curve and degradation category evolution trajectory of the current transformer are formed, so as to complete the performance evaluation of the current transformer based on the performance degradation curve and the degradation category evolution trajectory.

[0012] Preferably, the step of acquiring the magnetostrictive micro-vibration response signal and guided wave propagation response signal excited by the composite working excitation under the excitation state of the current transformer specifically includes:

[0013] High-frequency narrowband acoustic excitation or sweep frequency acoustic excitation is injected into the current transformer through the transmitting end of the ultrasonic transducer array, so that the guided wave propagates along the iron core region, winding region and support frame region of the current transformer.

[0014] The guided wave propagation response signal is obtained through the receiving end of the ultrasonic transducer array and the piezoelectric sheet;

[0015] The magnetostrictive micro-vibration response signal generated by the current transformer after applying a working current containing the fundamental frequency component, higher harmonic components, and slow DC bias component is obtained by using laser vibration measurement points or fiber optic grating measurement points.

[0016] Preferably, the step of performing signal processing on the magnetostrictive micro-vibration response signal and the guided wave propagation response signal to extract resonant mode change features, guided wave path propagation features, and local scattering features, and constructing a magnetoacoustic mode feature vector sequence based on the resonant mode change features, the guided wave path propagation features, and the local scattering features, specifically includes:

[0017] Based on the time window division method synchronized with the composite working excitation, the magnetostrictive micro-vibration response signal and the guided wave propagation response signal are preprocessed to extract resonant mode change features containing magnetostrictive effect information, guided wave path propagation features containing overall structural propagation path information, and local scattering features containing local structural discontinuities.

[0018] According to the preset channel order, frequency band division method and path numbering method, the resonant mode change characteristics, the guided wave path propagation characteristics and the local scattering characteristics are combined to construct the magnetoacoustic mode feature vector sequence.

[0019] Preferably, the step of inputting the magnetoacoustic mode feature vector sequence into a pre-constructed magnetoacoustic-performance mapping model with electromagnetic performance calibration results to obtain the predicted ratio difference, phase angle error, saturation characteristic offset, and harmonic error values ​​corresponding to the current transformer, and obtaining the degradation category probability distribution corresponding to the current transformer, specifically includes:

[0020] Within the magnetoacoustic-performance mapping model, a high-dimensional nonlinear feature transformation is performed on the magnetoacoustic modal feature vector sequence based on the time modeling unit to form a shared latent space representation that can characterize short-term local modal perturbations and long-term structural evolution trends.

[0021] Based on the shared latent space representation, the shared latent space representation is mapped to the corresponding electromagnetic performance prediction values ​​through the ratio difference regression branch for outputting the ratio difference prediction value, the phase angle error regression branch for outputting the phase angle error prediction value, the saturation characteristic offset regression branch for outputting the saturation characteristic offset prediction value, and the harmonic error regression branch for outputting the harmonic error prediction value, respectively.

[0022] Determine the real-value scores for multiple degradation categories corresponding to the current transformer;

[0023] The real-value scores of each degradation category are normalized by classifying the branches to form the probability distribution of each degradation category under the same time window.

[0024] Preferably, determining the changing trends of the predicted ratio difference, the predicted phase angle error, the predicted saturation characteristic offset, and the predicted harmonic error at different times, and combining this with the degradation category probability distribution, forms the performance degradation curve and degradation category evolution trajectory of the current transformer. This allows for the performance evaluation of the current transformer based on the performance degradation curve and the degradation category evolution trajectory. Specifically, this includes:

[0025] The predicted values ​​of the ratio difference, the predicted values ​​of the phase angle error, the predicted values ​​of the saturation characteristic offset, and the predicted values ​​of the harmonic error are arranged in chronological order to form a sequence of performance prediction results covering the continuous excitation state of the current transformer.

[0026] Identify the changing trends between different time windows in the performance prediction result sequence, and form a degradation category probability distribution sequence by combining the degradation category probability distributions that correspond one-to-one with the time windows of the performance prediction result sequence.

[0027] By performing correlation analysis between the changing trend of the performance prediction result sequence and the dominant change of the degradation category probability distribution sequence, the performance degradation curve and the degradation category evolution trajectory are generated, enabling the current transformer to be determined in the following stages: normal operation stage, early stage of latent degradation, accelerated degradation stage, and near-failure stage.

[0028] Preferably, the method further includes:

[0029] In the model calibration stage, a training dataset is constructed for current transformer samples with different degradation mechanisms and different life stages. Each target magnetoacoustic mode feature vector in the training dataset corresponds one-to-one with the target ratio difference calibration value, the target phase angle error calibration value, the target saturation characteristic offset calibration value, and the target harmonic error calibration value, and has a degradation category calibration label.

[0030] During the model training phase, a magnetoacoustic-performance mapping model with regression and classification branches is trained using the training dataset. The regression branch outputs the target ratio error calibration value, target phase angle error calibration value, target saturation characteristic offset calibration value, and target harmonic error calibration value corresponding to the target magnetoacoustic modal feature vector. The classification branch outputs the target degradation category probability distribution corresponding to the target magnetoacoustic modal feature vector.

[0031] Preferably, the method further includes:

[0032] Based on the rate of change and acceleration of change corresponding to the performance degradation curve, and the degree of abrupt change in the probability of key degradation categories in the degradation category evolution trajectory, the sampling period and sampling density of the magnetoacoustic excitation sequence are dynamically adjusted.

[0033] Based on the sampling period and sampling density of the adjusted magnetoacoustic excitation sequence, when the current transformer enters the implicit degradation acceleration stage or is nearing failure stage, the acquisition frequency of the magnetostrictive micro-vibration response signal and the guided wave propagation response signal is increased, and a failure risk warning is output.

[0034] A second aspect of this application provides a performance evaluation apparatus for a current transformer, the apparatus being used to perform the performance evaluation method for the current transformer, the apparatus comprising an acquisition module and a processing module, wherein...

[0035] The processing module is used to apply a composite working excitation to the primary side of the current transformer through a programmable current source, so that the current transformer is in an excitation state with saturation edge and non-periodic component under excitation conditions.

[0036] The acquisition module is used to acquire the magnetostrictive micro-vibration response signal and guided wave propagation response signal excited by the composite working excitation under the excitation state of the current transformer.

[0037] The processing module is further configured to perform signal processing on the magnetostrictive micro-vibration response signal and the guided wave propagation response signal to extract resonant mode change features, guided wave path propagation features and local scattering features, and to construct a magnetoacoustic mode feature vector sequence based on the resonant mode change features, the guided wave path propagation features and the local scattering features;

[0038] The processing module is further configured to input the magnetoacoustic mode feature vector sequence into a pre-constructed magnetoacoustic-performance mapping model with electromagnetic performance calibration results, to obtain the ratio difference prediction value, phase angle error prediction value, saturation characteristic offset prediction value and harmonic error prediction value corresponding to the current transformer, and to obtain the degradation category probability distribution corresponding to the current transformer.

[0039] The processing module is further configured to determine the changing trends of the ratio difference prediction value, the phase angle error prediction value, the saturation characteristic offset prediction value, and the harmonic error prediction value at different times, and combine them with the degradation category probability distribution to form the performance degradation curve and degradation category evolution trajectory of the current transformer, so as to complete the performance evaluation of the current transformer based on the performance degradation curve and the degradation category evolution trajectory.

[0040] A third aspect of this application provides an electronic device including a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, and both the user interface and the network interface are used to communicate with other devices. The processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method described above.

[0041] A fourth aspect of this application provides a non-transitory computer-readable storage medium storing instructions that, when executed, perform the method described above.

[0042] In summary, one or more technical solutions provided in this application have at least the following technical effects or advantages:

[0043] By simultaneously acquiring magnetostrictive micro-vibration response signals and guided wave propagation response signals under the actual excitation state of current transformers, latent degradation at the structural level can be captured before errors exceed limits. Resonance mode change characteristics, guided wave path propagation characteristics, and local scattering characteristics are extracted in the common dimension of the time domain and structural propagation path, and a magnetoacoustic mode feature vector sequence is constructed, forming a continuous and traceable unified representation between electromagnetic performance degradation and internal structural evolution. A magnetoacoustic-performance mapping model is used to establish the correlation between the magnetoacoustic mode feature vector sequence and the predicted values ​​of ratio difference, phase angle error, saturation characteristic offset, and harmonic error, while simultaneously outputting the degradation category probability distribution, ensuring a causal correspondence between performance offset and degradation mechanism. Based on the performance decay curve and degradation category evolution trajectory, life stage identification and latent degradation early warning are completed, enabling continuous prediction and interpretable location of degradation behavior of economical current transformers during long-term service, thereby improving metering and protection reliability and reducing maintenance and replacement lag. Therefore, this facilitates improved accuracy in performance evaluation throughout the entire lifespan of current transformers. Attached Figure Description

[0044] Figure 1 A flowchart illustrating a performance evaluation method for a current transformer provided in an embodiment of this application;

[0045] Figure 2 A schematic diagram of a performance evaluation device for a current transformer provided in an embodiment of this application;

[0046] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0047] Explanation of reference numerals in the attached figures: 21. Acquisition module; 22. Processing module; 31. Processor; 32. Communication bus; 33. User interface; 34. Network interface; 35. Memory. Detailed Implementation

[0048] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0049] In the description of the embodiments of this application, the words "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design that is described as "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design options. Rather, the use of the words "for example" or "for instance" is intended to present the relevant concepts in a specific manner.

[0050] In the description of the embodiments of this application, the term "multiple" means two or more. For example, multiple systems means two or more systems, and multiple screen terminals means two or more screen terminals. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0051] To address the aforementioned technical problems, this application provides a performance evaluation method for current transformers, referring to... Figure 1 , Figure 1 This is a flowchart illustrating a performance evaluation method for a current transformer provided in an embodiment of this application. The method is applied to a server and includes steps S110 to S150, as follows:

[0052] S110. Apply a composite working excitation to the primary side of the current transformer through a programmable current source so that the current transformer is in an excitation state with saturation edge and non-periodic component under excitation conditions.

[0053] Specifically, a server refers to a computing device or platform that executes control logic and data processing functions, such as an industrial computer or data center server equipped with an operating system, processor, memory, and communication interfaces. The server issues control commands, sets current waveform parameters and excitation condition parameters through a communication link with the lower-level programmable current source, and schedules and records the testing process of the current transformer when needed. For example, the server runs performance evaluation software in a power distribution testing laboratory, sending commands to the programmable current source in the rack via Ethernet or fieldbus, causing the programmable current source to excite the primary conductor of the current transformer according to the set current amplitude, phase, harmonic content, and time sequence.

[0054] Composite excitation refers to a current or voltage excitation formed by combining the fundamental frequency component, higher harmonic components, DC bias components, and possible short-term impulse components according to a specific amplitude ratio and phase relationship. In this technical solution, it specifically refers to the current excitation output by a programmable current source and applied to the primary side. Compared with a single sinusoidal current excitation, composite excitation can more realistically simulate the nonlinear loads, rectifier devices, and motor starting conditions existing in the power distribution system. For example, it can output a near-sinusoidal fundamental current for one test period, and superimpose a significant proportion of fifth and seventh harmonic currents for another period, inserting short-term current impulses to simulate motor restart, thereby reflecting the response of the current transformer under complex power quality environments.

[0055] The saturation edge refers to the region where the operating point of the iron core approaches the nonlinear saturation region of the magnetization curve of the magnetic material, but has not yet fully entered a deep saturation state. At the saturation edge, the permeability of the iron core begins to decrease, and the relationship between the incremental magnetic flux and the incremental excitation current deviates significantly from linearity, but still has a certain adjustment margin. In current transformers, if the peak value of the composite working excitation is high or a DC bias component is superimposed, part of the magnetic circuit of the iron core may enter the saturation edge near the peak value of the current waveform. At this time, the response to magnetostriction, magnetic noise, and harmonics will be much more sensitive than in the completely linear region. For example, in a current transformer with a normal rated current of a certain value, increasing the peak current to near the limit of the rated current and superimposing a DC bias may cause the iron core to be at the saturation edge at the current peak value.

[0056] Aperiodic components refer to those components in a composite excitation system that lack strict periodicity compared to an ideal periodic sine wave. These include DC bias components, aperiodic transient components, and random disturbance components. DC bias components cause the current waveform to deviate from its zero-symmetry position on the time axis; aperiodic transient components alter the waveform's symmetry and peak value for short periods; and random disturbance components superimpose irregular jitter onto the waveform. In power distribution systems, significant aperiodic components may appear in the current when three-phase loads are unbalanced, rectifier control strategies switch, or fault transients occur. This technical solution explicitly designs the aperiodic component in the programmable current source output, enabling the current transformer to exhibit more complete structural and electromagnetic responses under excitation conditions approaching real-world complex operating conditions.

[0057] S120. Under the excitation state of the current transformer, acquire the magnetostrictive micro-vibration response signal and guided wave propagation response signal excited by the composite working excitation.

[0058] Specifically, the magnetostrictive micro-vibration response signal refers to the time-series signal obtained by converting the minute mechanical vibrations generated in the structure of a current transformer under excitation due to the magnetostrictive effect of the core material into a sensor. Magnetostriction is a phenomenon in which the volume or shape of a ferromagnetic material undergoes minute changes due to the rearrangement of magnetic domains during magnetization. When the magnetic flux in the core changes with the combined working excitation, the core will produce extremely small expansion and contraction and bending in each excitation cycle. These minute deformations are transmitted between the core and the winding frame, insulation support, and outer shell, causing micron-level or even nanon-level vibrations in the entire or local areas of the current transformer.

[0059] The guided wave propagation response signal refers to the response signal generated in sensors located at different positions after a mechanical wave propagating inside a current transformer structure via guided waves is affected by structural discontinuities and material changes during propagation. In this embodiment, the guided wave is an ultrasonic wave or a mechanical wave in the ultrasonic frequency band propagating along the core, winding frame, or shell structure. It is injected into the structure by a dedicated ultrasonic transducer array. When the structure is continuous and uniform, the propagation time, attenuation degree, and waveform morphology of the guided wave along the predetermined path are relatively stable. However, when there are hidden structural defects such as winding slack, core microcracks, interlayer debonding, or insufficient bolt preload, the guided wave will be reflected, refracted, or scattered at the defect location, causing the guided wave propagation response signal at the receiving position to have an arrival time shift, abnormal amplitude attenuation, or waveform distortion.

[0060] Furthermore, in specific implementation, when injecting high-frequency narrowband acoustic excitation or frequency-sweeping acoustic excitation into the current transformer through the transmitting end of the ultrasonic transducer array, the acoustic coupling positions are first determined on the outer side of the core region, winding region, and support frame region of the current transformer. Several ultrasonic transducers are arranged circumferentially or axially on the current transformer shell as the transmitting end of the ultrasonic transducer array. The transmitting end is tightly attached to the current transformer shell using coupling adhesive or wedge structures to reduce acoustic energy loss. High-frequency narrowband acoustic excitation and frequency-sweeping acoustic excitation can be uniformly described as acoustic excitation signals with envelope modulation and frequency modulation. For example, the acoustic excitation signal can be constructed as follows:

[0061] ,

[0062] in, Indicates the time of the ultrasonic transducer array transmitter. The acoustic excitation signal output at the location, This represents the acoustic excitation amplitude, used to control the magnitude of guided wave energy. The value range is determined based on the structural bearing capacity and signal-to-noise ratio requirements of the current transformer. This represents a time window function used to limit the effective time interval of acoustic excitation and suppress sidelobe effects. Its value ranges from zero to one, and can be selected from Hanning window, cosine window or other smooth window. Indicates the fundamental center frequency, used to determine the frequency band of the dominant guided wave propagation mode; This represents the linear sweep rate, used to control the continuous change of frequency over time, so that the sweep acoustic excitation covers a wide frequency band. Indicates the first The frequency modulation coefficients of each modulation component are used to adjust the contribution of different modulation components to the instantaneous frequency disturbance, and the value is less than one to avoid excessive dispersion. Indicates the first The modulation frequency of each modulation component is used to introduce multi-scale frequency perturbations, so that the guided wave exhibits different sensitivities to structural defects at different spatial scales. Indicates the first The initial phase of each modulation component is used to adjust the phase relationship between different modulation components; This indicates the number of components participating in frequency modulation, which is set according to the requirements for spectral complexity; This represents the overall phase offset, used for phase alignment of the excitation waveform over time. By adjusting... , , , , These parameters allow guided waves to propagate along the core region, winding region, and support frame region in different frequency bands and modes, thereby enabling differentiated excitation of different types of latent structural defects.

[0063] In the acquisition of guided wave propagation response signals, some transducers in the ultrasonic transducer array are configured as receivers. Simultaneously, piezoelectric sheets are attached to the current transformer housing near the core, winding, and support frame areas to pick up surface vibrations generated by the guided waves on the structural surface. The voltage signals output from each receiver are connected to a multi-channel synchronous acquisition system via shielded cables. The multi-channel synchronous acquisition system operates at a sampling frequency... Samples are taken simultaneously from each receiving channel, and strictly time-aligned with the transmission time. For the first... The discrete guided wave propagation response signal of each receiving channel can be represented as a form containing the superposition of multiple propagation paths and noise perturbation:

[0064] ,

[0065] in, Indicates the first The receiving channel is in the first Discrete values ​​of the guided wave propagation response signal at each sampling point; Indicates the distance from the incentive position to the first position. The number of effective propagation paths that can be formed by a receiving location, including direct paths and indirect paths formed by reflection, refraction or scattering; Indicates the first The first receiving channel The magnitude coefficient of each path is used to describe the attenuation degree and coupling efficiency of the path in the structure; Indicates the first The first receiving channel The discrete propagation kernel corresponding to each path is used to describe the broadening, distortion and dispersion effects of the path on the excitation waveform. Its specific shape is determined by the structural material, geometry and defect state. Indicates the first The first receiving channel The discrete propagation delay of a path relative to the excitation time is used to reflect the changes in the propagation distance and propagation speed of the path; Indicates the first The noise terms for each receiving channel are used to describe electronic noise, environmental vibration, and unmodelable scattering contributions. This is achieved through... In multiple and Analyzing the distribution on the surface allows us to identify the dominant propagation paths along the core region, winding region, and support frame region, and to compare them with... , and The changes at different testing stages are used to extract the waveguide path propagation characteristics and local scattering characteristics.

[0066] In the acquisition of magnetostrictive micro-vibration response signals, laser vibration measurement points are placed near the corners of the iron core, the edges of the winding pressure plates, and the connection points of the support frame on the current transformer housing. Fiber grating measurement points are placed on the corresponding locations on the housing surface or the support frame surface, and are led out to the fiber optic demodulation equipment via optical fibers. During testing, the current transformer carries an operating current output from a programmable current source. This operating current includes the fundamental frequency component, higher harmonic components, and a slowly changing DC bias component, and can be superimposed with a slowly varying envelope and random disturbance components, which can be described as:

[0067] ,

[0068] in, Indicates time The operating current flowing through the primary side of the current transformer; It represents the amplitude of the fundamental frequency component and controls the main magnetization level of the current transformer; It represents the fundamental frequency of the power supply, which is equal to the rated frequency of the power distribution system; It represents the set of higher harmonic orders that participate in the superposition, such as the fifth, seventh, eleventh, etc. Indicates the first The amplitude of higher harmonic components is used to simulate current distortion caused by nonlinear loads or power electronic devices. Indicates the first The initial phase of higher harmonic components relative to the fundamental frequency component is used to describe the phase relationship between multiple frequency components. It represents the amplitude of the slow DC bias component, used to simulate the overall offset of the core operating point under conditions such as three-phase imbalance and magnetization bias. This represents the envelope modulation coefficient, used to control the degree of slow fluctuation in the amplitude of the operating current; This indicates the envelope modulation frequency, used to simulate slow load fluctuations or periodic switching of operating conditions; It represents random disturbance components and is used to describe the nondeterministic components of electromagnetic interference and current ripple in the field.

[0069] S130. Perform signal processing on the magnetostrictive micro-vibration response signal and the guided wave propagation response signal to extract resonant mode change characteristics, guided wave path propagation characteristics and local scattering characteristics, and construct a magnetoacoustic mode feature vector sequence based on the resonant mode change characteristics, guided wave path propagation characteristics and local scattering characteristics.

[0070] Specifically, resonant mode variation characteristics refer to the set of features that identify the variations in resonant frequency, resonant bandwidth, modal energy, and spatial mode shape of multiple resonant modes exhibited by a current transformer under different excitation conditions and structural states by analyzing the spectral and spatial distribution of the magnetostrictive micro-vibration response signal, and describe them parametrically. A resonant mode refers to the vibrational state of the entire structure at a specific frequency where energy is concentrated. Different resonant modes have different sensitivities to changes in stiffness and mass in different regions. When microcracks occur locally in the core or the winding support loosens, the resonant modes strongly coupled to these regions will exhibit a decrease in resonant frequency, an increase in resonant bandwidth, or a spatial redistribution of modal energy. These changes constitute the resonant mode variation characteristics.

[0071] Guided wave path propagation characteristics refer to the quantitative description of the propagation time, propagation speed, amplitude attenuation, and spectral changes of each effective guided wave propagation path from the excitation position to the receiving position by identifying the path and aligning the time of the guided wave propagation response signal, and tracking the changing trends of these parameters over time or with the service stage. Guided wave path propagation characteristics focus on characterizing the propagation behavior at the overall path level. For example, a guided wave path propagating along the interface of the core laminations has a stable propagation time and amplitude in the initial state, but exhibits a slight increase in propagation time and a significant attenuation in amplitude in the later stage of degradation due to changes in the contact condition between the laminations; another guided wave path propagating along the edge of the winding pressure plate may show a slight advance in arrival time and a shift in the spectral center of gravity after the winding clamping force is reduced.

[0072] Local scattering characteristics refer to the characteristic description of guided wave scattering behavior in guided wave propagation response signals caused by local discontinuities in the structure, abrupt changes in local material properties, and local stress concentrations, including local reflection, local mode conversion, and local energy diffusion. Unlike guided wave path propagation characteristics, which focus on the overall path, local scattering characteristics pay more attention to additional echo clusters, anomalous wave packets, or asymmetric waveform distortions that suddenly appear on a certain path within a specific time window. These signal components often directly correspond to local cracks, pores, debonded layers, or loose contact surfaces within the structure.

[0073] The magnetoacoustic modal feature vector sequence refers to the combination of resonant mode change features, guided wave path propagation features, and local scattering features extracted from each time window or test stage in multiple consecutive time windows or multiple test stages, according to a pre-set feature dimension and feature order. These multi-dimensional feature parameters are arranged into a feature vector, and the feature vectors of different time windows or different test stages are connected in chronological order to form a vector sequence describing the evolution trajectory of the magnetoacoustic characteristics of the current transformer during service or testing.

[0074] Furthermore, in the specific implementation process, firstly, based on the time window division method synchronized with the composite working excitation, when preprocessing the magnetostrictive micro-vibration response signal and the guided wave propagation response signal, the multi-channel magnetostrictive micro-vibration response signal is denoted as... The multi-channel guided wave propagation response signal is denoted as ,in Numbering of magnetostriction measurement points. Number the waveguide receiving channels; based on the periodic structure and event-triggered structure of the composite operating excitation, divide the time axis into several sets of non-overlapping or partially overlapping time windows. , order the The start and end times corresponding to each time window are: Within this time interval, the signals from each channel are truncated and resampled to obtain a discrete sequence. and To extract resonant mode variation features containing magnetostrictive effect information from magnetostrictive micro-vibration response signals, time-frequency decomposition was performed on the signals within a time window at each magnetostrictive measurement point, converting the discrete sequence... Mapped to time-frequency energy distribution Based on this, define the characterization of the first The modal energy distribution function in the energy set of the nth target resonance modes can be defined, for example, as the nth Each measurement point within the time window The Middle The normalized energy characteristics of each resonance mode are as follows:

[0075] ,

[0076] in, Indicates the first The magnetostriction measuring point at the first... Within the time window and the first The energy proportion characteristic corresponding to each resonance mode is in the range of zero and one. Indicates the relationship between the frequency axis and the time axis and the first... The set of time-frequency regions corresponding to each resonance mode is obtained through mode identification or pre-calibration; This represents the set of all frequency bands and all time regions that participate in energy statistics within this time window; Indicates the first The magnetostriction measuring point at the first... Within a time window, at frequency sampling points With time sampling points The time-frequency components at that point. Through the analysis of... Different and different By tracking the changes in the core, we can obtain the resonant mode variation characteristics that reflect the stiffness evolution and magnetostriction effect enhancement in the core region, winding region, and support frame region. At the same time, we can further calculate the energy distribution entropy or energy centroid offset in the time-frequency region of each mode as additional resonant mode variation characteristics, thereby forming a more refined characterization of the spatial and frequency band distribution of the magnetostriction effect.

[0077] For guided wave propagation response signals, to extract guided wave path propagation characteristics containing overall structural propagation path information and local scattering characteristics containing information on local structural discontinuities, cross-correlation analysis and waveform alignment processing are performed between different receiving channels for each preset path within each time window. This ensures precise temporal alignment of guided wave responses excited from the same transmitter to different receivers, and determines the equivalent propagation delay of the path based on the cross-correlation peak position. Furthermore, by integrating the energy of the first wave interval, the reflected wave interval, and the scattered wave interval, path-level propagation energy characteristics and scattering energy characteristics are constructed. For example, the overall propagation energy ratio and scattering energy ratio of each path can be used as part of the guided wave path propagation characteristics and local scattering characteristics of that path. For local scattering characteristics, the energy ratio of different frequency bands, wave packet duration, and waveform asymmetry can be parameterized within the scattered wave time window, making the scattering behavior caused by local cracks, lamination debonding, or support loosening significantly separable in the feature space. Through the above preprocessing based on synchronous time window division, each time window corresponds to a set of resonant mode change characteristics containing magnetostrictive effect information, guided wave path propagation characteristics containing overall structural propagation path information, and local scattering characteristics containing local structural discontinuities. Moreover, these characteristics are completely synchronized with the composite working excitation on the time axis.

[0078] After extracting the resonant mode variation characteristics, guided wave path propagation characteristics, and local scattering characteristics within each time window, the features are combined to construct a magnetoacoustic mode feature vector sequence according to a preset channel order, frequency band division method, and path numbering method. First, a fixed index number is assigned to each magnetostriction measurement point, each guided wave receiving channel, and each guided wave propagation path. This ensures that features belonging to the same physical location or the same path are always mapped to a fixed position in the magnetoacoustic mode feature vector across different time windows and different test batches. Let the... The magnetoacoustic modal eigenvectors corresponding to each time window are denoted as follows: All resonant modal variation characteristics are arranged sequentially according to both measurement point number and modal number to form a vector segment. Arrange all waveguide path propagation characteristics in order of path number and propagation direction to form a vector segment. Arrange all local scattering features in order of path number and scattering time window number to form a vector segment. When necessary, features of different categories can be normalized or weighted, for example, by constructing a concatenated vector of the following form:

[0079] ,

[0080] in, Indicates the first The magnetoacoustic modal feature vectors corresponding to each time window serve as the input feature basis for the subsequent magnetoacoustic-performance mapping model; Indicates the first The characteristic subvector of resonant mode change composed of all magnetostrictive measurement points and all target resonant modes within a time window; Indicates the first The waveguide path propagation characteristic sub-vector is composed of parameters such as propagation time, propagation energy, and spectral centroid of all effective waveguide paths within a time window; Indicates the first The local scattering feature vector within each time window is composed of parameters such as scattering energy, scattering packet duration, and scattering frequency band distribution extracted from all scattering-related time windows and path locations. , and These represent weighting coefficients that adjust for the relative importance of different feature components. Their values ​​range from zero to one and are determined through offline training or empirical settings. The system is constructed according to the above order. By time index on all time windows Arranged sequentially, this forms a sequence of magnetoacoustic mode feature vectors. This sequence is perfectly aligned with the composite working excitation in the time dimension and simultaneously contains information on magnetostriction effect, overall structural propagation path behavior, and local scattering behavior in the feature dimension. It provides a feature expression basis with strict sequential correlation for the joint inference of subsequent ratio difference prediction, phase angle error prediction, saturation characteristic offset prediction, harmonic error prediction, and degradation category probability distribution.

[0081] S140. Input the magnetoacoustic mode feature vector sequence into the pre-constructed magnetoacoustic-performance mapping model with electromagnetic performance calibration results to obtain the ratio difference prediction value, phase angle error prediction value, saturation characteristic offset prediction value and harmonic error prediction value corresponding to the current transformer, and obtain the degradation category probability distribution corresponding to the current transformer.

[0082] Specifically, the pre-constructed magnetoacoustic-performance mapping model with electromagnetic performance calibration results refers to a mapping model obtained by jointly training a large number of sample current transformers at different degradation stages with the magnetoacoustic mode feature vector sequences and corresponding electromagnetic performance calibration results. This mapping model establishes a correspondence between "externally observable magnetoacoustic features" and "internal electromagnetic metrological performance." During the model construction phase, a batch of current transformer samples with known structural states or whose degradation mechanisms can be confirmed through disassembly are selected. Magnetoacoustic mode feature vector sequences are obtained under different excitation conditions. Simultaneously, standard ratio difference tests, phase angle error tests, saturation characteristic tests, and harmonic error tests are performed on these samples under the same conditions to obtain electromagnetic performance calibration results as labels. By using the magnetoacoustic mode feature vectors, electromagnetic performance calibration results, and degradation category calibration results together for training, this magnetoacoustic-performance mapping model internally forms a set of parameters capable of estimating electromagnetic performance and degradation category based on input features.

[0083] The ratio difference prediction value refers to the estimate of the ratio difference of a current transformer within a given time window, given by the magnetoacoustic-performance mapping model based on the feature-performance relationship learned during previous training, when the model receives the magnetoacoustic mode feature vector corresponding to that time window. The ratio difference is one of the key metrological indicators of a current transformer, used to measure whether the transformer's transformation ratio deviates from its rated value; simply put, it is the deviation between the "actual transformation ratio" and the "nominal transformation ratio." In the field, the ratio difference is calculated by applying a standard current and simultaneously measuring the current on both the primary and secondary sides. However, in this technical solution, by inputting the magnetoacoustic mode feature vector sequence into the magnetoacoustic-performance mapping model, the model can output the ratio difference prediction value without the need for an additional high-current test circuit.

[0084] The predicted phase angle error refers to the estimate given by the magnetoacoustic-performance mapping model for the phase difference between the primary and secondary currents of a current transformer under the same operating condition, based on the magnetoacoustic mode eigenvectors corresponding to the same time window. Phase angle error is closely related to the correctness of relay protection operation, especially in protection configurations requiring precise current phase determination; excessive phase angle error can lead to misjudgment or failure to operate. In traditional testing, the phase difference between the primary and secondary sides needs to be measured using a phase meter or dedicated testing instruments. However, in this technical solution, because the magnetostrictive micro-vibration response signal and guided wave propagation response signal are highly sensitive to changes in the core magnetization state and leakage flux distribution, the model can infer the trend of phase angle error changes using this information and output the predicted phase angle error value.

[0085] The predicted saturation characteristic offset refers to the prediction result of the magnetoacoustic-performance mapping model based on the magnetoacoustic mode feature vector, indicating the degree of deviation of the critical inflection point position of the saturation characteristic of a current transformer relative to the design value. The saturation characteristic describes the process by which the transformer core gradually enters the saturation region from the linear operating region as the primary current gradually increases. The saturation inflection point determines whether the transformer can maintain an acceptable error level under high-current fault conditions. If the saturation characteristic offset occurs, the transformer may prematurely enter severe saturation before reaching the expected fault current level, affecting the reliability of protection criteria. By observing the combination of resonant mode change characteristics and local scattering characteristics, the magnetoacoustic-performance mapping model can identify local residual magnetism in the core, stress redistribution, and changes in the lamination contact state, thus providing the predicted saturation characteristic offset value.

[0086] Harmonic error prediction refers to the prediction result of the magnetoacoustic-performance mapping model on the degree of deviation of the transmission characteristics of each harmonic component of a current transformer when harmonic currents are present. This result reflects whether the transformer's amplification or attenuation of multi-frequency current components deviates from the normal range. Harmonic errors are particularly important in power distribution systems with a large number of power electronic loads, because harmonic currents not only affect billing accuracy but also affect equipment heating and protection actions. Through the magnetoacoustic mode feature vector sequence, the model can identify features associated with high-frequency vibration modes and high-frequency guided wave path propagation, and then infer the error changes of the transformer at different harmonic orders, outputting the harmonic error prediction value.

[0087] The degradation category probability distribution refers to the set of probability values ​​output by the degradation category discrimination branch within the magnetoacoustic-performance mapping model for each time window's corresponding magnetoacoustic mode feature vector. This distribution characterizes the current transformer's likelihood of being dominated by a specific degradation mechanism. Degradation categories can be classified based on engineering experience and failure mode analysis into core demagnetization-dominated degradation, winding relaxation-dominated degradation, dielectric polarization migration-dominated degradation, and composite degradation, among others. Each component in the degradation category probability distribution represents the probability that the transformer belongs to a particular degradation category within the current time window; all components are summed to one.

[0088] Furthermore, in specific implementation, when performing high-dimensional nonlinear feature transformation on the magnetoacoustic modal feature vector sequence through the time modeling unit within the magnetoacoustic-performance mapping model, the magnetoacoustic modal feature vector sequence corresponding to the continuous time window is first represented as... ,in Indicates the first The magnetoacoustic mode feature vectors corresponding to each time window. This represents the number of time windows involved in the temporal modeling. The temporal modeling unit encodes the sequence using a combination of a bidirectional gated recursive structure and an attention mechanism. The forward hidden state and the reverse hidden state are represented as follows: and Furthermore, local temporal dependencies can be constructed using gating functions and nonlinear mappings. For example, the following can be defined:

[0089] ,

[0090] ,

[0091] in, and They represent the first The hidden state vectors of each time window in the forward and backward recursion; and Indicates the first The gating vector corresponding to each time window has components ranging from zero to one, and is used to weight the historical hidden states and the current nonlinear transformation. This indicates element-wise multiplication; This represents an element-wise nonlinear transformation function, which can be either a hyperbolic tangent or a smooth activation function. , , and , , Let represent the weight matrix and bias vector for the forward and backward recursion, respectively. Their dimensions are determined by the latent space dimension and the input feature dimension. The forward and backward latent states are concatenated to obtain... Then, a weighted aggregation is performed across different time windows using a time-series self-attention structure. The coupling relationship between short-term disturbances and long-term trends is characterized by both bilinear and nonlinear correlation, for example, defined as follows:

[0092] ,

[0093] ,

[0094] ,

[0095] in, Indicates the first The first time window and the first A scalar measure of the correlation between hidden states within a time window; This represents the bilinear correlation matrix, used to characterize the coupling strength in different directions within the latent space; and These represent the weight vector and bias vector of the non-linear branch in the attention scoring, respectively. This represents the weight matrix that maps the concatenated hidden states to the attention hidden space; Indicating the construction of the first When a time window shares a hidden space representation, it assigns the first... The weighting coefficients of the hidden state within a time window; This represents the weight matrix used to perform a linear transformation on the weighted hidden states; Indicates the first The reconstructed gating vector for each time window, through get, This represents a nonlinear function that compresses element by element between zero and one. and Let represent the weight matrix and bias vector of the gating transformation, respectively. Through the above structure, the following is obtained: Simultaneously containing the following in each time window The local modal perturbation information represented by and derived from The aggregation represents the long-term structural evolution trend information, thereby forming a shared latent space representation for subsequent multi-task regression and classification.

[0096] Representation in shared implicit space Based on this, to output the predicted values ​​of ratio difference, phase angle error, saturation characteristic offset, and harmonic error, a multi-task regression structure is constructed within the model. Mapping to a more suitable common representation for electromagnetic performance prediction through one or more nonlinear transformations. For example, the definition:

[0097] ,

[0098] in, Indicates the first The common representation vector of tasks for each time window; and These represent the weight matrices for the two levels of nonlinear transformation; and This represents the corresponding bias vector; This refers to a nonlinear activation function that can suppress saturation and retain high-frequency information, such as a parameterized rectified function; The weight matrix represents the weights of the direct connection path, used to preserve the weights from the direct connection path beyond deep nonlinear transformations. The linear component. For each specific electromagnetic performance index, a regression mapping with a quadratic term and a neighborhood regularization term is constructed, such as the comparison difference prediction value. Predicted phase angle error saturation characteristic offset prediction value Harmonic error prediction values It can be uniformly described as:

[0099] ,

[0100] in, Indicates the first The time window in the first Predicted values ​​for each electromagnetic performance index It can be any one of the four types: ratio difference prediction, phase angle error prediction, saturation characteristic offset prediction, and harmonic error prediction. Indicates the first The linear regression weight vector corresponding to each electromagnetic performance index is used to characterize the common representation of the task. The first-order relationship with this performance metric; Indicates the first The symmetric quadratic weight matrix corresponding to each electromagnetic performance index is used to characterize the influence of second-order coupling between different components of the task's common representation on the performance index. Representation of offsets in shared implicit space The relevant weight vector, This represents the empirical mean of the shared latent space representation in the training data, used to compensate for the overall drift between different operating conditions. Indicates the first The time smoothing coefficients corresponding to each performance index are taken as non-negative values ​​in practical applications to control the smoothness of the predicted sequence in the time dimension. and These represent the predicted values ​​for the same performance index within adjacent time windows, facilitating the introduction of local second-order time differences to suppress drastic, non-physical fluctuations. During model training, data with calibration results are used to jointly constrain various... Approximating the actual electromagnetic performance calibration results, this structure simultaneously provides the predicted values ​​of ratio difference, phase angle error, saturation characteristic offset, and harmonic error based on the shared implicit space representation.

[0101] When determining the real-valued scores of multiple degradation categories corresponding to a current transformer, a joint feature vector is first constructed. This vector is represented by the shared latent space. Public representation of tasks And it is composed of four electromagnetic performance prediction values ​​spliced ​​together in a fixed order, for example For each predefined degradation category Construct a scoring function that combines quadratic terms, multi-core nonlinear terms, and bias terms, and then... Mapped to the real-valued score of this degradation category For example, it can be defined as:

[0102] ,

[0103] in, Indicates the first The first time window corresponds to the first Each degradation category has a real-valued score, and the value can be any real number. The larger the value, the higher the degree of matching between the current feature state and the degradation category. Indicates the first The quadratic weight matrix corresponding to each degradation category is used to characterize the influence of the interaction between different components of the joint feature vector on the degradation category; Indicates the first The degradation category in the first The weight coefficients on the kernel function can be positive or negative, and are used to control the contribution of different kernel components to the score. Indicates The center of the first A kernel function at point The value at the given point can be chosen, for example, a Gaussian kernel, a Laplace kernel, or a polynomial kernel can be selected to enhance the nonlinear discrimination capability. The kernel function center vector is derived from several representative feature patterns in the training data. This indicates the number of kernel functions involved in the scoring; Indicates the first The bias term of each degradation category score is used to adjust the baseline score level of that category when there is no feature preference. Through this structure, the real-valued scores of degradation categories comprehensively consider structural mode changes, waveguide characteristics, electromagnetic performance predictions, and their multi-scale interactions, making the degradation categories dominated by core demagnetization, winding relaxation, dielectric polarization migration, and composite degradation categories distributable in the score space.

[0104] After obtaining the real-value scores for all degradation categories, the real-value scores for each degradation category are normalized through classification branching to form the probability distribution of degradation categories within the same time window. In the... Within a time window, for the set of degradation categories We calculate the unnormalized weight value for each degradation category, and introduce a category-dependent temperature factor and prior weights to obtain a more flexible probability allocation structure. For example, we can define:

[0105] ,

[0106] in, Indicates the first The current transformer belongs to the first time window. The probability value for each degradation category is between zero and one. Indicates the first The weight coefficients of each degradation category under prior statistical significance can reflect the frequency of occurrence of the degradation pattern in the overall sample based on historical engineering data; The exponential coefficient representing the degree of influence of prior weights, when When taking zero, prior knowledge is not considered. When the value is positive, the prior weights have a stronger influence on the probability distribution; Indicates the first A temperature scaling factor for each degradation category is used to adjust the sensitivity of the real-valued score for that category in the exponential map. When the value is larger, the corresponding category probability is more sensitive to changes in the score; the denominator is the sum of the exponential mappings of all degradation categories after prior correction, used to ensure that the sum of all degradation category probability values ​​is one within this time window. The degradation category probability distribution obtained in this way gives the relative probability of occurrence of each degradation mechanism in each time window, and forms a continuous probability evolution trajectory on the time axis. Together with the aforementioned electromagnetic performance prediction values, it is used to construct the performance degradation curve and the degradation category evolution trajectory, realizing the joint assessment of performance changes and dominant degradation mechanism changes during the life stage of the current transformer.

[0107] S150. Determine the changing trends of the ratio difference prediction value, phase angle error prediction value, saturation characteristic offset prediction value, and harmonic error prediction value at different times. Combined with the degradation category probability distribution, form the performance degradation curve and degradation category evolution trajectory of the current transformer, so as to complete the performance evaluation of the current transformer based on the performance degradation curve and degradation category evolution trajectory.

[0108] Specifically, a performance degradation curve is a graphical representation of the comprehensive degradation process of the electromagnetic performance of a current transformer over time, plotted or arranged in chronological order by the predicted ratio difference, phase angle error, saturation characteristic offset, and harmonic error. The performance degradation curve can be a time series curve of a single indicator or a composite curve obtained by weighting multiple indicators to form a comprehensive performance score. Through the performance degradation curve, the continuous evolution of the transformer from a stable performance stage, a slight degradation stage, a significant degradation stage, and finally to a near-failure stage can be observed. For example, in a certain engineering application, the predicted ratio difference and phase angle error are weighted according to their importance to form a comprehensive performance index. As the transformer operates for a long time, this index remains at a low level for several years, then slowly rises over several months, and finally rapidly approaches a preset threshold in a short period. This curve clearly shows the staged characteristics of the transformer's performance degradation.

[0109] The degradation category evolution trajectory refers to arranging the probability distribution of degradation categories corresponding to each time window in chronological order, and displaying the trajectory of how the dominance of different degradation categories changes over time. The degradation category evolution trajectory can be understood as plotting the change curves of the probability of each degradation category on a time axis, or as using color or bandwidth to represent the proportion of a certain degradation category at different times on a two-dimensional plane. For example, in the early stages of a current transformer's service life, the degradation category evolution trajectory may show that the probabilities of all categories are close to a healthy state or a slightly degraded category; as service time progresses, the probability of the core demagnetization-dominated degradation category gradually increases within multiple time windows; after further operation, the probability of the winding slack-dominated degradation category begins to increase significantly and shares the dominant position with the core demagnetization-dominated degradation category; near the failure stage, the probability of the composite degradation category increases sharply, indicating the existence of multiple degradation mechanisms superimposed.

[0110] Furthermore, when arranging the predicted values ​​of ratio difference, phase angle error, saturation characteristic offset, and harmonic error in chronological order to form a performance prediction result sequence covering the continuous excitation state of the current transformer, a strictly monotonically increasing time index is first assigned to each time window on the output side of the magnetoacoustic-performance mapping model. The predicted values ​​for the ratio difference, phase angle error, saturation characteristic offset, and harmonic error corresponding to each time window are denoted as follows: , , and and construct the first A vector of performance prediction results for each time window:

[0111] ,

[0112] Arrange the performance prediction result vectors across all time windows in ascending order of time index to form a performance prediction result sequence. ,in This refers to the number of time windows involved in the evaluation. To ensure the comparability of performance prediction results with different dimensions and numerical scales, the statistical mean and statistical fluctuation range can be pre-calculated for each type of performance prediction result based on historical calibration data or long-term operating data. A standardization transformation can then be performed on each dimension of the performance prediction result; for example, a standardized vector can be defined for each type of performance prediction result. ,

[0113] in , , , They were respectively in the second The dimensionless normalized results obtained by performing translation and scaling transformation on the predicted values ​​of phase difference, phase angle error, saturation characteristic offset, and harmonic error in each time window have values ​​concentrated in a finite interval, which is convenient for subsequent trend analysis and construction of comprehensive performance indicators. Through this construction method, the performance prediction result sequence not only retains the original prediction results of each time window, but also aligns the relative offset of different indicators in a unified feature space, thereby fully covering the time evolution of the performance prediction results of the current transformer under continuous excitation.

[0114] In identifying the changing trends between different time windows in the performance prediction result sequence, and simultaneously constructing a degradation category probability distribution sequence that corresponds one-to-one with the time windows of the performance prediction result sequence, a smoothing trend and an incremental trend are built for each performance prediction indicator to characterize the slow drift and rapid change of that indicator on the time axis. An exponential smoothing trend sequence can be constructed for each type of performance prediction result, making... Indicates the first Class performance prediction results in the first Smoothing values ​​under a time window ,definition:

[0115] ,

[0116] in Indicates the first Class performance prediction results in the first The exponential smoothing trend value under each time window reflects the smooth change of the performance index in adjacent time windows; Indicates the first The smoothing coefficient corresponding to the performance prediction results ranges from zero to one. The closer it is to one, the better the smoothing trend predicts the current time window value. The more sensitive, The smaller the value, the more the smoothing of the trend depends on the long historical sequence; This represents the smoothed trend value corresponding to the previous time window. An incremental trend is defined based on the smoothed trend to characterize the rate of change of the performance prediction result between adjacent time windows, letting... Indicates the first Class performance prediction results in the first The incremental trend value of each time window is defined as follows:

[0117] ,

[0118] in The sign of the value reflects the direction of the performance prediction over time, indicating whether it is rising or falling, while the magnitude of the value reflects the rate of change. Maintaining a small value over multiple consecutive time windows can be considered a stable performance indicator. A sustained increase over multiple consecutive time windows can be considered an accelerated degradation of the performance metric. The probability distribution of degradation categories corresponding to the performance prediction result sequence is represented as a set of probability values ​​for each time window, let... ,in Indicates the first The current transformer belongs to the first time window. The probability of each degradation category, This represents the number of degenerate categories, with each value ranging from zero to one, and all components summing to one. For After arranging the degradation category probability distribution sequence by time index, the dominant change in degradation mechanism is characterized by defining the intensity of distribution change between adjacent time windows. For example, a weighted total variation metric can be used to define the intensity of overall degradation category distribution change. ,as follows:

[0119] ,

[0120] in Indicates the first The intensity of the overall change in the probability distribution of degradation categories in each time window relative to the previous time window; the larger the value, the more significant the shift in the dominant category of degradation mechanism. Indicates the first The weight coefficients for each degradation category, taking values ​​in the range of non-negative real numbers, are used to emphasize the sensitivity to changes in important degradation categories, such as composite degradation categories or near-failure related degradation categories, and the absolute value of the weights is not subject to normalization constraints. and The first two consecutive time windows are respectively the first two consecutive time windows. The probability value of each degradation category. The trend of the performance prediction result sequence as described above. , With the probability distribution sequence of degradation categories and the intensity of its change The combined description allows us to obtain the changing trends of various performance indicators of the current transformer and the dominant changes in degradation categories within each time window.

[0121] By correlating the changing trends of the performance prediction result sequence with the dominant changes in the degradation category probability distribution sequence, a performance degradation curve and a degradation category evolution trajectory are generated. This allows for the determination of the current transformer's stage based on the performance degradation curve and degradation category evolution trajectory. Firstly, a comprehensive performance degradation index is constructed in the trend space, mapping the smoothed trend and incremental trend of the multi-dimensional performance prediction results into a single performance degradation intensity sequence. This can be applied to the [specific stage / stage]. The time window constructs the overall performance degradation intensity For example, a weighted quadratic form can be constructed based on the normalized trend of each performance prediction result:

[0122] ,

[0123] in Indicates the first The comprehensive performance degradation intensity corresponding to each time window is used to reflect the degree and rate of change of ratio difference, phase angle error, saturation characteristic shift and harmonic error on a single scalar. Indicates the first The weighting coefficient of the smoothing trend term in the performance prediction results is a non-negative real number, used to adjust the contribution of each performance index to the overall performance degradation intensity. and They represent the first The statistical mean and statistical fluctuation scale of the performance prediction results in historical healthy samples or benchmark samples are used to standardize the smoothing trend so that the degree of deviation of different indicators can be compared on the same scale. Indicates the first The weighting coefficient of the incremental trend term in the performance prediction results is used to adjust the importance of the rate of change in the overall performance degradation intensity. Indicates the first Typical fluctuation scales of incremental trends in performance prediction results under healthy or slowly degrading states are used to analyze... Standardization is performed to allow comparisons of the rates of change of different performance indicators on the same scale. (Comprehensive performance degradation intensity sequence) As the numerical basis for the performance degradation curve, it is plotted on the time axis. This allows us to obtain the performance degradation trajectory of the current transformer from the initial stage to the current stage. Simultaneously, to characterize the evolution of degradation category dominance over time, the degradation category probability distribution sequence is obtained. Unfolding along the timeline reveals the evolutionary trajectory of degenerate categories, which can be analyzed by selecting the dominant degenerate category index. To represent it, as follows:

[0124] ,

[0125] in Indicates the first The index of the degradation category with the highest probability value within a given time window is used to identify the dominant degradation mechanism within that time window. Furthermore, the probability values ​​of each degradation category can be... With overall performance attenuation intensity This approach, used in conjunction with other methods, divides the operating state of current transformers into four stages: normal operation, early latent degradation, accelerated degradation, and near-failure, by setting multi-level performance strength thresholds and degradation category probability thresholds. For example, in engineering implementation, a set of judgment thresholds for each of the four stages can be selected. ,satisfy And preset probability thresholds for degradation categories associated with near-failure. ,when long-term below When the probabilities of all unhealthy degradation categories are at a low level, this time window is considered the normal operation phase; when Between and The probability of a single degradation category is slowly increasing but has not yet exceeded [a certain threshold]. When this time window is defined as the early stage of latent degradation; when Exceed When the probability of one or more degradation categories increases significantly over multiple consecutive time windows, this time window is identified as the accelerated degradation phase; when Approaching or exceeding Furthermore, the probability of composite degradation categories associated with near-failure exceeds [a certain threshold]. When the time window is reached, it is determined to be near failure. By correlating the changing trend of the performance prediction result sequence with the dominant changes in the degradation category probability distribution sequence, we can read "how the degradation intensity evolves over time" from the performance decay curve and "which degradation mechanism begins to dominate at what time" from the degradation category evolution trajectory. This provides a comprehensive judgment result with quantitative basis and mechanistic explanation for the performance evaluation and maintenance decisions of current transformers at different life stages.

[0126] In one possible implementation, during the model calibration phase, a training dataset is constructed for current transformer samples with different degradation mechanisms and different life stages. Each target magnetoacoustic mode feature vector in the training dataset corresponds one-to-one with the target ratio difference calibration value, target phase angle error calibration value, target saturation characteristic offset calibration value, and target harmonic error calibration value, and has a degradation category label. During the model training phase, a magnetoacoustic-performance mapping model with regression and classification branches is trained using the training dataset. The regression branch outputs the target ratio difference calibration value, target phase angle error calibration value, target saturation characteristic offset calibration value, and target harmonic error calibration value corresponding to the target magnetoacoustic mode feature vector, and the classification branch outputs the target degradation category probability distribution corresponding to the target magnetoacoustic mode feature vector.

[0127] Specifically, during the model calibration phase, when constructing a training dataset for current transformer samples with different degradation mechanisms and lifespan stages, a set of current transformer samples covering multiple degradation mechanisms and lifespan stages is first selected. These samples include those in normal operation at the initial factory stage, those undergoing high-temperature and high-humidity aging, overload impact, long-term harmonic stress, or mechanical vibration fatigue treatment at different degradation levels, and in-service recycled samples that have been in operation for many years and whose degradation mechanisms have been confirmed through disassembly and inspection. For each current transformer sample, a programmable current source, ultrasonic transducer array, piezoelectric sheet, laser vibration measurement device, and fiber optic demodulation device are configured on the experimental rig. The current transformer sample is continuously excited under multiple preset composite operating excitation conditions. Corresponding magnetostrictive micro-vibration response signals and guided wave propagation response signals are collected. A magnetoacoustic modal feature vector sequence is constructed based on a time window division method synchronized with the composite operating excitation, and multiple target magnetoacoustic modal feature vectors are extracted or slidably selected from this sequence. For each target magnetoacoustic mode characteristic vector, under the same transformer samples, excitation conditions, and time window, a standard electromagnetic performance test circuit was constructed. The target ratio difference calibration value was obtained using a ratio difference test method conforming to metrological specifications. The target phase angle error calibration value was obtained using the same test system. The target saturation characteristic offset calibration value was obtained by gradually increasing the primary current to near the saturation range and recording the saturation inflection point. The target harmonic error calibration value was obtained by measuring the secondary response under multi-order harmonic current and calculating the harmonic errors of each order. This was combined with the mutual inductance... The manufacturing information, accelerated aging process, disassembly and inspection results and diagnostic records of the instrument sample are used to determine the dominant degradation mechanism of the sample under this time window. Categories such as core demagnetization, winding relaxation, dielectric polarization migration, composite degradation or near-health are used as degradation category labels. If necessary, the degree of degradation is further subdivided into different levels so that each target magnetoacoustic mode feature vector in the training dataset corresponds one-to-one with the target ratio difference calibration value, target phase angle error calibration value, target saturation characteristic offset calibration value, target harmonic error calibration value and degradation category label.

[0128] During the model training phase, when training the magnetoacoustic-performance mapping model with regression and classification branches using the completed training dataset, all target magnetoacoustic modal feature vectors in the training dataset are uniformly divided into training set, validation set and test set to ensure that the distribution of different current transformer samples and different degradation mechanisms in each subset after division is representative. The input to the magnetoacoustic-performance mapping model is the target magnetoacoustic modal feature vector or a temporally organized sequence of magnetoacoustic modal feature vectors. The intermediate structure of the model extracts a shared latent space representation through a temporal modeling unit and a high-dimensional nonlinear feature transformation unit. The regression branch, based on the shared latent space representation, includes sub-branches for ratio difference regression, phase angle error regression, saturation characteristic shift regression, and harmonic error regression. Each sub-branch receives the shared latent space representation and outputs the corresponding predicted values ​​for ratio difference, phase angle error, saturation characteristic shift, and harmonic error. These values ​​are compared with the corresponding target ratio difference calibration values, target phase angle error calibration values, target saturation characteristic shift calibration values, and target harmonic error calibration values ​​in the training dataset to calculate the multi-task regression loss, which constrains the model's fitting accuracy on different performance indices. The classification branch outputs a real-valued score for the degradation category based on the shared latent space representation and, if necessary, the joint features of the performance prediction results. The degradation category probability distribution is obtained through normalization and compared with the corresponding degradation category labels in the training dataset to form the classification loss, constraining the model's ability to distinguish between different degradation mechanisms. The entire magnetoacoustic-performance mapping model is trained using a joint optimization approach. This involves weighting and summing the multi-task regression loss and classification loss according to preset weights, and then iteratively updating the model parameters using a gradient descent optimization algorithm. This allows the regression branch to gradually approach the target ratio error calibration value, the target phase angle error calibration value, the target saturation characteristic offset calibration value, and the target harmonic error calibration value, while the classification branch gradually approaches the target degradation category probability distribution. During training, the model's error changes and overfitting signs are monitored using a validation set, and the loss weights and regularization strength are dynamically adjusted. Finally, the model's generalization ability on untrained samples is validated on a test set. This results in a magnetoacoustic-performance mapping model capable of simultaneously outputting multiple electromagnetic performance prediction results and degradation category probability distributions on unknown current transformers, providing a reliable model foundation for subsequent online performance evaluation.

[0129] In one possible implementation, the sampling period and sampling density of the magnetoacoustic excitation sequence are dynamically adjusted based on the rate of change and acceleration of change corresponding to the performance decay curve, as well as the degree of abrupt change in the probability of key degradation categories in the degradation category evolution trajectory. According to the adjusted sampling period and sampling density of the magnetoacoustic excitation sequence, when the current transformer enters the implicit degradation acceleration stage or is close to failure stage, the acquisition frequency of the magnetostrictive micro-vibration response signal and the guided wave propagation response signal is increased, and a failure risk warning is output.

[0130] Specifically, in the implementation process, a continuous-time update mechanism is first established for the performance degradation curve and degradation category evolution trajectory of the current transformer. This ensures that the comprehensive performance degradation intensity, rate of change indicator, acceleration indicator, and degradation category probability corresponding to each time window are updated and cached in real time. The server uses the changing trend of the performance degradation curve as the core monitoring indicator, and performs statistical analysis on its trend in the most recent multiple time windows to determine whether the comprehensive performance degradation intensity has transitioned from a slow change to a rapid increase, whether the rate of change has maintained a continuous high level, and whether the acceleration has remained positive and exceeded the empirical threshold. At the same time, the server continuously tracks the probability changes of key degradation categories representing core demagnetization-dominated degradation, winding relaxation-dominated degradation, dielectric polarization migration-dominated degradation, and composite degradation in the degradation category evolution trajectory, identifying whether the probability of key degradation categories has significantly increased in a short period of time or shown a sudden jump from low probability to medium-high probability.

[0131] Based on the above multi-parameter joint judgment results, the server automatically assigns a dynamic monitoring level to the current transformer's current state. Different monitoring levels correspond to different magnetoacoustic excitation sequence sampling periods and sampling densities. When the performance degradation curve changes smoothly and the degradation category probability is uniformly distributed, the server sets the magnetoacoustic excitation sequence sampling period to a longer period, implementing fewer sets of composite working excitations within each period to reduce online monitoring resource consumption and interference burden. When the performance degradation curve shows signs of acceleration or the probability of key degradation categories rises slowly, the server automatically shortens the magnetoacoustic excitation sequence sampling period and increases the number of excitation sets per period, increasing the magnetoacoustic modal data acquisition density, enabling sensitive tracking capability at the very beginning of the degradation acceleration trend. When a steep rise in the performance degradation curve or a significant abrupt change in the probability of key degradation categories is detected within a short period, indicating that performance degradation has entered a critical stage, the server further compresses the sampling period, enabling the magnetoacoustic excitation sequence to cover multiple frequency bands, multiple structural propagation paths, and multiple operating conditions with high density, forming a high-frequency, high-resolution dynamic performance monitoring mode.

[0132] During acquisition and monitoring based on the adjusted magnetoacoustic excitation sequence, the server synchronously adjusts the triggering strategies and sampling parameters of the laser vibration measurement points, fiber optic grating measurement points, ultrasonic transducer array receivers, and piezoelectric sheets to match the acquisition frequency of the magnetostrictive micro-vibration response signal and the guided wave propagation response signal with the monitoring level. In the latent degradation acceleration phase, the server executes multiple sets of composite working excitations within each shortened sampling period, causing the transformer to be repeatedly excited under multiple harmonic content, magnetic bias, and load combinations, and acquiring magnetoacoustic modal data with higher sampling resolution, thereby capturing the frequent subtle structural disturbances that occur during this phase. When entering the near-failure phase, the server further improves the sampling rate, dynamic range, and time synchronization accuracy of the acquisition channels, focusing on capturing high-frequency nonlinear anomalies such as local structural instability, demagnetization propagation, and sudden evolution of latent defects, so that the performance degradation curve and degradation category evolution trajectory can accurately reflect the degree of approaching failure.

[0133] After each update, the server also performs reasoning analysis on the performance degradation curve and the evolution trajectory of degradation categories. When the overall performance degradation intensity continues to exceed the risk threshold or the probability of degradation categories related to the critical degradation mechanism remains high, it automatically outputs a failure risk warning, providing the current monitoring level, the direction of performance trend change, the dominant degradation mechanism and its probability, and generating operation and maintenance suggestions, including reducing the operating load, arranging reinforcement measures or carrying out planned replacements, thereby realizing proactive defensive monitoring and forward-looking fault warning driven by performance degradation status.

[0134] This application also provides a performance evaluation device for a current transformer, referring to... Figure 2 , Figure 2This is a schematic diagram of a performance evaluation device for a current transformer provided in an embodiment of this application. The device is a server, which includes an acquisition module 21 and a processing module 22. The processing module 22 is used to apply a composite working excitation to the primary side of the current transformer through a programmable current source, so that the current transformer is in an excitation state with saturation edge and containing a non-periodic component under excitation conditions. The acquisition module 21 is used to acquire the magnetostrictive micro-vibration response signal and the guided wave propagation response signal excited by the composite working excitation under the excitation state of the current transformer. The processing module 22 is also used to perform signal processing on the magnetostrictive micro-vibration response signal and the guided wave propagation response signal to extract resonant mode change characteristics, guided wave path propagation characteristics, and local scattering characteristics, and to perform signal processing on the resonant mode change characteristics, The waveguide path propagation characteristics and local scattering characteristics are used to construct a magnetoacoustic mode feature vector sequence. The processing module 22 is also used to input the magnetoacoustic mode feature vector sequence into a pre-constructed magnetoacoustic-performance mapping model with electromagnetic performance calibration results to obtain the ratio difference prediction value, phase angle error prediction value, saturation characteristic offset prediction value, and harmonic error prediction value corresponding to the current transformer, and to obtain the degradation category probability distribution corresponding to the current transformer. The processing module 22 is also used to determine the changing trends of the ratio difference prediction value, phase angle error prediction value, saturation characteristic offset prediction value, and harmonic error prediction value at different times, and combine them with the degradation category probability distribution to form the performance decay curve and degradation category evolution trajectory of the current transformer, so as to complete the performance evaluation of the current transformer based on the performance decay curve and degradation category evolution trajectory.

[0135] It should be noted that the above embodiments of the apparatus are only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0136] This application also provides an electronic device, with reference to... Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include: at least one processor 31, at least one network interface 34, a user interface 33, a memory 35, and at least one communication bus 32.

[0137] The communication bus 32 is used to enable communication between these components.

[0138] The user interface 33 may include a display screen and a camera. Optionally, the user interface 33 may also include a standard wired interface and a wireless interface.

[0139] The network interface 34 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0140] The processor 31 may include one or more processing cores. The processor 31 connects to various parts of the server via various interfaces and lines, executing instructions, programs, code sets, or instruction sets stored in the memory 35, and calling data stored in the memory 35 to perform various server functions and process data. Optionally, the processor 31 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 31 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content to be displayed on the screen; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 31 and may be implemented as a separate chip.

[0141] The memory 35 may include random access memory (RAM) or read-only memory. Optionally, the memory 35 may include a non-transitory computer-readable storage medium. The memory 35 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 35 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 35 may also be at least one storage device located remotely from the aforementioned processor 31. Figure 3As shown, the memory 35, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application program for evaluating the performance of a current transformer.

[0142] exist Figure 3 In the electronic device shown, the user interface 33 is mainly used to provide an input interface for the user and to obtain the user input data; while the processor 31 can be used to call an application program stored in the memory 35 for evaluating the performance of a current transformer. When executed by one or more processors, the electronic device performs one or more methods as described in the above embodiments.

[0143] This application also provides a non-transitory computer-readable storage medium storing instructions. When executed by one or more processors, these instructions cause an electronic device to perform one or more of the methods described in the above embodiments.

[0144] The foregoing description is merely an exemplary embodiment of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Those skilled in the art will readily conceive of other embodiments of this disclosure upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure. The specification and embodiments are considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

Claims

1. A method for evaluating the performance of a current transformer, characterized in that, The method includes: A composite operating excitation is applied to the primary side of the current transformer by a programmable current source, so that the current transformer is in an excitation state with saturation edge and non-periodic component under excitation conditions. Under the excitation state of the current transformer, the magnetostrictive micro-vibration response signal and the guided wave propagation response signal excited by the composite working excitation are obtained; Signal processing is performed on the magnetostrictive micro-vibration response signal and the guided wave propagation response signal, and features are extracted. Based on the extracted features, a magnetoacoustic mode feature vector sequence is constructed. The magnetoacoustic mode feature vector sequence is input into a pre-constructed magnetoacoustic-performance mapping model with electromagnetic performance calibration results to obtain the ratio difference prediction value, phase angle error prediction value, saturation characteristic offset prediction value and harmonic error prediction value corresponding to the current transformer, and to obtain the degradation category probability distribution corresponding to the current transformer. The changing trends of the ratio difference prediction value, the phase angle error prediction value, the saturation characteristic offset prediction value, and the harmonic error prediction value at different times are determined. Combined with the degradation category probability distribution, the performance degradation curve and degradation category evolution trajectory of the current transformer are formed, so as to complete the performance evaluation of the current transformer based on the performance degradation curve and the degradation category evolution trajectory.

2. The performance evaluation method for a current transformer according to claim 1, characterized in that, The process of acquiring the magnetostrictive micro-vibration response signal and guided wave propagation response signal excited by the composite working excitation under the excitation state of the current transformer specifically includes: High-frequency narrowband acoustic excitation or sweep frequency acoustic excitation is injected into the current transformer through the transmitting end of the ultrasonic transducer array, so that the guided wave propagates along the iron core region, winding region and support frame region of the current transformer. The guided wave propagation response signal is obtained through the receiving end of the ultrasonic transducer array and the piezoelectric sheet; The magnetostrictive micro-vibration response signal generated by the current transformer after applying a working current containing the fundamental frequency component, higher harmonic components, and slow DC bias component is obtained by using laser vibration measurement points or fiber optic grating measurement points.

3. The performance evaluation method for a current transformer according to claim 1, characterized in that, The step involves performing signal processing on the magnetostrictive micro-vibration response signal and the guided wave propagation response signal, extracting features, and constructing a magnetoacoustic mode feature vector sequence based on the extracted features. Specifically, this includes: Based on the time window division method synchronized with the composite working excitation, the magnetostrictive micro-vibration response signal and the guided wave propagation response signal are preprocessed to extract resonant mode change features containing magnetostrictive effect information, guided wave path propagation features containing overall structural propagation path information, and local scattering features containing local structural discontinuities. According to the preset channel order, frequency band division method and path numbering method, the resonant mode change characteristics, the guided wave path propagation characteristics and the local scattering characteristics are combined to construct the magnetoacoustic mode feature vector sequence.

4. The performance evaluation method for a current transformer according to claim 1, characterized in that, The step involves inputting the magnetoacoustic mode feature vector sequence into a pre-constructed magnetoacoustic-performance mapping model with electromagnetic performance calibration results to obtain the predicted ratio difference, phase angle error, saturation characteristic offset, and harmonic error values ​​corresponding to the current transformer, and to obtain the degradation category probability distribution corresponding to the current transformer. Specifically, this includes: Within the magnetoacoustic-performance mapping model, a high-dimensional nonlinear feature transformation is performed on the magnetoacoustic modal feature vector sequence based on the time modeling unit to form a shared latent space representation that can characterize short-term local modal perturbations and long-term structural evolution trends. Based on the shared latent space representation, the shared latent space representation is mapped to the corresponding electromagnetic performance prediction values ​​through the ratio difference regression branch for outputting the ratio difference prediction value, the phase angle error regression branch for outputting the phase angle error prediction value, the saturation characteristic offset regression branch for outputting the saturation characteristic offset prediction value, and the harmonic error regression branch for outputting the harmonic error prediction value, respectively. Determine the real-value scores for multiple degradation categories corresponding to the current transformer; The real-value scores of each degradation category are normalized by classifying the branches to form the probability distribution of each degradation category under the same time window.

5. The performance evaluation method for a current transformer according to claim 1, characterized in that, The process of determining the changing trends of the predicted ratio difference, the predicted phase angle error, the predicted saturation characteristic offset, and the predicted harmonic error at different times, combined with the degradation category probability distribution, forms the performance degradation curve and degradation category evolution trajectory of the current transformer. This is used to complete the performance evaluation of the current transformer based on the performance degradation curve and the degradation category evolution trajectory. Specifically, this includes: The predicted values ​​of the ratio difference, the predicted values ​​of the phase angle error, the predicted values ​​of the saturation characteristic offset, and the predicted values ​​of the harmonic error are arranged in chronological order to form a sequence of performance prediction results covering the continuous excitation state of the current transformer. Identify the changing trends between different time windows in the performance prediction result sequence, and form a degradation category probability distribution sequence by combining the degradation category probability distributions that correspond one-to-one with the time windows of the performance prediction result sequence. By performing correlation analysis between the changing trend of the performance prediction result sequence and the dominant change of the degradation category probability distribution sequence, the performance degradation curve and the degradation category evolution trajectory are generated, enabling the current transformer to be determined in the following stages: normal operation stage, early stage of latent degradation, accelerated degradation stage, and near-failure stage.

6. The performance evaluation method for a current transformer according to claim 1, characterized in that, The method further includes: In the model calibration stage, a training dataset is constructed for current transformer samples with different degradation mechanisms and different life stages. Each target magnetoacoustic mode feature vector in the training dataset corresponds one-to-one with the target ratio difference calibration value, the target phase angle error calibration value, the target saturation characteristic offset calibration value, and the target harmonic error calibration value, and has a degradation category calibration label. During the model training phase, a magnetoacoustic-performance mapping model with regression and classification branches is trained using the training dataset. The regression branch outputs the target ratio error calibration value, target phase angle error calibration value, target saturation characteristic offset calibration value, and target harmonic error calibration value corresponding to the target magnetoacoustic modal feature vector. The classification branch outputs the target degradation category probability distribution corresponding to the target magnetoacoustic modal feature vector.

7. The performance evaluation method for a current transformer according to claim 1, characterized in that, The method further includes: Based on the rate of change and acceleration of change corresponding to the performance degradation curve, and the degree of abrupt change in the probability of key degradation categories in the degradation category evolution trajectory, the sampling period and sampling density of the magnetoacoustic excitation sequence are dynamically adjusted. Based on the sampling period and sampling density of the adjusted magnetoacoustic excitation sequence, when the current transformer enters the implicit degradation acceleration stage or is nearing failure stage, the acquisition frequency of the magnetostrictive micro-vibration response signal and the guided wave propagation response signal is increased, and a failure risk warning is output.

8. A performance evaluation device for a current transformer, characterized in that, The device is used to perform the performance evaluation method for a current transformer as described in any one of claims 1 to 7.

9. An electronic device, characterized in that, The electronic device includes a processor, a memory, a user interface, and a network interface. The memory is used to store instructions. The user interface and the network interface are both used to communicate with other devices. The processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores instructions that, when executed, perform the method as described in any one of claims 1 to 7.

Images