Method for evaluating mechanical strength of transformer winding under reclosing condition

Abstract

The present application relates to the technical field of mechanical property test of power transformer winding material, and particularly relates to a transformer winding mechanical strength evaluation method under reclosing conditions, which comprises the following steps: obtaining and preprocessing historical operation data, finite element simulation data and laboratory simulation test data; deploying micro foil strain gauges, MEMS acceleration sensors and flexible piezoresistive sensors based on the preprocessed data, and synchronously collecting radial strain, axial acceleration and pad pressure data; after noise filtering, time alignment and feature extraction, the multi-directional deformation synergistic influence is analyzed through Pearson correlation coefficient, partial least squares regression and other methods; and the evaluation model is trained in combination with historical damage samples, and the mechanical strength degradation grade is output in real time. The present application solves the problems of synchronous collection and correlation analysis of multi-directional deformation data under reclosing impact, and realizes accurate evaluation of overall mechanical strength change caused by cumulative deformation.

Description

Technical Field

[0001] This invention relates to the field of mechanical property testing technology for power transformer winding materials, and in particular to a method for evaluating the mechanical strength of transformer windings under reclosing conditions. Background Technology

[0002] The reclosing condition refers to the automatic reclosing operation performed after a circuit breaker in a power system trips due to a fault. This process generates transient overvoltages and inrush currents, which, when applied to transformer windings, can cause imbalances in the distribution of electromagnetic forces within the windings, distortion of the electric field in the insulation medium, and abnormal local temperature rises. By analyzing changes in winding inductance parameters to reflect the degree of inter-turn displacement, and considering that the resonant frequency formed by the winding inductance and inter-turn capacitance shifts due to winding deformation, high-frequency response analysis is used to collect frequency domain characteristics at different closing moments. Combined with a winding deformation and frequency response characteristic correlation model established by finite element simulation, the impact of the reclosing condition on the winding's mechanical integrity and insulation performance can be quantitatively assessed. Furthermore, the time domain characteristics of partial discharge signals correspond to the state of insulation medium degradation. By capturing parameters such as the amplitude and phase distribution of discharge pulses using ultra-high frequency sensors, the damage evolution of the winding insulation layer under reclosing impact can be indirectly determined.

[0003] In power transformers operating under repeated reclosing inrush currents, multi-dimensional testing and analysis methods for the degradation of winding mechanical strength present technical challenges. Existing technologies primarily test and analyze the radial or axial deformation of the winding under short-circuit impacts. However, in actual operation, reclosing impacts simultaneously induce multiple mechanical deformations, such as radial compression and axial vibration. These deformations are coupled and accelerate the degradation of winding mechanical performance. There is a lack of testing methods that can simultaneously collect and correlate multi-directional deformation data, making it difficult to accurately assess the overall change in winding mechanical strength due to cumulative deformation. For example, existing tests only monitor radial deformation using strain gauges and cannot simultaneously acquire data on the loosening of pads caused by axial vibration. This makes it impossible to comprehensively analyze the synergistic effect of radial compression and axial loosening on winding mechanical strength, and consequently, to determine whether the winding has lost its safe operating capability due to the accumulation of multi-dimensional deformations. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method for evaluating the mechanical strength of transformer windings under reclosing conditions. This method solves the problem of synchronously collecting and correlating multi-directional deformation data of power transformer windings under reclosing impact, so as to accurately assess the overall mechanical strength changes caused by cumulative deformation.

[0005] To solve the above-mentioned technical problems, the specific technical solution of the present invention is as follows:

[0006] The present invention provides a method for evaluating the mechanical strength of transformer windings under reclosing conditions, comprising:

[0007] Step 1: Acquire basic data including historical operation data, finite element simulation data and laboratory simulation test data. After deduplication and format unification preprocessing, high stress area data and easily damaged location data are generated to determine the sensor deployment location, as well as historical damage samples and simulation and test verification data for model training.

[0008] Step 2: Based on the high-stress area data and easily damaged location data obtained from preprocessing, micro foil strain gauges, MEMS accelerometers and flexible piezoresistive sensors are deployed in the winding area. The original time series data of radial strain, axial acceleration and pad pressure are synchronously acquired through FPGA synchronous controller.

[0009] Step 3: Perform sliding window mean filtering, 1ms time window alignment based on synchronization timestamp, and feature extraction on the collected raw time series data in sequence.

[0010] Step 4: Based on the multi-dimensional deformation characteristics such as the maximum radial deformation, the axial dominant frequency, and the pressure fluctuation range of the pad block extracted by feature extraction, the correlation between radial deformation and axial vibration amplitude is analyzed by Pearson correlation coefficient, and the correlation between axial dominant frequency and pad block pressure fluctuation is analyzed by partial least squares regression to identify the parameters that synergistically affect deformation in multiple directions.

[0011] Step 5: Input the identified multi-directional deformation synergistic influence parameters and actual damage samples from historical operating data into the training model to train an evaluation model that outputs the mechanical strength degradation level. Based on the real-time collected deformation characteristics and synergistic influence parameters, the evaluation model outputs the real-time winding mechanical strength evaluation result.

[0012] Furthermore, the method for evaluating the mechanical strength of transformer windings under reclosing conditions described in this invention also includes:

[0013] The historical operating data consists of records of reclosing events of transformers over 5 years and corresponding disassembly and inspection reports, which are used to provide a database of reclosing conditions and actual damage samples required for model training.

[0014] The finite element simulation data simulates the distribution of winding electromagnetic force under different reclosing conditions through a multi-physics coupling model, and outputs the location of the maximum radial stress, axial vibration frequency and pad contact pressure threshold to identify the high-stress area of ​​sensor deployment.

[0015] The laboratory simulation test data uses a 220kV transformer scaled-down model and a controllable impulse current generator to simulate reclosing conditions, records winding deformation data, and disassembles to verify actual damage, in order to verify the consistency between simulation data and actual operating data.

[0016] Furthermore, the transformer winding mechanical strength assessment method under reclosing conditions described in this invention uses sensors including miniature foil strain gauges, MEMS accelerometers, and flexible piezoresistive sensors.

[0017] The micro foil strain gauges are deployed in the region of maximum radial stress determined by the finite element simulation data in step 1, and are used to monitor radial tensile or compressive strain values ​​in real time.

[0018] The MEMS accelerometer is deployed in the gap of the disc pad where the axial vibration frequency shift is significant, as recorded in the historical dismantling detection report in step 1, to monitor the acceleration amplitude and frequency of axial vibration.

[0019] The flexible piezoresistive sensor is deployed in the region with the lowest contact pressure threshold of the pad determined by the finite element simulation data in step 1, and indirectly reflects the degree of loosening of the pad through pressure changes.

[0020] All sensors are aligned to microsecond-level timestamps via an FPGA synchronization controller, enabling the acquisition of radial strain, axial acceleration, and pad pressure data at the same time reference.

[0021] Furthermore, in the method for evaluating the mechanical strength of transformer windings under reclosing conditions described in this invention, step 1 includes:

[0022] Noise filtering employs sliding window mean filtering to remove high-frequency electromagnetic interference noise while retaining low-frequency deformation characteristics related to reclosing impact.

[0023] Time alignment is based on the synchronization timestamp generated by the FPGA synchronization controller in step 2. The radial strain, axial acceleration and pad pressure data are divided into time windows in 1ms units and matched with the reclosing event timestamps of the historical running data in step 1 to form a time-aligned multi-dimensional dataset.

[0024] Feature extraction includes:

[0025] Extract the maximum deformation and the rate of change of deformation within each time window;

[0026] Perform FFT transformation on the acceleration signal to extract the dominant frequency and vibration amplitude;

[0027] Extract the fluctuation range and pressure drop rate of the pressure signal;

[0028] The extracted multidimensional deformation features are used as input parameters for the correlation analysis in step 4.

[0029] Furthermore, the method for evaluating the mechanical strength of transformer windings under reclosing conditions described in this invention also includes:

[0030] The correlation between the radial deformation rate of change and the axial vibration amplitude is calculated using the Pearson correlation coefficient, and the radial and axial coupling coefficients are output.

[0031] Partial least squares regression analysis was used to analyze the correlation between the pressure fluctuation range of the pad block and the axial dominant frequency, and the correlation coefficient between the axial direction and the pad block was output.

[0032] By establishing a multivariate covariance matrix, the covariances of radial, axial and pad deformation characteristics are calculated to identify the dominant degradation factors;

[0033] The coupling coefficient, correlation coefficient, and dominant degradation factor output by the correlation analysis are used as input parameters for model training in step 5.

[0034] Furthermore, the mechanical strength assessment method for transformer windings under reclosing conditions described in this invention includes the following input parameters for the mechanical strength assessment model: the extracted maximum radial deformation, the axial main frequency offset, and the pressure drop rate of the pad block.

[0035] The multi-directional deformation coupling strength output by the correlation analysis includes radial and axial coupling coefficients, and axial and pad correlation coefficients.

[0036] The mechanical strength assessment model outputs parameters representing the mechanical strength degradation level, where:

[0037] First mechanical strength degradation level: maximum radial deformation ≤ preset threshold, axial dominant frequency offset ≤ 5%, pad pressure drop rate ≤ 0.5 N / s and multi-directional deformation coupling strength ≤ 0.3;

[0038] Second mechanical strength degradation level: Any parameter exceeds the threshold of the first mechanical strength degradation level but does not meet the standard of the third mechanical strength degradation level;

[0039] Third mechanical strength degradation level: maximum radial deformation > 2 times the preset threshold or axial main vibration frequency offset > 10% or pad pressure drop rate > 1.0 N / s;

[0040] The mechanical strength assessment model is trained using a random forest algorithm or a long short-term memory network. It uses actual damage samples from historical operating data as labels and optimizes tree depth or time step parameters through cross-validation to capture the nonlinear relationship between multi-dimensional features and degradation level.

[0041] Furthermore, in the method for evaluating the mechanical strength of transformer windings under reclosing conditions described in this invention, the mechanical strength evaluation model is verified using 20% ​​of the samples in the laboratory simulation test data that were not used for training, and the degradation level output by the mechanical strength evaluation model is compared with the actual disassembly damage level.

[0042] When the matching degree is ≥90%, the matching degree is the percentage of samples whose output level of the mechanical strength assessment model matches the actual level, and the model calibration is completed.

[0043] The calibrated mechanical strength assessment model is used for real-time assessment in step 5.

[0044] Furthermore, the method for evaluating the mechanical strength of transformer windings under reclosing conditions described in this invention includes real-time evaluation of:

[0045] Receive real-time sensor data synchronously acquired by the FPGA synchronous controller in step 2;

[0046] After noise filtering, time alignment and feature extraction in step 3, the real-time data is input into the correlation analysis in step 4.

[0047] Input the preprocessed and correlation-analyzed real-time feature parameters into the mechanical strength assessment model trained in step 5, and output the real-time mechanical strength degradation level.

[0048] Real-time mechanical strength degradation levels include:

[0049] First mechanical strength degradation level: Real-time data and evaluation results are recorded in the database to maintain the regular inspection cycle;

[0050] Second mechanical strength degradation level: Generate early warning information to prompt a shortening of the local monitoring cycle;

[0051] Third mechanical strength degradation level: Generate maintenance suggestions for pushing to the operation and maintenance system.

[0052] Furthermore, the method for evaluating the mechanical strength of transformer windings under reclosing conditions described in this invention also includes: historical operating data is preprocessed in step 1 to form a sample library of reclosing conditions and actual damage, providing real damage labels for model training in step 5.

[0053] The finite element simulation data, through preprocessing in step 1, clarifies the high-stress region and the theoretical deformation threshold, providing a theoretical basis for sensor deployment in step 2.

[0054] The laboratory simulation test data is preprocessed in step 1 to verify the consistency between the simulation data and the actual operation data.

[0055] Furthermore, the method for evaluating the mechanical strength of transformer windings under reclosing conditions described in this invention also includes:

[0056] The FPGA synchronization controller achieves hardware clock synchronization via PTP;

[0057] By combining software timestamp correction, time offsets caused by differences in hardware characteristics of different sensors are eliminated;

[0058] Through the synergy of hardware synchronization and software correction, the time alignment in step 3 is used to accurately match the timestamps of historical reclosing events, thereby avoiding biases in correlation analysis caused by time misalignment.

[0059] Beneficial effects of this invention;

[0060] This invention effectively addresses the technical challenges of synchronous acquisition and correlation analysis of multi-directional deformation data under reclosing impact through a multi-source data fusion and multi-dimensional analysis framework: Preprocessing and coordinating historical operating data, finite element simulation data, and laboratory simulation test data provide multi-dimensional information on high-stress areas and vulnerable locations for sensor deployment, avoiding the limitations of a single data source; Synchronous deployment of micro-foil strain gauges, MEMS accelerometers, and flexible piezoresistive sensors, aligned with FPGA timestamps, enables microsecond-level synchronous acquisition of radial deformation, axial vibration, and pad loosening data, resolving the problem of time misalignment in multi-directional deformation data; The application of sliding window filtering, time window alignment, and feature extraction techniques... The method extracts key features reflecting instantaneous stress intensity, structural stiffness changes, and contact stability, providing quantitative parameters for correlation analysis. Joint analysis of Pearson correlation coefficient, partial least squares regression, and multivariate covariance matrix quantifies the synergistic effects of multi-directional deformation, such as radial and axial coupling and axial and pad-block correlation, overcoming the limitations of single-directional deformation analysis. Training and real-time evaluation of random forest or LSTM models, by capturing the nonlinear relationship between multi-dimensional features and degradation levels, accurately assesses the overall mechanical strength changes of windings caused by cumulative deformation, providing decision-making basis for power transformer operation and maintenance based on multi-dimensional test data, meeting the technical requirements for multi-dimensional analysis and testing of materials or objects. Attached Figure Description

[0061] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on the drawings without creative effort.

[0062] Figure 1 This is a flowchart of a method for evaluating the mechanical strength of transformer windings under reclosing conditions, provided in an embodiment of the present invention. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention. The technical solutions provided by various embodiments of this invention will be described in detail below with reference to the accompanying drawings. To better understand the objectives of this invention, it will be described in further detail below.

[0064] Please see Figure 1 The present invention provides a method for evaluating the mechanical strength of transformer windings under reclosing conditions, comprising:

[0065] Step 1: Acquire basic data including historical operation data, finite element simulation data and laboratory simulation test data. After deduplication and format unification preprocessing, high stress area data and easily damaged location data are generated to determine the sensor deployment location, as well as historical damage samples and simulation and test verification data for model training.

[0066] Step 2: Based on the high-stress area data and easily damaged location data obtained from preprocessing, micro foil strain gauges, MEMS accelerometers and flexible piezoresistive sensors are deployed in the key areas of the winding. The original time series data of radial strain, axial acceleration and pad pressure are synchronously collected through the FPGA synchronous controller.

[0067] Step 3: Perform sliding window mean filtering, 1ms time window alignment based on synchronization timestamp, and feature extraction on the collected raw time series data in sequence.

[0068] Step 4: Based on the multi-dimensional deformation characteristics such as the maximum radial deformation, the axial dominant frequency, and the pressure fluctuation range of the pad block extracted by feature extraction, the correlation between radial deformation and axial vibration amplitude is analyzed by Pearson correlation coefficient, and the correlation between axial dominant frequency and pad block pressure fluctuation is analyzed by partial least squares regression to identify the parameters that synergistically affect deformation in multiple directions.

[0069] Step 5: Input the identified multi-directional deformation synergistic influence parameters and actual damage samples from historical operating data into the training model to train an evaluation model that outputs the mechanical strength degradation level. Based on the real-time collected deformation characteristics and synergistic influence parameters, the evaluation model outputs the real-time winding mechanical strength evaluation result.

[0070] Step 1, data fusion preprocessing, includes: constructing a multi-dimensional basic database by integrating three heterogeneous data sources: historical transformer operating data, finite element simulation data, and laboratory simulation test data. Historical operating data is extracted from reclosing event sequences over five years recorded by the power grid dispatch system and corresponding transformer disassembly and inspection reports, including physical damage labels such as actual winding deformation locations and insulation pad displacement. Finite element simulation data calculates the winding electromagnetic force distribution under different reclosing phase angles based on a multi-physics coupling model (electromagnetic, structural, and thermal), outputting radial stress cloud maps, axial vibration mode frequencies, and pad contact pressure distribution thresholds. Laboratory simulation test data uses a 220kV transformer scaled-down model, reproducing the reclosing transient process using a controllable impulse current generator, synchronously recording winding deformation sensor data, and calibrating the actual mechanical damage level through disassembly. The three types of data undergo ETL processes for redundant record deduplication, unified timestamp conversion, and dimensional standardization, ultimately generating:

[0071] High-stress region coordinate set: composed of the radial stress peak coordinates and the axial vibration energy concentration region coordinates output by finite element simulation.

[0072] Vulnerable location mapping table: established based on the frequently occurring line disc numbers and pad displacement positions in historical dismantling reports.

[0073] Damage Sample Feature Library: Historical damage samples are associated with and stored with corresponding operating parameters (impact current amplitude, reclosing interval duration).

[0074] This step provides a theoretical basis for optimized sensor deployment through multi-source data spatial registration and feature alignment, and builds a physically labeled dataset for model training.

[0075] Step 2, multi-sensor collaborative deployment and synchronous data acquisition, includes:

[0076] Based on the high-stress region coordinate set and easily damaged location mapping table output in step 1, multiple types of sensors are embedded and deployed on the transformer winding entity:

[0077] Miniature foil strain gauges: These are bonded to the region of maximum radial stress (such as the outer surface of the coil in the middle of a low-voltage winding) by epoxy resin to monitor the transient response of radial tensile / compressive strain of the winding.

[0078] MEMS accelerometer: Installed in the gap between the pads of the wire disc where the axial vibration frequency deviation exceeds 5% as recorded in the historical dismantling report, the time-domain waveform of axial vibration is captured by triaxial acceleration measurement.

[0079] Flexible piezoresistive sensor: Embedded in the area where the contact pressure of the pad is 20% below the design threshold as displayed by finite element simulation, it measures the dynamic pressure change of the insulating pad.

[0080] All sensor signals are connected to the FPGA synchronization controller via shielded cables. This controller uses the IEEE 1588 PTP protocol to achieve hardware clock synchronization and writes a microsecond-precision timestamp when the data acquisition card is triggered. Through the FPGA's parallel sampling architecture, the original time series of radial strain, axial acceleration, and pad pressure signals are captured synchronously, eliminating the data phase difference caused by traditional time-division sampling.

[0081] Step 3, signal processing and feature engineering, includes:

[0082] A three-stage processing pipeline is executed on the raw time series acquired by the FPGA:

[0083] Noise filtering: A sliding mean filter with adjustable window width is used to filter out high-frequency noise (>2kHz) caused by magnetostriction of the transformer core, while retaining the mechanical deformation characteristics in the 0 and 500Hz frequency bands.

[0084] Time alignment: Based on the synchronization timestamp written by the FPGA, the data from the three sensors are re-sliced ​​into 1ms time windows and matched with the time reference of historical reclosing events to form a multi-channel dataset with strict time alignment.

[0085] Feature extraction:

[0086] Radial strain channel: Calculates the maximum absolute value of deformation within each time window and the deformation rate between adjacent windows.

[0087] Axial acceleration channel: The 0 and 200Hz main oscillation frequency components and their amplitude envelopes are extracted by FFT transformation.

[0088] Pressure channel of pad block: statistical pressure fluctuation range (maximum and minimum values) and pressure drop slope fitted by linear regression.

[0089] The multi-dimensional feature vector output in this step serves as the input for subsequent correlation analysis. Feature design focuses on characterizing the instantaneous strength, rate of change of structural stiffness, and contact stability of the winding's mechanical response.

[0090] Step 4, the multidimensional deformation coupling analysis, includes:

[0091] Based on the feature vectors extracted in step 3, a coupling analysis model between deformation directions is established:

[0092] Radial and axial correlation analysis: The linear correlation between radial deformation rate and axial vibration amplitude is calculated using the Pearson correlation coefficient, and the radial and axial coupling coefficients (with a scale of 1 to 1) are output to quantify the energy transfer effect of winding radial compression on axial vibration.

[0093] Axial and pad correlation modeling: Partial least squares regression (PLSR) is applied to analyze the mapping relationship between the axial dominant frequency and the pad pressure fluctuation range, and the axial and pad correlation coefficient is output to characterize the degree of pad loosening caused by axial vibration.

[0094] Multivariate covariance analysis: Construct a three-dimensional covariance matrix of radial deformation, axial dominant frequency offset, and pad pressure drop rate. Identify the primary factors that dominate mechanical degradation through eigenvalue decomposition (such as the dimension corresponding to the largest eigenvalue). The coupling coefficient, correlation coefficient, and dominant factors output in this step together constitute a set of parameters for the synergistic influence of multi-directional deformation, revealing the interaction mechanism between different deformation modes.

[0095] Step 5, the assessment of mechanical strength degradation, includes:

[0096] Establish a machine learning-based mechanical strength assessment model:

[0097] Input layer: Receives the basic features of radial maximum deformation, axial main frequency offset, and pad pressure drop rate extracted in step 3, as well as the coupling coefficient, correlation coefficient, and collaborative parameters output in step 4.

[0098] Model architecture: A multi-decision tree ensemble model is constructed using the random forest algorithm, or a long short-term memory network (LSTM) is applied to process time series features; the actual damage level in the historical damage sample library is used as the supervision label.

[0099] Training mechanism: The model hyperparameters (such as the tree depth of random forest and the time step of LSTM) are optimized through k-fold cross-validation to learn the nonlinear mapping between multidimensional features and mechanical strength degradation.

[0100] Real-time assessment: The system receives preprocessed sensor data streams online, performs feature extraction and correlation analysis, then inputs the data into the trained model, outputting mechanical strength degradation levels I and III.

[0101] Level I: All parameters are below the safety threshold; maintain routine monitoring.

[0102] Level II: A single parameter exceeds the limit but does not reach the critical value, and high-frequency local monitoring is activated.

[0103] Level III: Key parameters exceed the failure threshold, generating maintenance instructions. This step achieves a quantitative assessment of accumulated mechanical damage to the windings by integrating multi-dimensional deformation characteristics and their synergistic effect parameters.

[0104] Specifically, the method for evaluating the mechanical strength of transformer windings under reclosing conditions described in this invention includes multiple types of sensors such as miniature foil strain gauges, MEMS accelerometers, and flexible piezoresistive sensors.

[0105] The micro foil strain gauges are deployed in the region of maximum radial stress determined by the finite element simulation data in step 1, and are used to monitor radial tensile or compressive strain values ​​in real time.

[0106] The MEMS accelerometer is deployed in the gap of the disc pad where the axial vibration frequency shift is significant, as recorded in the historical dismantling detection report in step 1, to monitor the acceleration amplitude and frequency of axial vibration.

[0107] The flexible piezoresistive sensor is deployed in the region with the lowest contact pressure threshold of the pad determined by the finite element simulation data in step 1, and indirectly reflects the degree of loosening of the pad through pressure changes.

[0108] All sensors are aligned to microsecond-level timestamps via an FPGA synchronization controller, enabling the acquisition of radial strain, axial acceleration, and pad pressure data at the same time reference.

[0109] The sensor selection and deployment logic includes:

[0110] The deployment of the micro foil strain gauges is based on the radial stress distribution cloud map output from the finite element simulation in step 1. The region of the wire disc where the stress concentration exceeds 80% of the material's yield strength is selected as the installation site. In practice, the strain gauge substrate is bonded to the outer surface of the winding conductor using high-temperature resistant epoxy resin, with its grid wire direction parallel to the radial force direction. This allows for real-time capture of the micro-strain signal of the winding's radial elastic deformation caused by reclosing impact. This deployment method directly correlates with the mechanical weak points revealed by the finite element simulation, ensuring that the measured data reflects the actual strain state of the theoretically predicted high-risk areas.

[0111] The placement of MEMS accelerometers was based on axial vibration anomaly records from historical dismantling inspection reports. For the gaps between the winding blocks where frequency deviations exceeded ±5% of the reference value after multiple reclosing events, triaxial accelerometers were embedded at the corresponding coordinate points on the three-phase windings. The sensors are mechanically coupled to the block surface via rigid supports, with the measurement direction perpendicular to the block plane, and are used to acquire the time-domain acceleration waveform and spectral characteristics of the winding's axial vibration. This deployment strategy maps historical damage locations to real-time monitoring points, establishing a direct correlation between axial vibration data and known mechanical fault modes.

[0112] The integration location of the flexible piezoresistive sensor is determined by the critical contact pressure region of the pad determined by finite element simulation. A thin-film piezoresistive sensing unit is embedded on the pad surface where the simulation results show the pressure value is 15% lower than the design clamping force threshold. The upper and lower surfaces of the sensor are in direct contact with the insulating pad and the end face of the wire disc, and the dynamic pressure distribution of the pad is inverted through the rate of change of resistance. When the pad loosens due to a reclosing impact, the pressure drop process is converted into a continuous electrical signal output, indirectly characterizing the degree of stability degradation of the insulation support structure.

[0113] Multi-source data synchronous acquisition mechanisms include:

[0114] The output signals from three types of sensors are connected to the multi-channel acquisition card of the FPGA synchronous controller via shielded twisted-pair cables. The controller has a built-in IEEE 1588 Precision Time Protocol (PTP) module, which receives the master clock signal through a fiber optic network to achieve hardware clock synchronization for each acquisition channel. Upon each reclosing event, the controller injects a microsecond-precision timestamp into all data channels. Synchronous acquisition is achieved using a parallel sampling architecture.

[0115] Analog voltage signal of radial strain channel (corresponding to 0 and 5000με range).

[0116] Triaxial acceleration digital signal (±50g range);

[0117] Piezoresistive sensor bridge output signal (0 and 10MPa range).

[0118] The raw data is transmitted to the buffer in a fixed frame format, with each frame including the sensor ID, synchronization timestamp, and sampled value. This mechanism eliminates the phase error caused by traditional time-division sampling, ensuring that the transient responses of different physical quantities are under the same time reference, thus establishing a temporal consistency basis for the multidimensional feature correlation analysis in step 3.

[0119] Specifically, the method for evaluating the mechanical strength of transformer windings under reclosing conditions described in this invention uses sliding window mean filtering to filter out high-frequency electromagnetic interference noise and retain low-frequency deformation characteristics related to reclosing impact.

[0120] The time alignment is based on the synchronization timestamp generated by the FPGA synchronization controller in step 2. The radial strain, axial acceleration and pad pressure data are divided into time windows in 1ms units and matched with the reclosing event timestamps of the historical running data in step 1 to form a time-aligned multi-dimensional dataset.

[0121] The feature extraction includes:

[0122] Extract the maximum deformation and the rate of change of deformation within each time window;

[0123] Perform FFT transformation on the acceleration signal to extract the dominant frequency and vibration amplitude;

[0124] Extract the fluctuation range and pressure drop rate of the pressure signal;

[0125] The extracted multidimensional deformation features are used as input parameters for the correlation analysis in step 4.

[0126] Noise filtering mechanisms include:

[0127] A sliding window mean filtering algorithm was used to process the raw sensor signals, with the window width set according to the spectral characteristics of the transformer core magnetostrictive noise. The filtering process was performed separately on the radial strain, axial acceleration, and pad pressure signals: continuous sampling points were grouped according to a fixed window length, and the arithmetic mean of each group was calculated as the output value. This operation effectively suppressed high-frequency noise components (>2kHz band) caused by core vibration and electromagnetic interference, while retaining low-frequency deformation characteristics (0 and 500Hz band) related to the mechanical impact of reclosing. The filtered signal retained effective mechanical response information such as the radial elastic deformation of the winding, axial structural vibration, and changes in pad contact pressure, providing a denoised data basis for subsequent feature extraction.

[0128] Multi-channel time alignment methods include:

[0129] Based on the microsecond-precision timestamps generated by the FPGA synchronization controller in step 2, the time base of the three sensor data streams is unified. Specific implementation includes:

[0130] The filtered radial strain, axial acceleration, and pad pressure data were divided into discrete time windows at 1ms intervals.

[0131] The data in each window is marked with a start timestamp and matched with the time coordinates of the reclosing events in the historical running database of Step 1.

[0132] Phase compensation is performed on cross-window data points using a linear interpolation algorithm.

[0133] A strictly time-aligned multi-dimensional dataset is formed, establishing a correspondence between the transient responses of different physical quantities under the same time coordinate. This alignment mechanism solves the data misalignment problem caused by sensor response delays and provides temporal consistency assurance for multi-directional deformation collaborative analysis.

[0134] The multi-dimensional feature extraction process includes:

[0135] Perform feature quantization within the time-aligned data window:

[0136] Radial strain channel: The maximum absolute value of strain within the calculation window is used as the instantaneous deformation strength index. The strain change rate between adjacent sampling points is calculated using the backward difference method to characterize the dynamic change characteristics of the radial stiffness of the winding.

[0137] Axial acceleration channel: Perform Fast Fourier Transform (FFT) on the acceleration time domain signal to identify the main frequency components with energy accounting for more than 60% of the total energy in the 0 and 200 Hz frequency bands, and simultaneously extract the peak value of the vibration amplitude envelope corresponding to the frequency.

[0138] Pressure channel of pad block: The fluctuation range of pressure sampling values ​​within the statistical window (the difference between the maximum and minimum values), and the slope of the pressure-time curve fitted by the least squares method is used as the pressure drop rate.

[0139] The extracted feature parameters constitute a multidimensional vector characterizing the mechanical response, including radial deformation intensity, axial vibration spectrum characteristics, and pad contact stability index. This vector serves as the input for the correlation analysis in step 4, realizing the transformation from the original signal to quantitative parameters of mechanical behavior.

[0140] Specifically, the method for evaluating the mechanical strength of transformer windings under reclosing conditions described in this invention includes the following sub-analyses in its correlation analysis, which takes the radial deformation rate of change, axial vibration amplitude, pad pressure fluctuation range, and axial dominant frequency from the extracted multi-dimensional deformation features as input:

[0141] The correlation between the radial deformation rate of change and the axial vibration amplitude is calculated using the Pearson correlation coefficient, and the radial and axial coupling coefficients are output.

[0142] Partial least squares regression analysis was used to analyze the correlation between the pressure fluctuation range of the pad block and the axial dominant frequency, and the correlation coefficient between the axial direction and the pad block was output.

[0143] By establishing a multivariate covariance matrix, the covariances of radial, axial and pad deformation characteristics are calculated to identify the dominant degradation factors;

[0144] The coupling coefficient, correlation coefficient, and dominant degradation factor output by the correlation analysis are used as input parameters for model training in step 5.

[0145] The quantification of radial and axial coupling effects includes:

[0146] Based on the radial deformation rate of change and axial vibration amplitude sequences extracted in step 3, the Pearson correlation coefficient algorithm is used to calculate their linear correlation. Specifically, within a single reclosing event time window, the radial deformation rate of change is used as the independent variable X, and the axial vibration amplitude is used as the dependent variable Y. The ratio of covariance to standard deviation is calculated. The radial and axial coupling coefficients are output (ranging from 1 to 1). Positive values ​​indicate that radial compression intensifies axial vibration, while negative values ​​reflect that radial stretching inhibits axial vibration. This coefficient quantifies the modulation effect of winding radial deformation on axial dynamic behavior, revealing the law of mechanical energy transmission in orthogonal directions.

[0147] The axial and pad block association modeling methods include:

[0148] For the two characteristic parameters of axial dominant frequency and pressure fluctuation range of the pad block, a multivariate mapping model is established using partial least squares regression (PLSR). The axial dominant frequency is used as the predictor variable matrix, and the pressure fluctuation range of the pad block is used as the response variable matrix. A regression equation is established by iteratively extracting latent variables. The axial-pad correlation coefficient (standardized regression coefficient) is output, characterizing the change in pressure fluctuation amplitude caused by a unit frequency change. This coefficient reflects the efficiency of axial vibration energy transfer to the pad block contact interface and is used to assess the stability degradation trend of the insulation support system.

[0149] Identification of multidimensional degradation-dominant factors includes:

[0150] A three-dimensional covariance matrix was constructed, comprising the maximum radial deformation, axial dominant frequency offset, and pad pressure drop rate. The calculation process involved first standardizing the three eigenvectors using Z and scores, then calculating the covariance between any two eigenvectors to fill the matrix elements. The eigenvalues ​​and eigenvectors of the covariance matrix were solved using the Jacobi iteration method, and the eigenvector component corresponding to the largest eigenvalue was used as the weight of the dominant degradation factor. For example, if the radial deformation component weight exceeded 0.7, radial deformation was determined to be the primary cause of mechanical strength degradation. This analysis reveals the dominant failure mode in the synergistic effect of multi-directional deformation.

[0151] Specifically, the mechanical strength assessment method for transformer windings under reclosing conditions described in this invention includes the following input parameters for the mechanical strength assessment model: the extracted maximum radial deformation, the axial main frequency offset, and the pressure drop rate of the pad block.

[0152] The multi-directional deformation coupling strength output by the correlation analysis includes radial and axial coupling coefficients, and axial and pad correlation coefficients.

[0153] The mechanical strength assessment model outputs parameters as the mechanical strength degradation level, where:

[0154] First mechanical strength degradation level: maximum radial deformation ≤ preset threshold, axial dominant frequency offset ≤ 5%, pad pressure drop rate ≤ 0.5 N / s and multi-directional deformation coupling strength ≤ 0.3;

[0155] Second mechanical strength degradation level: Any parameter exceeds the threshold of the first mechanical strength degradation level but does not meet the standard of the third mechanical strength degradation level;

[0156] Third mechanical strength degradation level: maximum radial deformation > 2 times the preset threshold or axial main vibration frequency offset > 10% or pad pressure drop rate > 1.0 N / s;

[0157] The mechanical strength assessment model is trained using a random forest algorithm or a long short-term memory network. It uses actual damage samples from historical operating data as labels and optimizes tree depth or time step parameters through cross-validation to capture the nonlinear relationship between multi-dimensional features and degradation level.

[0158] Model input parameter construction includes:

[0159] The input layer of the mechanical strength assessment model receives two sets of parameters: the first set consists of the basic deformation features output by the feature extraction module in step 3, including the absolute value of the maximum deformation in the radial strain channel, the percentage offset of the dominant frequency of the axial acceleration channel relative to the reference value, and the pressure drop rate per unit time in the pad pressure channel; the second set consists of the synergistic influence parameters generated by the correlation analysis in step 4, including the Pearson coupling coefficient between the radial deformation rate of change and the axial vibration amplitude, and the partial least squares correlation coefficient between the axial dominant frequency and the pad pressure fluctuation range. After standardization, the input parameters form a five-dimensional feature vector, comprehensively characterizing the instantaneous mechanical response of the winding under reclosing impact and the intensity of multi-directional deformation interaction.

[0160] The criteria for classifying degradation levels include:

[0161] The model output uses a discretized mechanical strength degradation level classification:

[0162] The first level corresponds to the winding structure integrity maintenance state, requiring that the radial deformation does not exceed the material elastic limit threshold, the axial vibration frequency deviation is within the design allowable tolerance zone (≤5%), the pad pressure decay rate is maintained in the linear viscoelastic deformation stage (≤0.5N / s), and the multi-directional coupling effect does not cause abnormal energy accumulation (coupling strength ≤0.3).

[0163] The second level indicates local mechanical property deterioration, which is triggered when any basic parameter exceeds the threshold of the first level but does not reach the critical failure value. For example, radial deformation enters the plastic deformation range but does not exceed twice the elastic limit.

[0164] The third level is defined as a structural failure risk state, which is activated when the radial deformation exceeds twice the material yield strength threshold, or the axial vibration frequency shift leads to mechanical resonance (>10%), or the pad pressure decay rate reflects the instability of the support structure (>1.0 N / s).

[0165] This classification logic is based on the material mechanical property threshold and the dynamic stability theory of mechanical systems, and realizes the quantitative classification of damage degree.

[0166] Machine learning model training mechanisms include:

[0167] The evaluation model is constructed using random forest or long short-term memory network architecture:

[0168] Random Forest Implementation: Using actual damage samples from the historical operational database as supervisory labels (Levels I and III), multiple decision trees are constructed for ensemble learning. Hyperparameters such as tree depth and minimum number of samples per split node are optimized through grid search and k-fold cross-validation, enabling the model to learn the nonlinear decision boundary between multi-dimensional features and degradation levels.

[0169] LSTM Implementation: The time-window sequence features are input into a recurrent neural network, utilizing hidden state memory units to capture the temporal dependence of deformation parameters. The training process is optimized through time step adjustment and gradient pruning to extract the dynamic evolutionary patterns of mechanical performance degradation.

[0170] The trained model can output the corresponding mechanical strength degradation level probability distribution based on the real-time input five-dimensional feature vector, and take the highest probability level as the evaluation result.

[0171] Specifically, the mechanical strength assessment method for transformer windings under reclosing conditions described in this invention uses 20% of the laboratory simulation test data that were not used in training to verify the mechanical strength assessment model, and compares the degradation level output by the mechanical strength assessment model with the actual disassembly damage level.

[0172] When the matching degree is ≥90%, the matching degree is the percentage of samples whose output level of the mechanical strength assessment model matches the actual level, and the model calibration is completed.

[0173] The calibrated mechanical strength assessment model is used for real-time assessment in step 5.

[0174] The validation sample construction mechanism includes:

[0175] The model validation phase utilizes an independent dataset generated from laboratory simulation experiments, which was separated from the original experimental data during the preprocessing step 1. Specifically, 20% of the test samples from the reclosing simulation experiments conducted using a scaled-down 220kV transformer model were randomly selected as the validation set. The deformation characteristic parameters and co-influence parameters corresponding to these samples were not included in the model training process. Each validation sample was associated with the actual mechanical damage level recorded in the post-test winding disintegration detection report, including damage labels of Level I and III converted from physical measurements such as radial permanent deformation and pad displacement distance. The validation set construction process ensured that the data distribution was from the same source as the training set but independent, avoiding model evaluation bias.

[0176] The methods for calculating the matching degree include:

[0177] The validation set samples are input into the trained mechanical strength assessment model to obtain the degradation level prediction results output by the model. The comparative analysis process is as follows:

[0178] Compare the model output level with the actual disintegration damage level on a sample-by-sample basis;

[0179] The proportion of samples with identical statistical ranks to the total validation samples;

[0180] When the proportion reaches the preset threshold (≥90%), the model is deemed to meet the accuracy requirements.

[0181] The matching degree calculation only recognizes samples with a perfect level match, excluding partial matches. For example, if the model outputs level II and the actual level is level II, it is considered a match; if the output is level II but the actual level is level I or III, it is considered a mismatch. This strict standard ensures the reliability of the evaluation results.

[0182] Model calibration technology implementation includes:

[0183] When the matching degree does not reach the threshold, the model calibration procedure is initiated:

[0184] Analyze the characteristic distribution patterns of mismatched samples to identify high-frequency misjudgment parameter combinations;

[0185] Adjust the feature weight allocation of the random forest or the time window length of the LSTM;

[0186] Add augmented samples of similar operating conditions to the training set;

[0187] Retrain and test using the same validation set;

[0188] The calibration cycle continues until the matching degree meets the standard. The calibrated model is integrated into the online monitoring system, receives the sensor data stream acquired in real time in step 2, and after feature extraction in step 3 and correlation analysis in step 4, performs real-time assessment of the mechanical strength degradation level in step 5.

[0189] Specifically, the method for evaluating the mechanical strength of transformer windings under reclosing conditions according to the present invention includes the following real-time evaluation:

[0190] Receive real-time sensor data synchronously acquired by the FPGA synchronous controller in step 2;

[0191] After the real-time data undergoes noise filtering, time alignment, and feature extraction in step 3, it is input into the correlation analysis in step 4.

[0192] Input the preprocessed and correlation-analyzed real-time feature parameters into the mechanical strength assessment model trained in step 5, and output the real-time mechanical strength degradation level.

[0193] Perform the following operations based on the output level:

[0194] First mechanical strength degradation level: Real-time data and evaluation results are recorded in the database to maintain the regular inspection cycle;

[0195] Second mechanical strength degradation level: Generate early warning information to prompt a shortening of the local monitoring cycle;

[0196] Third mechanical strength degradation level: Generate maintenance suggestions for pushing to the operation and maintenance system.

[0197] Real-time data stream processing mechanisms include:

[0198] The online monitoring system continuously receives the raw sensor data stream transmitted by the FPGA synchronous controller in step 2. This data stream includes synchronous sampling sequences of three channels: radial strain, axial acceleration, and pad pressure. Each frame of data carries a timestamp with microsecond-level precision. The data processing engine calls the signal processing pipeline defined in step 3: first, it applies a sliding window mean filter to eliminate electromagnetic interference noise; then, it slices and aligns the three signals at 1ms intervals according to the timestamps; finally, it executes a feature extraction algorithm to generate feature vectors such as maximum radial deformation, axial dominant frequency offset, and pad pressure drop rate. The processed feature set is input into the correlation analysis module in step 4 to calculate the coupling coefficient and correlation coefficient, forming a complete set of real-time evaluation input parameters.

[0199] Model reasoning and ranking output include:

[0200] The real-time generated feature parameter set is input into the calibrated mechanical strength assessment model trained in step 5. The model performs inference operations based on a random forest or long short-term memory network architecture: the random forest uses a multi-decision tree voting mechanism to determine the probability distribution of degradation levels, and the LSTM captures the feature evolution trend through time-series memory units. Discretized evaluation results are output according to a preset threshold logic.

[0201] Level 1: The model determines that all input parameters are within a safe threshold range;

[0202] Level 2: One or more parameters exceed the safety threshold but do not reach the critical failure value;

[0203] Level 3: Key parameters exceed the material or structural failure threshold;

[0204] The output includes confidence level indicators and trigger parameter identifiers, forming a structured evaluation report.

[0205] The implementation of a tiered response strategy includes:

[0206] Activate differentiated operation and maintenance response protocols based on output level:

[0207] Level 1 response: Compress and store the raw sensor data, characteristic parameters and evaluation results into the time series database, maintain the preset routine inspection cycle (such as full data collection once every 24 hours), and the system maintains the baseline monitoring status.

[0208] Level 2 Response: Automatically generates standardized early warning messages. The message content includes the type of abnormal parameter (e.g., "axial frequency offset 7.5%"), location information (associated sensor installation coordinates), and monitoring strategy upgrade instructions. The instructions trigger an increase in the sensor sampling rate to 1kHz in a specific area and initiate high-frequency data recording mode.

[0209] Level 3 Response: The maintenance management system interface is invoked to push maintenance work orders. The structured fields of the work order include the damage location topology code (e.g., "high voltage winding A phase 3rd wire disc"), failure mode classification (radial deformation / axial loosening / shield instability), and the disposal suggestion library index code (e.g., "DZ and 203" correspond to the shim re-tightening process). The work order simultaneously activates the equipment interlocking protection logic chain.

[0210] Specifically, in the method for evaluating the mechanical strength of transformer windings under reclosing conditions described in this invention, the historical operating data is preprocessed in step 1 to form a sample library of reclosing conditions and actual damage, providing real damage labels for model training in step 5.

[0211] The finite element simulation data, through the preprocessing in step 1, clarifies the high-stress region and the theoretical deformation threshold, providing a theoretical basis for the sensor deployment in step 2.

[0212] The laboratory simulation test data is preprocessed in step 1 to verify the consistency between the simulation data and the actual operating data.

[0213] The preprocessing and functional implementation of historical operational data include:

[0214] Historical operational data is derived from reclosing event archives and corresponding transformer dismantling and inspection reports recorded by the power grid dispatching system for more than five years. The preprocessing stage involves key information extraction: extracting operating parameters such as impact current amplitude and reclosing phase angle from the event archives; and identifying physical damage indicators such as winding radial permanent deformation and pad displacement distance from the dismantling reports. A mapping table between "operating parameters and damage indicators" is established to create a sample library labeled with real damage data. This sample library provides the training labels needed for supervised learning of the mechanical strength assessment model in step 5, enabling the model to learn the evolution of mechanical damage in actual operating scenarios. The sample library is stored in a relational database, supporting sample retrieval and retrieval by voltage level and winding type.

[0215] The processing and application paths of finite element simulation data include:

[0216] Finite element simulation data is generated through multiphysics coupling modeling: a three-dimensional solid model of the transformer, including the core, windings, and insulating pads, is established, and electromagnetic, structural, and thermal multi-field coupling boundary conditions are set. Under typical reclosing conditions (such as a 90° closing phase angle), the electromagnetic force distribution cloud map of the windings is solved, outputting the radial stress peak coordinate set, axial vibration modal frequency spectrum, and pad contact pressure distribution threshold. The preprocessing stage extracts the coordinates of areas where stress exceeds 85% of the material's allowable stress, sensitive sections of vibration frequency deviation, and pad locations where pressure is 20% lower than the design value. These parameters provide spatial positioning basis for sensor deployment in step 2, guiding the optimized placement strategy of miniature foil strain gauges and flexible piezoresistive sensors.

[0217] The verification logic of the laboratory simulation experiment includes:

[0218] The laboratory simulation test used a scaled-down model of a 220kV transformer and reproduced the reclosing transient process using a controllable impulse current generator. During the preprocessing stage, data consistency verification was performed: the deformation data (radial strain waveform, axial vibration spectrum) collected from the experiment were compared with the finite element simulation results in the time-frequency domain to calculate the error rate of key indicators (e.g., radial stress peak error ≤ 8%); simultaneously, the physical damage measured after the experiment (e.g., the location of the wire disc deformation) was compared with the damage patterns in the historical operation sample library to verify the spatial distribution matching degree (location overlap ≥ 90%). This verification process confirmed the physical consistency between the simulation data and the actual operating data, ensuring the engineering applicability of the sensor deployment strategy in step 2 and the correlation analysis model in step 4.

[0219] Specifically, in the method for evaluating the mechanical strength of transformer windings under reclosing conditions described in this invention, the FPGA synchronization controller achieves hardware clock synchronization through PTP.

[0220] By combining software timestamp correction, time offsets caused by differences in hardware characteristics of different sensors are eliminated;

[0221] Through the synergy of hardware synchronization and software correction, the time alignment in step 3 is used to accurately match the timestamps of historical reclosing events, thereby avoiding biases in correlation analysis caused by time misalignment.

[0222] Hardware clock synchronization mechanisms include:

[0223] The FPGA synchronization controller integrates an IEEE 1588 Precision Time Protocol (PTP) module, receiving the time synchronization signal from the master clock source via a fiber optic network. Internally, the controller deploys hardware clock synchronization logic: within each PTP synchronization cycle, it calculates the path transmission delay and clock offset between the slave and master clocks, dynamically adjusting the local clock counter. This process achieves microsecond-level precision time base unification at the physical layer, providing a consistent hardware time reference source for all connected sensor data channels. The synchronization signal is transmitted to each data acquisition card via a dedicated clock distribution circuit, eliminating time jitter caused by traditional network communication protocols.

[0224] Software timestamp correction methods include:

[0225] To address the inherent delay differences in signal processing links from different sensors, software correction logic is implemented at the FPGA firmware layer. Specific implementation includes:

[0226] During the system initialization phase, standard step test signals are injected into each sensor channel;

[0227] Measure the actual response delay time from signal input to the FPGA acquisition buffer;

[0228] The delay value is dynamically compensated during the timestamp marking process;

[0229] For example, the signal from a miniature foil strain gauge experiences a fixed delay due to the bridge balancing circuit. The correction logic subtracts this delay from its timestamp. This compensation mechanism eliminates timing deviations caused by sensor hardware characteristics, ensuring precise alignment of the transient responses of different physical quantities in the time dimension.

[0230] The specific implementation process of this invention is based on the mechanical strength testing requirements of power transformers under reclosing conditions, especially addressing the technical bottleneck of synchronous acquisition and synergistic impact quantitative analysis of multi-directional deformation data. The implementation first establishes a multi-source data foundation: extracting reclosing event records from the power grid dispatching system for more than five years and corresponding transformer disassembly and inspection reports to form a historical operation database; simulating winding stress distribution under different reclosing phase angles using an electromagnetic, structural, and thermal multiphysics coupling model to generate radial stress peak coordinates, axial vibration sensitive frequency bands, and pad pressure thresholds; conducting laboratory simulation tests using a 220kV transformer scaled-down model and a controllable impulse current generator to record deformation data and verify actual damage locations through disassembly. After deduplication, timestamp unification, and dimensional standardization preprocessing of the three types of data, the following outputs are provided: a high-stress area coordinate set (to guide sensor placement), a damage-prone location mapping table (associating with historical fault points), and a damage sample feature library (including operating parameters and physical damage labels).

[0231] During the sensor deployment phase, precise placement is executed based on the preprocessing results: On the surface of the wire disc where the radial stress, as determined by finite element simulation, exceeds 85% of the material's allowable stress, micro-foil strain gauges are bonded to monitor transient radial strain values; MEMS accelerometers are installed in the gaps between pads where the axial vibration frequency deviation is ≥5% in historical dismantling reports to capture triaxial acceleration signals; flexible piezoresistive sensors are embedded in areas where the pad contact pressure, as shown in simulations, is 20% lower than the design value to measure pressure fluctuations. All sensor signals are connected to an FPGA synchronization controller, which achieves hardware clock synchronization via the IEEE 1588 PTP protocol and injects sensor response delay compensation values ​​(e.g., strain gauge bridge delay of 0.2ms), outputting a synchronized data stream with microsecond-level corrected timestamps.

[0232] The signal processing flow includes three levels of operations: applying a 20ms sliding mean filter to the raw data to suppress high-frequency noise caused by magnetostriction of the iron core; aligning the three signals in 1ms window slices based on the FPGA timestamp and matching the historical event time base through linear interpolation; extracting the radial maximum deformation, deformation rate of change, axial dominant frequency (FFT analysis of the 0 and 200Hz bands), vibration amplitude, pressure fluctuation range of the pad block, and pressure drop rate (least square fitting slope) within the alignment window. The feature extraction results are input into the correlation analysis module: calculating the correlation between the radial deformation rate of change and the axial vibration amplitude using the Pearson correlation coefficient, and outputting the radial and axial coupling coefficients; using partial least squares regression analysis to analyze the correlation between the axial dominant frequency and the pressure fluctuation range of the pad block, and outputting the axial and pad block correlation coefficients; constructing a three-dimensional covariance matrix of radial deformation, axial dominant frequency offset, and pad block pressure drop rate, and identifying the dominant degradation factors (such as the dimension weight corresponding to the largest eigenvalue > 0.7) through eigenvalue decomposition.

[0233] The mechanical strength assessment model is trained using a random forest architecture: the input layer receives the aforementioned five basic features and two collaborative parameters; using Level I and Level III labels from the historical damage sample library as supervisory data, the tree depth (typical values ​​10 and 15) and minimum number of split samples are optimized through grid search; validation is performed using 20% ​​of the samples not involved in training from laboratory simulation experiments, and calibration is completed when the model output level matches the actual disassembly damage level by ≥90%. During online assessment, real-time feature parameters are input to calibrate the model output degradation level: Level I (all parameters within the safety threshold) triggers regular data recording; Level II (single parameter exceeding the limit) generates an early warning with location information and increases the local sampling rate to 1kHz; Level III (radial deformation > 2 times the yield threshold / axial frequency offset > 10% / pad pressure drop rate > 1.0N / s) pushes a maintenance work order to the operation and maintenance system, the work order including the damage location topology code (e.g., "low-voltage winding L3 phase 5th wire cake") and the treatment process index code.

[0234] This implementation process comprehensively addresses the lack of assessment of the multi-directional deformation synergy impact pointed out by the background technology through multi-source data collaboration to guide optimal sensor deployment, hardware synchronization and software calibration to ensure data timing accuracy, multi-dimensional feature correlation analysis to reveal deformation coupling mechanisms, and machine learning models to achieve quantitative assessment of cumulative damage. The technical parameters and limiting features at each stage provide corresponding support; for example, the sensor deployment basis corresponds to the stress peak coordinates output by finite element simulation, the correlation analysis is achieved through Pearson correlation coefficient calculation, and the random forest model architecture and three-level degradation level output echo the core assessment logic.

[0235] This invention systematically solves the problem of synchronous acquisition and correlation analysis of multi-directional deformation data under reclosing impact by employing a multi-source data fusion and multi-dimensional analysis framework. First, step 1 acquires and preprocesses three types of basic data: historical operational data provides a sample library of "reclosing conditions and actual damage" (including event records and dismantling reports from over 5 years); finite element simulation data clarifies high-stress areas of the winding (such as the middle section of the low-voltage winding with the highest radial stress) and theoretical deformation thresholds; and laboratory simulation test data verifies the consistency between simulation and reality (such as scaled-down model disassembly verifying the deformation position matching rate). After deduplication and format unification, these three types of data respectively form high-stress area data (simulation results), vulnerable location data (historical records), and real damage labels for model training (historical samples), providing multi-dimensional support for subsequent data acquisition and analysis.

[0236] Secondly, steps 2 and 3 are used to achieve synchronous acquisition and feature extraction of multi-directional deformation data. Based on the high-stress region data preprocessed in step 1, micro foil strain gauges (monitoring radial tensile / compressive strain values), MEMS accelerometers (monitoring axial vibration acceleration amplitude and frequency), and flexible piezoresistive sensors (monitoring pad pressure changes) are deployed in key areas of the winding. Microsecond-level timestamp alignment (hardware clock synchronization + software correction) is achieved through an FPGA synchronous controller, enabling radial, axial, and pad data to be acquired under the same time reference. The acquired raw data undergoes sliding window mean filtering (filtering out high-frequency interference), 1ms time window alignment (matching historical event timestamps), and feature extraction (such as maximum radial deformation, axial dominant frequency offset, and pad pressure drop rate) in step 3 to form quantitative parameters reflecting multi-directional deformation characteristics, providing input for correlation analysis.

[0237] Finally, steps 4 and 5 are used to perform correlation analysis of multi-directional deformation and overall mechanical strength assessment. Step 4 uses the extracted multi-dimensional deformation features as input, calculates the correlation between radial and axial deformation using Pearson correlation coefficient (output coupling coefficient), analyzes the correlation between axial vibration and pad loosening using partial least squares regression (output correlation coefficient), and quantifies the synergistic effect using multivariate covariance matrix (identifying dominant degradation factors), transforming independent deformation data into synergistic effect parameters. Step 5 uses these parameters and historical damage samples as a training set, employing random forest or LSTM algorithms to train a mechanical strength assessment model, outputting degradation levels (Level I normal, Level II slight degradation, Level III severe degradation). After the model is embedded in the online monitoring system, it receives sensor data in real time, preprocesses and performs correlation analysis before inputting it into the model, outputting the real-time winding mechanical strength status, thus achieving an accurate assessment of the overall mechanical strength change caused by cumulative deformation.

Claims

1. A method for evaluating the mechanical strength of transformer windings under reclosing conditions, characterized in that, include: Step 1: Acquire basic data including historical operation data, finite element simulation data and laboratory simulation test data. After deduplication and format unification preprocessing, high stress area data and easily damaged location data are generated to determine the sensor deployment location, as well as historical damage samples and simulation and test verification data for model training. Step 2: Based on the high-stress area data and easily damaged location data obtained from preprocessing, micro foil strain gauges, MEMS accelerometers and flexible piezoresistive sensors are deployed in the winding area. The original time series data of radial strain, axial acceleration and pad pressure are synchronously acquired through FPGA synchronous controller. Step 3: Perform sliding window mean filtering, 1ms time window alignment based on synchronization timestamp, and feature extraction on the collected raw time series data in sequence. Step 4: Based on the multi-dimensional deformation characteristics of the radial maximum deformation, axial dominant frequency, and pad pressure fluctuation range extracted by feature extraction, the correlation between radial deformation and axial vibration amplitude is analyzed by Pearson correlation coefficient, and the correlation between axial dominant frequency and pad pressure fluctuation is analyzed by partial least squares regression to identify parameters that synergistically affect deformation in multiple directions. Step 5: Input the identified multi-directional deformation synergistic influence parameters and actual damage samples from historical operating data into the training model to train an evaluation model that outputs the mechanical strength degradation level. Based on the real-time collected deformation characteristics and synergistic influence parameters, the evaluation model outputs the real-time winding mechanical strength evaluation result. Sensors include miniature foil strain gauges, MEMS accelerometers, and flexible piezoresistive sensors; The micro foil strain gauges are deployed in the region of maximum radial stress determined by the finite element simulation data in step 1, and are used to monitor radial tensile or compressive strain values ​​in real time. The MEMS accelerometer is deployed in the gap of the disc pad where the axial vibration frequency shift is significant, as recorded in the historical dismantling detection report in step 1, to monitor the acceleration amplitude and frequency of axial vibration. The flexible piezoresistive sensor is deployed in the region with the lowest contact pressure threshold of the pad determined by the finite element simulation data in step 1, and indirectly reflects the degree of loosening of the pad through pressure changes. The sensor achieves microsecond-level timestamp alignment through an FPGA synchronization controller, enabling the acquisition of radial strain, axial acceleration, and pad pressure data under the same time reference. Also includes: The correlation between the radial deformation rate of change and the axial vibration amplitude is calculated using the Pearson correlation coefficient, and the radial and axial coupling coefficients are output. Partial least squares regression analysis was used to analyze the correlation between the pressure fluctuation range of the pad block and the axial dominant frequency, and the correlation coefficient between the axial direction and the pad block was output. By establishing a multivariate covariance matrix, the covariances of radial, axial and pad deformation characteristics are calculated to identify the dominant degradation factors; The coupling coefficient, correlation coefficient, and dominant degradation factor output by the correlation analysis are used as input parameters for model training in step 5.

2. The method for evaluating the mechanical strength of transformer windings under reclosing conditions as described in claim 1, characterized in that, Also includes: The historical operating data consists of records of reclosing events of transformers over 5 years and corresponding disassembly and inspection reports, which are used to provide a database of reclosing conditions and actual damage samples required for model training. The finite element simulation data simulates the distribution of winding electromagnetic force under different reclosing conditions through a multi-physics coupling model, and outputs the location of the maximum radial stress, axial vibration frequency and pad contact pressure threshold to identify the high-stress area of ​​sensor deployment. The laboratory simulation test data uses a 220kV transformer scaled-down model and a controllable impulse current generator to simulate reclosing conditions, records winding deformation data, and disassembles to verify actual damage, in order to verify the consistency between simulation data and actual operating data.

3. The method for evaluating the mechanical strength of transformer windings under reclosing conditions as described in claim 2, characterized in that: Step 1 includes: Noise filtering employs sliding window mean filtering to remove high-frequency electromagnetic interference noise while retaining low-frequency deformation characteristics related to reclosing impact. Time alignment is based on the synchronization timestamp generated by the FPGA synchronization controller in step 2. The radial strain, axial acceleration and pad pressure data are divided into time windows in 1ms units and matched with the reclosing event timestamps of the historical running data in step 1 to form a time-aligned multi-dimensional dataset. Feature extraction includes: Extract the maximum deformation and the rate of change of deformation within each time window; Perform FFT transformation on the acceleration signal to extract the dominant frequency and vibration amplitude; Extract the fluctuation range and pressure drop rate of the pressure signal; The extracted multidimensional deformation features are used as input parameters for the correlation analysis in step 4.

4. The method for evaluating the mechanical strength of transformer windings under reclosing conditions as described in claim 3, characterized in that: The input parameters of the mechanical strength assessment model include: the extracted radial maximum deformation, the axial principal frequency offset, and the pressure drop rate of the pad block; The multi-directional deformation coupling strength output by the correlation analysis includes radial and axial coupling coefficients, and axial and pad correlation coefficients. The mechanical strength assessment model outputs parameters representing the mechanical strength degradation level, where: First mechanical strength degradation level: maximum radial deformation ≤ preset threshold, axial dominant frequency offset ≤ 5%, pad pressure drop rate ≤ 0.5 N / s and multi-directional deformation coupling strength ≤ 0.3; Second mechanical strength degradation level: Any parameter exceeds the threshold of the first mechanical strength degradation level but does not meet the standard of the third mechanical strength degradation level; Third mechanical strength degradation level: maximum radial deformation > 2 times the preset threshold or axial main vibration frequency offset > 10% or pad pressure drop rate > 1.0 N / s; The mechanical strength assessment model is trained using a random forest algorithm or a long short-term memory network. It uses actual damage samples from historical operating data as labels and optimizes tree depth or time step parameters through cross-validation to capture the nonlinear relationship between multi-dimensional features and degradation level.

5. The method for evaluating the mechanical strength of transformer windings under reclosing conditions as described in claim 4, characterized in that: The mechanical strength assessment model was validated using 20% ​​of the laboratory simulation test data that were not used for training. The degradation level output by the mechanical strength assessment model was compared with the actual disassembly damage level. When the matching degree is ≥90%, the matching degree is the percentage of samples whose output level of the mechanical strength assessment model matches the actual level, and the model calibration is completed. The calibrated mechanical strength assessment model is used for real-time assessment in step 5.

6. The method for evaluating the mechanical strength of transformer windings under reclosing conditions as described in claim 5, characterized in that, Real-time assessment includes: Receive real-time sensor data synchronously acquired by the FPGA synchronous controller in step 2; After noise filtering, time alignment and feature extraction in step 3, the real-time data is input into the correlation analysis in step 4. Input the preprocessed and correlation-analyzed real-time feature parameters into the mechanical strength assessment model trained in step 5, and output the real-time mechanical strength degradation level. Real-time mechanical strength degradation levels include: First mechanical strength degradation level: Real-time data and evaluation results are recorded in the database to maintain the regular inspection cycle; Second mechanical strength degradation level: Generate early warning information to prompt a shortening of the local monitoring cycle; Third mechanical strength degradation level: Generate maintenance suggestions for pushing to the operation and maintenance system.

7. The method for evaluating the mechanical strength of transformer windings under reclosing conditions as described in claim 6, characterized in that, Also includes: history The operational data is preprocessed in step 1 to form a database of reclosing conditions and actual damage samples, providing real damage labels for model training in step 5. The finite element simulation data, through preprocessing in step 1, clarifies the high-stress region and the theoretical deformation threshold, providing a theoretical basis for sensor deployment in step 2. The laboratory simulation test data is preprocessed in step 1 to verify the consistency between the simulation data and the actual operation data.

8. The method for evaluating the mechanical strength of transformer windings under reclosing conditions as described in claim 7, characterized in that, Also includes: The FPGA synchronization controller achieves hardware clock synchronization via PTP; By combining software timestamp correction, time offsets caused by differences in hardware characteristics of different sensors are eliminated; Through the synergy of hardware synchronization and software correction, the time alignment in step 3 is used to accurately match the timestamps of historical reclosing events, thereby avoiding biases in correlation analysis caused by time misalignment.

Images