A non-contact multi-point vibration monitoring method and related equipment for substation transformers based on millimeter-wave radar
By adopting a non-contact multi-point vibration monitoring method based on millimeter-wave radar, the installation complexity and synchronous measurement challenges of substation transformer vibration monitoring have been solved. This method enables synchronous measurement and condition assessment of vibration in multiple areas of the transformer enclosure, reducing system complexity and power consumption. It is applicable to the field of power system equipment condition monitoring.
Patent Information
- Application Number
- CN202611323410.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-28
- Publication Date
- 2026-11-17
AI Technical Summary
In the existing technology, the vibration monitoring methods for substation transformers have problems such as requiring power outage for installation, easy damage to sensors, difficulty in reflecting the distribution characteristics of the enclosure with single-point measurement, and high complexity of multi-point deployment. In addition, non-contact measurement equipment is expensive or cannot simultaneously measure multi-point vibration.
A non-contact multi-point vibration monitoring method based on millimeter-wave radar is adopted. By transmitting frequency-modulated continuous wave signals, a range-dimensional fast Fourier transform is performed to decompose the vibration displacement sequence of vibration response points. Frequency compensation is performed using an exponential correction model, and beamforming is combined with near-field steering vectors to achieve synchronous measurement and condition assessment of multi-point vibration.
It enables non-contact synchronous measurement of multiple areas on the transformer tank surface, improves the correlation between vibration characteristics and operating conditions, reduces system complexity and power consumption, and meets the low power consumption and carbon footprint requirements of green substations.
Smart Images

Figure CN123111446A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of power system equipment condition monitoring, and relates to a non-contact multi-point vibration monitoring method and related equipment for substation transformers based on millimeter-wave radar. Background Technology
[0002] Power transformers are core equipment in substations, and reliable monitoring of their operating status is crucial for power grid safety. Vibration analysis is one of the important methods for transformer condition assessment. Traditional transformer vibration monitoring mainly uses contact sensors such as accelerometers, which are directly attached to the surface of the transformer tank. This method has the following problems: (1) Power outage is required during installation, and the sensor is exposed to strong electromagnetic fields, high temperature and oily environment for a long time, which can easily cause insulation deterioration, signal drift or even detachment; (2) Single-point measurement is difficult to reflect the vibration distribution characteristics of different areas of the transformer box, while multi-point deployment of sensors greatly increases the system complexity and maintenance cost.
[0003] In recent years, non-contact vibration measurement technology has attracted attention. Laser Doppler vibration meters (LDV) can achieve high-precision non-contact measurements, but their equipment is expensive, optical path alignment is difficult, and they can only scan a single point sequentially, making it impossible to simultaneously measure multi-point vibrations.
[0004] Vibration measurement technology based on millimeter-wave radar has been gradually introduced into the field of structural health monitoring due to its advantages such as good penetration, immunity to light, and ability to monitor multiple targets simultaneously.
[0005] For example, CN110161494A discloses a method for multi-point vibration monitoring of bridges based on FMCW radar, which separates vibration points at different radial distances through range resolution. However, this method can only distinguish vibration sources with radial distance differences greater than the radar range resolution, and cannot separate multiple sources located at the same distance layer (radial distance difference less than the resolution) but with different angles. To solve this problem, CN112630769A proposes attaching multiple corner reflectors to the surface of the transformer tank and using the radar's angle resolution capability to distinguish the vibrations corresponding to different reflectors. However, this method requires additional auxiliary devices, increasing the on-site installation workload, and the corner reflectors may introduce interference in complex electromagnetic environments. Summary of the Invention
[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a non-contact multi-point vibration monitoring method and related equipment for substation transformers based on millimeter-wave radar. This method can realize non-contact synchronous measurement of vibration in multiple areas on the surface of the power transformer enclosure and construct vibration characteristics that can stably reflect the operating conditions.
[0007] To achieve the above objectives, the present invention employs the following technical solution: A non-contact multi-point vibration monitoring method for substation transformers based on millimeter-wave radar includes: The complex intermediate frequency signal is obtained by transmitting a frequency-modulated continuous wave signal to a transformer via millimeter-wave radar and receiving the echo. A range-dimensional fast Fourier transform is performed on the complex intermediate frequency signal to obtain a one-dimensional range image and the complex echo sequence corresponding to each range gate; Based on the number of local energy peaks and full width at half maximum (FWHM) in a one-dimensional range image, vibration displacement sequences of at least two vibration response points are decomposed from the complex echo sequence. Spectral analysis was performed on the vibration displacement sequence at each vibration response point to obtain the amplitude spectrum of each vibration response point. In the amplitude spectrum of each vibration response point, local peaks are searched in a preset neighborhood with the theoretical fundamental frequency as the reference center. The amplitude of the local peaks is extracted as the main frequency amplitude feature, and the actual peak frequency corresponding to the local peaks is extracted. Calculate the frequency offset of the actual peak frequency relative to the theoretical fundamental frequency; Based on the main frequency offset, an exponential correction model is used to compensate for the main frequency amplitude characteristics, and the compensated vibration characteristic values are obtained. The compensated vibration characteristic values are correlated with the synchronously acquired transformer operating reference values to output the transformer operating status assessment results.
[0008] Optionally, based on the number of local energy peaks and full width at half maximum (FWHM) in the one-dimensional range image, a vibration displacement sequence of at least two vibration response points is decomposed from the complex echo sequence, including: When there are multiple independent energy peaks in a one-dimensional range image and the full width at half maximum (FWHM) of each independent energy peak is less than or equal to twice the system range resolution, an incoherent energy accumulation is performed on the one-dimensional range image along the slow time dimension to construct a statistical energy curve. Extract the original complex echo sequences corresponding to the distance gates of each independent energy peak; The original complex echo sequence is unwrapped and differentiated to generate a radially separable vibration displacement sequence for each vibration response point; When a single broadening peak exists in a one-dimensional range image and the full width at half maximum (FWHM) of the single broadening peak is greater than twice the system range resolution, a near-field steering vector is constructed based on the multi-channel original complex echo signal within the target range layer where the single broadening peak is located. Using the near-field steering vector, Bartlett beamforming spatial spectrum focusing is performed to search for multiple local peaks in the spatial spectrum and separate vibration response points at different angles in the same layer aliasing scene; The complex focusing signal of each vibration response point is extracted and subjected to phase demodulation and displacement conversion to generate the vibration displacement sequence of each vibration response point.
[0009] Optionally, Bartlett beamforming spatial spectrum focusing can be performed using a near-field steering vector, including: The search range of candidate points is limited to the spatial surface determined by the radial distance value of the target distance layer; The radial distance coordinates of each candidate point are fixed as the radial distance values corresponding to the target distance layer; The azimuth and elevation angles of each candidate point are traversed and searched to calculate the Bartlett spatial spectrum of the corresponding location.
[0010] Optionally, after constructing a statistical energy curve by incoherent energy accumulation of the one-dimensional distance image along the slow time dimension, the following steps are also included: Extract multiple local peak points from the statistical energy curve; Multiple local peak points are filtered by applying minimum gate spacing constraints in spatial order to determine multiple target distance gates that are independent of each other in the radial direction. The minimum gate spacing constraint threshold is 2 to 5 times the system distance resolution. The process of extracting the original complex echo sequence of each independent energy peak corresponding to the range gate is performed based on multiple target range gates.
[0011] Optionally, an exponential correction model is used to compensate for the main frequency amplitude characteristics, including the following calculation formula:
[0012] in, F The compensated vibration characteristic value, A Main frequency amplitude characteristics, This is the main frequency offset. c For compensation coefficient, These are the first-order frequency compensation coefficients. These are second-order frequency compensation coefficients, and c , and All system parameters were determined through offline joint optimization and calibration.
[0013] Optional, c , and The offline joint optimization calibration process includes: Multiple sets of vibration signals were collected under various load rate conditions of the transformer. Obtain the main frequency amplitude characteristics, main frequency offset, and load rate per unit value under various load rate conditions; Introducing a normalization factor Build Minimal The objective function is , where The characteristic of the main frequency amplitude under the i-th operating condition is... Let be the main frequency offset under the i-th operating condition. Let i be the per-unit value of the load rate under the i-th operating condition; Simultaneously solve using numerical optimization algorithms , c , and The optimal value; Discard the optimal value The optimal values of the remaining parameters are solidified into the compensation coefficients, first-order frequency compensation coefficients, and second-order frequency compensation coefficients of the exponential correction model.
[0014] Optionally, before transmitting the frequency-modulated continuous wave signal through the millimeter-wave radar front end, the following steps are also included: The millimeter-wave radar front end is activated according to the preset monitoring schedule to perform data acquisition operations on the transmitted signals; During non-acquisition periods after data acquisition operations are completed, the millimeter-wave radar front end is placed in a low-power sleep mode; Record the cumulative runtime and power consumption data of the monitoring process, and calculate the cumulative carbon emissions based on the power grid carbon emission factor and power consumption data.
[0015] A non-contact multi-point vibration monitoring system for substation transformers based on millimeter-wave radar, including a millimeter-wave radar front-end module; The communication end of the millimeter-wave radar front-end module is connected to a data acquisition and preprocessing module. The output end of the data acquisition and preprocessing module is connected to a multi-point vibration separation module. The output end of the multi-point vibration separation module is connected to a vibration feature extraction and state assessment module. The power supply terminals of the millimeter-wave radar front-end module, the data acquisition and preprocessing module, the multi-point vibration separation module, and the vibration feature extraction and state assessment module are all connected to communication and power supply modules. The control terminal of the millimeter-wave radar front-end module is connected to a green management module; The millimeter-wave radar front-end module includes virtual array elements arranged in a two-dimensional area. The millimeter-wave radar front-end module is located outside the transformer box and has an insulating distance gap with the surface of the transformer box. The main beam direction of the millimeter-wave radar front-end module forms an angle of less than or equal to 8° with the normal direction of the transformer box surface.
[0016] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the non-contact multi-point vibration monitoring method for substation transformers based on millimeter-wave radar.
[0017] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the non-contact multi-point vibration monitoring method for substation transformers based on millimeter-wave radar.
[0018] Compared with the prior art, the present invention has the following beneficial effects: This invention achieves non-contact synchronous measurement of multiple vibration response points by a single radar by transmitting a frequency-modulated continuous wave signal to the transformer and processing the complex intermediate frequency signal to obtain a one-dimensional range profile. Based on the number of energy peaks and the full width at half maximum (FWHM), the vibration displacement sequence is decomposed. After acquiring the amplitude spectrum at each point, the main frequency offset of the actual peak frequency relative to the theoretical fundamental frequency is calculated, and an exponential correction model is used to compensate for the main frequency amplitude characteristics. This compensation mechanism eliminates the influence of frequency offset on vibration amplitude fluctuations, outputting corrected vibration characteristic values and improving the correlation between the transformer operating status assessment results and actual operating conditions.
[0019] Furthermore, the relationship between the full width at half maximum (FWHM) of the energy peak and the system range resolution is used as a criterion to distinguish between radially separable scenarios and intra-layer aliasing scenarios. For radially separable scenarios, the original complex echo sequence is extracted and demodulated by incoherent energy accumulation along the slow time dimension. For single broadened peaks with a FWHM greater than twice the system range resolution, a near-field steering vector is introduced for Bartlett beamforming spatial spectrum focusing. This process utilizes the difference in near-field spherical wavefront curvature to generate a detectable phase difference, separating multiple vibration response points located in the same range layer but with different spatial angles, thus solving the problem of indistinguishable aliasing of multiple vibration sources within a single range cell.
[0020] Furthermore, when using the near-field steering vector for Bartlett beamforming spatial spectrum focusing, the radial distance coordinates of each candidate point are fixed based on the prior radial distance value of the target distance layer. This reduces the dimensionality of the three-dimensional search to the spatial surface determined by the radial distance value. By performing a search only on the azimuth and elevation angles, the number of grid points for spatial spectrum calculation is controlled while preserving the near-field path difference information to distinguish response points at different angles, thus reducing the computational load of the beamforming algorithm.
[0021] Furthermore, after constructing the statistical energy curve and extracting local peak points, a minimum gate interval constraint with a minimum interval threshold of 2 to 5 times the system distance resolution is applied in spatial order. Based on the distance resolution characteristics, this constraint filters out false neighboring peaks caused by the sidelobe effect of windowing and selects target distance gates that are independent of each other in the radial direction. This prevents errors in vibration displacement sequence extraction due to misjudgment of distance units and ensures the accuracy of multi-point vibration separation.
[0022] Furthermore, an exponential correction model incorporating first-order and second-order frequency compensation coefficients is used to compensate for the dominant frequency amplitude characteristics. The first-order and second-order compensation coefficients quantify the first and second nonlinear effects of frequency offset on amplitude attenuation, respectively. Substituting the dominant frequency offset into this exponential formula, the measured dominant frequency amplitude characteristics are restored to the reference characteristic value under the state of no frequency offset, thus eliminating the interference of power grid fundamental frequency fluctuations on vibration amplitude measurement.
[0023] Furthermore, compensation coefficients, first-order frequency compensation coefficients, and second-order frequency compensation coefficients are determined through offline joint optimization calibration. In the calibration process, which includes multiple load factor conditions, a normalization factor is introduced to construct the objective function, and the optimal values of each coefficient are simultaneously solved using a numerical optimization algorithm. This process enables the model parameters to be automatically tuned based on measured data of the transformer equipment under different operating conditions. After calibration, the normalization factor is discarded, and only the compensation coefficients are fixed for online operation, establishing a clear mapping relationship between the output vibration characteristic values and the transformer load factor.
[0024] Furthermore, the millimeter-wave radar front-end is activated to perform data acquisition according to a preset monitoring schedule, and placed into a low-power sleep mode during non-acquisition periods. This on-demand scheduling logic controls the working time of the radar front-end and reduces the average operating power consumption of the monitoring system throughout its entire lifecycle. Simultaneously, by recording operating time and power consumption data and combining this with the grid carbon emission factor to calculate cumulative carbon emissions, quantitative monitoring data on auxiliary equipment energy consumption and carbon emissions are output. Attached Figure Description
[0025] Figure 1 This is a block diagram of the overall structure of the non-contact multi-point vibration monitoring system for substation transformers based on millimeter-wave radar according to the present invention. Figure 2 This is a schematic diagram of the deployment of the monitoring system of the present invention at a substation site; Figure 3 This is a schematic diagram of the distance spectrum and multi-point extraction of a radially separable scene according to the present invention; Figure 4(a) is a schematic diagram of the distance constraint candidate point search surface in the same layer aliasing scene of the present invention; Figure 4 () shows the spatial spectrum and multi-point separation effect of the same-layer cascade scene of the present invention; Figure 5 This is a flowchart of the adaptive feature construction and state evaluation process of the present invention; Figure 6 This is a flowchart of the offline calibration (joint optimization) of the parameters of the exponential compensation model of the present invention. Detailed Implementation
[0026] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.
[0027] The following disclosure provides many different embodiments or examples for implementing various structures of the invention. To simplify the disclosure, specific examples of components and arrangements are described below. These are merely examples and are not intended to limit the invention. Furthermore, reference numerals and / or letters may be repeated in different examples; such repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed. In addition, examples of various specific processes and materials are provided in this invention, but those skilled in the art will recognize the application of other processes and / or the use of other materials.
[0028] Example 1 This embodiment provides a non-contact multi-point vibration monitoring system for substation transformers based on millimeter-wave radar, such as... Figure 1 As shown, it includes a millimeter-wave radar front-end module 100, a data acquisition and preprocessing module 200, a multi-point vibration separation module 300, a vibration feature extraction and state assessment module 400, a communication and power supply module 500, and a green management module 600.
[0029] The millimeter-wave radar front-end module 100, integrated within a single housing, is installed outside a transformer enclosure, maintaining a safe distance from the enclosure, with the angle between the main beam direction and the normal direction of the enclosure surface not exceeding 8°, wherein the normal direction refers to the direction of the normal to the enclosure surface at the center point of the measurement area; the millimeter-wave radar front-end module 100 is used to transmit frequency-modulated continuous wave signals to the target measurement area and receive echoes; the millimeter-wave radar front-end module 100 includes multiple virtual array elements arranged in a two-dimensional area array.
[0030] The data acquisition and preprocessing module 200 is communicatively connected to the millimeter-wave radar front-end module 100. It is used to perform analog-to-digital conversion and range-dimensional fast Fourier transform on the complex intermediate frequency signal to obtain a one-dimensional range image and complex echo sequences corresponding to each range gate.
[0031] A multi-point vibration separation module 300, communicatively connected to the data acquisition and preprocessing module 200, is used to decompose at least two vibration displacement sequences of vibration response points that are distinguishable in the radial distance dimension or in the angular dimension within the same radial distance layer from the complex echo sequences based on the one-dimensional range image and the complex echo sequences of each range gate. The multi-point vibration separation module 300 includes a radially separable scene processing unit, a same-layer aliased scene processing unit, and an adaptive switching unit for automatically selecting the radially separable scene processing unit or the same-layer aliased scene processing unit. The same-layer aliasing scene processing unit is configured to: perform near-field Bartlett beamforming based on the multi-channel original complex echo signals within the target range layer. Each component of the guide vector of the near-field Bartlett beamforming is calculated based on the true Euclidean distance from the candidate point to each array element. The search space of the candidate points is limited to a spatial surface determined by the radial distance value corresponding to the target range layer. The radial distance coordinates of each candidate point on the spatial surface are fixed to the radial distance value corresponding to the target range layer, and the search is performed only on the two-dimensional coordinates of azimuth and pitch angles.
[0032] The radially separable scene processing unit is used to perform statistical energy accumulation on the one-dimensional range image along the slow time dimension, extract multiple local peak points from the obtained statistical energy curve, and apply minimum gate spacing constraints in spatial order for filtering to obtain multiple target range gates that are independent of each other in the radial direction. The original complex echo sequence corresponding to each target range gate is extracted and obtained by phase demodulation and displacement conversion to obtain the vibration displacement sequence of each radial response point. The minimum gate spacing constraint minimum spacing threshold is 2 to 5 times the range resolution.
[0033] The adaptive switching unit is configured to select the same-layer aliasing scene processing unit when the full width at half maximum (FWHM) of the independent energy peaks in the one-dimensional range image is greater than twice the range resolution; otherwise, the radially separable scene processing unit is selected.
[0034] The same-layer aliasing scene processing unit calculates the Bartlett spatial spectrum according to the following formula:
[0035] In the formula, candidate points The spherical coordinates are represented as Its radial distance coordinates are fixed as the radial distance value corresponding to the target distance layer. azimuth and pitch angle Traverse within the preset search range with a preset step size; This is the near-field steering vector. It is the conjugate transpose operator, its first... i Each component is ,ini = 1,…, Na , Na The effective number of array elements; Candidate points To the i The Euclidean distance of each array element It is the radar operating wavelength. It is the covariance matrix. ,in For the target distance layer The multi-channel raw complex echo vector of a slow-time sampling point Indicates slow time index = 1,…, Nt Calculate the statistical expectation. Nt The number of slow-time sampling frames; the same-layer aliasing scene processing unit separates multiple spatially distinguishable vibration response points by searching multiple local peaks of the Bartlett spatial spectrum in the spatial surface within the same distance layer, extracts the complex focusing signal corresponding to each spatial response point, and obtains the vibration displacement sequence of each spatial response point through phase demodulation and displacement conversion.
[0036] The vibration feature extraction and state assessment module 400 is communicatively connected to the multi-point vibration separation module 300. It is used to perform spectral analysis on each vibration displacement sequence, search for local peaks in a preset neighborhood with the theoretical fundamental frequency as the reference center to obtain the main frequency amplitude feature, and use an exponential correction model determined by offline joint optimization calibration to adaptively compensate the main frequency amplitude feature. Then, it performs correlation analysis with the compensated vibration feature value and the synchronously acquired transformer operation reference value to output the operation state assessment result.
[0037] The exponential correction model is expressed as ,in The measured amplitude at 100Hz (the maximum peak amplitude obtained by searching around 100Hz); These are second-order frequency compensation coefficients, used to capture the quadratic effects of frequency shift; Main frequency offset The , and The normalization factor was determined through offline joint optimization calibration experiments. and , and Together they are used as optimization variables to minimize For the goal, among which This is the per-unit value of the load rate for the i-th calibration condition; after calibration, only the... , and The parameters are fixed into system parameters and stored in non-volatile memory. It is not stored as an online runtime parameter.
[0038] The communication and power supply module 500 is used to realize data transmission and power supply.
[0039] The green management module 600 includes a low-power management unit and a carbon emission monitoring unit. The low-power management unit is used to periodically wake up the millimeter-wave radar front-end module 100 to collect vibration data according to a preset monitoring schedule, and to set at least some modules to a low-power sleep mode during non-collection periods. The carbon emission monitoring unit is used to record the cumulative running time of the system and the power consumption data of each module, calculate the carbon emissions, and transmit them to the substation carbon emission management system.
[0040] This embodiment also discloses a non-contact multi-point vibration assessment method for substation transformers based on millimeter-wave radar, including the following steps: S1. The millimeter-wave radar front end transmits a frequency-modulated continuous wave signal and receives the echo to obtain a complex intermediate frequency signal. A range-dimensional fast Fourier transform is performed to obtain a one-dimensional range image and a complex echo sequence corresponding to each range gate.
[0041] S2. Based on the one-dimensional range image and the complex echo sequences of each range gate, decompose at least two vibration displacement sequences of vibration response points that are distinguishable in the radial distance dimension or in the angular dimension within the same radial distance layer from the complex echo sequences.
[0042] S3. Perform spectral analysis on the vibration displacement sequence of each vibration response point to obtain the amplitude spectrum of each vibration response point.
[0043] S4. In the amplitude spectrum of each vibration response point, local peaks are searched in a preset neighborhood with the theoretical fundamental frequency as the reference center. The amplitude of the local peak is used as the main frequency amplitude feature. The actual peak frequency corresponding to the local peak is extracted. The main frequency offset of the frequency relative to the theoretical fundamental frequency is calculated. Based on the main frequency offset, the main frequency amplitude feature is adaptively compensated using an exponential correction model determined by offline joint optimization calibration to obtain the compensated vibration characteristic value.
[0044] S5. Perform correlation analysis between the compensated vibration characteristic values and the synchronously acquired transformer operation reference values, and output the transformer operation status evaluation results.
[0045] S2 specifically includes: If there are multiple local energy peaks in the one-dimensional range image whose radial distance intervals exceed the range resolution and whose peak shapes are independent, then the radially separable separation mode is adopted: the one-dimensional range image is incoherently accumulated along the slow time dimension, a statistical energy curve is constructed, the original complex echo sequence of each peak corresponding to the range gate is extracted, and the vibration displacement sequence of each radial response point is obtained through phase unwrapping and difference. If a single broadened peak exists in a one-dimensional range image and its full width at half maximum (FWHM) is greater than twice the range resolution, it is determined to be a aliasing scene within the same range layer. Based on the multi-channel original complex echo signal within this range layer, a near-field steering vector is constructed and Bartlett beamforming spatial spectrum focusing is performed. The search range of candidate points is limited to the spatial surface determined by the radial distance value of the target range layer. The radial distance coordinates of the candidate points are fixed to the radial distance value corresponding to this range layer. Only the azimuth and pitch angles are searched. Vibration response points at different angles are separated by searching multiple local peaks in the spatial spectrum.
[0046] The exponential correction model described in S4 is determined by the following offline calibration method: multiple sets of vibration signals are collected under different load rate conditions of the transformer. After each condition is stabilized, the collection time is not less than 0.5 minutes, the sampling rate is not less than 1kHz, and the number of FFT points is not less than 4096. Obtain the main frequency amplitude under various operating conditions frequency offset and per-unit load factor ; Introducing a normalization factor To minimize To achieve the objective, we simultaneously solve the problem through numerical optimization. and , and and the solution obtained , and The parameters are fixed into system parameters and stored in non-volatile memory. It is not stored as an online runtime parameter.
[0047] S5 also includes: periodically waking up the radar to collect data according to a preset monitoring schedule, setting a low-power sleep mode during non-collection periods; recording the system's cumulative runtime and power consumption data of each module, calculating carbon emissions using the configured or default grid carbon emission factor, and uploading the data to the substation carbon emission management system.
[0048] This embodiment achieves non-contact single-radar multi-point synchronous monitoring, requiring only one millimeter-wave radar to simultaneously acquire vibration information from multiple areas on the surface of the enclosure, without contact with live equipment or the need to deploy sensors or reflectors at multiple points. Through the organic combination and adaptive switching of two complementary multi-point separation modes, it achieves, for the first time, completely non-contact synchronous measurement of multiple vibration sources with arbitrary spatial distribution under a single radar architecture.
[0049] This embodiment achieves complete coverage and adaptive switching between two modes: radially separable and in-layer aliasing. Breaking away from conventional far-field beamforming approaches, it creatively enables a near-field model at physical distances that meet far-field conditions. By utilizing prior information in the distance dimension to define the search surface, it transforms minute differences in the curvature of the spherical wavefront into robustly extractable angular information, achieving in-layer multi-source separation. Under conditions of 60 GHz, ≥8 array elements, ≥30,000 slow-time sampling frames, and ≥15 dB signal-to-noise ratio, the measured resolvable angular spacing is better than 5°, significantly better than the traditional far-field Rayleigh limit (approximately 14.3°), effectively overcoming the severe limitations of existing technologies that can only distinguish radially separable sources or require external reflectors.
[0050] The joint optimization calibration method proposed in this embodiment simultaneously solves for the normalization factor and compensation coefficient, avoiding subjective settings and achieving truly individualized parameter tuning for each machine. Experiments show that this method improves the correlation coefficient between vibration characteristics and load rate from 0.62–0.75 in the traditional method to 0.83–0.91, and can be reproduced across different devices.
[0051] This embodiment integrates a green management module 600, introducing for the first time a low-power mechanism based on time-schedule management and carbon emission metering function into a transformer-specific vibration monitoring system. The average power consumption throughout the entire cycle is less than 0.5W, meeting the requirements of green substations for low power consumption and quantifiable carbon footprint of auxiliary equipment. The collaborative design of low-power scheduling and vibration monitoring time-schedule management enables on-demand wake-up and deep sleep energy management, reducing power consumption by approximately 90% compared to conventional continuous operation schemes.
[0052] Example 2 1. Overall system structure.
[0053] like Figure 1 As shown in the figure, the non-contact multi-point vibration monitoring system for substation transformers based on millimeter-wave radar provided in this embodiment mainly consists of six parts: a millimeter-wave radar front-end module 100, a data acquisition and preprocessing module 200, a multi-point vibration separation module 300, a vibration feature extraction and condition assessment module 400, a communication and power supply module 500, and a green management module 600. All modules are integrated into a single housing to form an integrated monitoring terminal, which can be quickly installed on-site.
[0054] In this embodiment, the millimeter-wave radar front-end module 100 uses a 60GHz band FMCW radar with a working bandwidth of 5GHz and a range resolution of 0.03m. The millimeter-wave radar front-end module integrates two transmitting antennas and four receiving antennas, employing time-division multiplexing (TDM) to form eight virtual array elements. Each virtual array element has a two-dimensional L-shaped array layout, with four elements arranged horizontally (element spacing of half a wavelength λ / 2, forming a horizontal aperture of approximately 3λ) and the other four elements arranged vertically (element spacing of half a wavelength λ / 2, forming a vertical aperture of approximately 3λ). The two directions intersect at the array reference point. The array reference point is defined as the origin of a coordinate system established with the geometric center of the antenna array as its origin. This two-dimensional array layout enables the system to have angular resolution in both azimuth and elevation dimensions; key parameters are shown in Table 1. The data acquisition and preprocessing module 200 includes an analog-to-digital converter and an FPGA processor for performing range-dimensional FFT.
[0055] Table 1 Summary of Key Parameters
[0056] Second, on-site deployment methods.
[0057] like Figure 2 As shown, in actual deployment, the millimeter-wave radar front-end module is placed on an insulated tripod and installed outside the transformer enclosure, maintaining a safe insulation distance of 3.5 meters from the enclosure surface. To ensure that the measured vibration displacement is mainly a radial component perpendicular to the enclosure surface, it is necessary to ensure that the angle between the radar main beam direction and the normal direction of the enclosure surface does not exceed 8°. The normal direction refers to the direction of the normal to the enclosure surface at the center point of the measurement area.
[0058] The angle is limited based on the following: the actual vibration displacement vector direction is perpendicular to the box surface, the radial component measured by radar is d·cosθ, and the relative error is (1-cosθ)×100%. When θ=8°, the error is approximately 0.97%, which corresponds to the generally acceptable upper limit of 1% accuracy for transformer vibration measurement in the power industry (see GB / T 1094.10 or IEC 60076-10 regarding the general requirements for permissible measurement errors). Therefore, 8° is taken as the upper limit. In this embodiment, the tripod head is adjusted with the assistance of a laser rangefinder to keep the included angle within 5°.
[0059] Third, radially divisible multi-point extraction.
[0060] like Figure 3 As shown, in a radially separable scenario, the specific implementation of this method is as follows: Step 1: The multi-point vibration separation module 300 first performs a distance-dimensional FFT on the intermediate frequency signal to obtain a one-dimensional distance image. The curve contains three significant energy peaks (P1, P2, P3), with radial positions of 2.77m, 3.81m, and 4.38m respectively. The radial distance difference between adjacent peaks exceeds the system distance resolution (0.03m), and the full width at half maximum (FWHM) of each peak is measured to be 0.04m, 0.04m, and 0.05m respectively, all less than twice the distance resolution (0.06m), satisfying the independent narrow peak condition. Therefore, they can be independently distinguished in the distance dimension.
[0061] Step 2: The multiple points The vibration separation module performs incoherent energy accumulation on the range image along slow time, defining statistical energy:
[0062] Detection conditions based on local peak values:
[0063] Extract candidate peak values and then apply minimum gate spacing constraints. (That is, the minimum interval threshold is twice the distance resolution, corresponding to an actual distance of approximately 0.06m), resulting in three independent target distance gates K1, K2, and K3. The lower limit of the minimum gate interval is set to twice the distance resolution to avoid false proximity peaks caused by windowed sidelobes being misjudged as independent targets.
[0064] Step 3: The adaptive switching unit detects three independent narrow peaks, and the full width at half maximum (FWHM) of each peak does not exceed twice the distance resolution, thus determining it as a radially separable scene. It then automatically selects the radially separable scene processing unit. The phase sequences of the original complex echo sequences corresponding to K1, K2, and K3 are extracted respectively, and phase unwrapping and dynamic phase difference operations are performed to recover the radial vibration displacement of each measuring point. Furthermore, frequency domain features are generated through FFT to obtain the dominant frequency amplitude of each measuring point near 100Hz.
[0065] Fourth, near-field spatial separation in the same-layer cascaded scene.
[0066] As shown in Figures 4(a) and 4(b), in the same-layer aliasing scenario (the radial distance difference between two response points A and B is only 0.02m), the distance-dimensional energy spectrum exhibits a single broadened peak. The full width at half maximum (FWHM) of this broadened peak is measured to be 0.08m, which is more than twice the distance resolution (0.03m), making separation impossible based on the distance dimension. In this case, the adaptive switching unit of the multi-point vibration separation module automatically switches to the same-layer aliasing processing unit according to the judgment conditions.
[0067] Explanation of the peak broadening criterion: In this embodiment, the full width at half maximum (FWHMmix) is greater than twice the range resolution as the aliasing criterion. The theoretical basis for this is as follows: For a single ideal point scattering source, the main lobe width of the response after windowed FFT is approximately equal to the Rayleigh range resolution δR = c / (2B). When the distance between two scattering sources is less than the resolution, the superposition of their echo vectors leads to broadening of the synthesized peak. Under the narrowband approximation, the FWHMmix of the synthesized peak is related to the amplitude ratio, phase difference, and distance between the two sources. Statistical analysis based on numerous Monte Carlo simulations (2000 random phase and amplitude ratios, distance 0–1.5 times the resolution) shows that: When the spacing is ≤0.5 times the resolution, the probability that the synthesized peak FWHM is ≥1.8 times the main lobe width of the single source is greater than 90%; When the spacing is ≤0.3 times the resolution, the probability that FWHM is ≥2.0 times the single-source main lobe width is greater than 95%.
[0068] Based on this, this embodiment uses 2x resolution as the criterion threshold. This threshold can be adjusted within the range of 1.8 to 2.5x according to the actual signal-to-noise ratio and window function type. In this embodiment, using 2x is a preferred value that balances sensitivity and false positive rate.
[0069] The technical principles and feasibility of near-field beamforming: In this embodiment, the millimeter-wave radar front-end adopts an L-shaped array with 8 virtual array elements, operating at a frequency of 60 GHz and a wavelength λ ≈ 0.005 m. The slow-time sampling frame number Nt is 30,000 frames (corresponding to 0.5 minutes of acquisition), and the signal-to-noise ratio of the receiving channel is approximately 18 dB. Classical far-field criterion:
[0070] The given distance is approximately 0.16m, while the actual distance of 3.5m far exceeds this requirement. If the conventional far-field plane wave assumption is used, the steering vectors of all candidate points within the same distance layer are only related to the angle, making multi-source separation within the same layer impossible. This invention counterintuitively employs a near-field spherical wavefront model, fixing the radial distance of candidate points to the target distance layer value, making the steering vector sensitive to the spherical path difference. At 60GHz, two points with an azimuth angle difference of 5° can have a path difference of tens of micrometers to the two end array elements, producing a detectable phase difference of approximately several degrees.
[0071] Analysis of the necessity of cumulative frame count: The signal-to-noise ratio (SNR) of a single-frame signal is limited, and weak phase differences are easily masked by noise. Through slow-time incoherent accumulation (statistical averaging of the Bartlett spectrum), the SNR is improved by approximately:
[0072] When Nt = 30000, the theoretical signal-to-noise ratio is improved by approximately 45 dB, which can stably convert minute phase differences into resolvable peaks in the spatial spectrum. Actual test results are shown in Table 2. Table 2 Actual Test Results
[0073] In this embodiment, 30000 is chosen as the equilibrium point, which ensures both angle separation capability and keeps the single measurement time within 0.5 minutes, thus meeting the real-time requirements of transformer condition monitoring.
[0074] Search step size and computational complexity: The search range for azimuth and elevation angles is ±15° each, with a step size of 0.5°, resulting in approximately 61 × 61 = 3721 grid points. Each Bartlett spectrum calculation involves a weighted summation of Na array elements, with a complexity of approximately... Assuming Nt = 30000, approximately 30000 × 64 × 3721 ≈ 7.15 × 10⁻⁶ 9 Complex number multiplication and addition can be completed in seconds with FPGA or GPU acceleration, meeting the needs of offline analysis or near real-time monitoring.
[0075] Specific implementation steps: Step 1: Construct an ROI distance layer centered on the distance cell where the single broadening peak is located, determine the radial distance value R0 corresponding to the distance layer, and select the original complex echo data of multiple channels within the layer.
[0076] Step 2: Obtain the physical coordinates of each virtual array element Constructing the candidate point search space: candidate points The radial distance coordinates are fixed as Only the azimuth angle and pitch angle Perform a traversal search with a step size of 0.5°. The Cartesian coordinates of candidate points are obtained by transforming spherical coordinates. For any candidate point... Calculate its up to the th The true Euclidean distance of each array element:
[0077] And the radial distance to the array reference point (the origin of the array coordinate system):
[0078] Therefore, the near-field steering vector components are constructed:
[0079] Spatial spectrum calculated based on Bartlett beamforming:
[0080] in For the target distance layer A multi-channel raw complex echo vector with slow time sampling points.
[0081] Step 4: Search for different grid coordinates corresponding to the peak position in the spatial spectrum to obtain two spatially distinguishable response points (point A and point B) within the same radial distance layer. Extract the complex signal at each point to complete the displacement inversion.
[0082] Quantitative explanation of angular resolution: Under the conditions of the 8-element L-shaped array, slow-time sampling frames ≥30,000, and signal-to-noise ratio ≥15dB, the measured resolvable angular interval is approximately 4.8° (verified using two corner reflectors controlled by stepper motors), which is better than the far-field Rayleigh limit of approximately 14.3°. When the angular interval between the two sources is greater than 5°, the separation success rate is higher than 95%.
[0083] 5. Adaptive robust vibration characteristic compensation (including joint parameter optimization calibration).
[0084] Based on the vibration displacements of multiple measurement points obtained from the measured sequence, a robust feature construction method (such as...) is introduced. Figure 5 (As shown). In this embodiment, the parameters c and k of the exponential compensation model, as well as the intermediate variable—the reference eigenvalue normalization factor α—used only in the calibration process, are determined through offline joint optimization calibration experiments.
[0085] like Figure 6 As shown, the process of joint optimization and offline calibration of the exponential compensation model parameters is as follows: The exponential correction model is as follows:
[0086] Under a 50Hz power grid, the theoretical fundamental frequency is 100Hz. The maximum peak amplitude found within the range [95Hz, 105Hz] .
[0087] The core of this calibration method lies in: instead of pre-setting the normalization factor α, it is compared with... , and Solve together. Construct the objective function during calibration:
[0088] The i index represents 6 load factor conditions. Let be the per-unit load factor. The L-BFGS algorithm is used to solve this, with initial values set to [1.0, 1.0, 1.0] and a convergence threshold of 10. -6 .
[0089] The optimal parameters for calibrating an SSZ11-50000 / 110 transformer are shown in Table 3. Table 3 Optimal Parameter Results
[0090] After calibration, only , , It is stored in the system's non-volatile memory and can be accessed online. No data is stored. This process ensures that online operation is free of subjective, empirical parameters.
[0091] The 30-day monitoring data, compared with existing methods, are shown in Table 4: Table 4 Monitoring Data Results
[0092] When Δf is large, the double exponential model of the present invention can still maintain high accuracy, while the accuracy of the single exponential model in the published paper has decreased significantly. The method of the present invention has significant advantages and can be reproduced across devices.
[0093] Sixth, the green management module operates collaboratively.
[0094] like Figure 1 The green management module 600 in this embodiment operates according to the following strategy: Low-power management: A preset monitoring schedule is implemented, collecting data every 30 minutes from 8:00 AM to 8:00 PM daily, with each data collection session lasting 0.5 minutes (corresponding to Nt=30000 frames). The system remains in deep sleep mode during the remaining time. During sleep mode, the radar front-end is completely powered off, and the standby power consumption of the data acquisition module is <0.5mW. The average system power consumption is less than 0.5W. Compared to traditional continuous operation solutions (approximately 5W), power consumption is reduced by approximately 90%.
[0095] b. Carbon Emission Measurement: The carbon emission monitoring unit accumulates the energy consumption (Etotal) of each module. Users can configure the local power grid carbon emission factor (ε) through the communication interface; if not configured, the system's preset default factor is used (refer to the latest guidelines issued by the Ministry of Ecology and Environment of China). The cumulative carbon emissions are calculated and reported every 24 hours.
[0096]
[0097] c. Operational status correlation: The compensated vibration characteristic value is correlated with the load rate. When the Pearson correlation coefficient is below 0.6 for 3 consecutive days, an early warning is triggered, indicating that the sensor may have an angular shift or a change in surface scattering characteristics.
[0098] Example 3 This embodiment provides a computer device including a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), or field-programmable gate arrays (FPGAs). Gate Array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., are the computing and control core of computer equipment. They are suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions to realize corresponding method flows or corresponding functions. The processor described in this embodiment of the invention can be used in the operation of a non-contact multi-point vibration monitoring method for substation transformers based on millimeter-wave radar, including: acquiring a complex intermediate frequency signal obtained by transmitting a frequency-modulated continuous wave signal to the transformer through millimeter-wave radar and receiving the echo; performing a range-dimensional fast Fourier transform on the complex intermediate frequency signal to obtain a one-dimensional range image and the complex echo sequence corresponding to each range gate; and performing a local... The number of energy peaks and their full width at half maximum (FWHM) are used to decompose the vibration displacement sequences of at least two vibration response points from the complex echo sequence. Spectral analysis is performed on the vibration displacement sequences of each vibration response point to obtain its amplitude spectrum. Within the amplitude spectrum of each vibration response point, local peaks are searched in a preset neighborhood with the theoretical fundamental frequency as the reference center. The amplitude of these local peaks is extracted as the dominant frequency amplitude feature, and the corresponding actual peak frequency is also extracted. The dominant frequency offset of the actual peak frequency relative to the theoretical fundamental frequency is calculated. Based on the dominant frequency offset, an exponential correction model is used to compensate for the dominant frequency amplitude feature, resulting in compensated vibration characteristic values. The compensated vibration characteristic values are then correlated with synchronously acquired transformer operating reference values to output the transformer operating status assessment results.
[0099] Example 4 This embodiment provides a computer-readable storage medium (Memory), which is a memory device in a computer device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the operating system of the computer device. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device.
[0100] One or more instructions stored in the computer-readable storage medium can be loaded and executed by the processor to implement the corresponding steps of the non-contact multi-point vibration monitoring method for substation transformers based on millimeter-wave radar in the above embodiments; one or more instructions in the computer-readable storage medium are loaded and executed by the processor to perform the following steps: acquiring a complex intermediate frequency signal obtained by transmitting a frequency-modulated continuous wave signal to the transformer through millimeter-wave radar and receiving the echo; performing a range-dimensional fast Fourier transform on the complex intermediate frequency signal to obtain a one-dimensional range image and a complex echo sequence corresponding to each range gate; decomposing at least two local energy peaks and their full width at half maximum (FWHM) from the complex echo sequence based on the number of local energy peaks and their FWHM in the one-dimensional range image. The vibration displacement sequence of each vibration response point is obtained; spectral analysis is performed on the vibration displacement sequence of each vibration response point to obtain the amplitude spectrum of each vibration response point; in the amplitude spectrum of each vibration response point, local peaks are searched in a preset neighborhood with the theoretical fundamental frequency as the reference center, and the amplitude of the local peaks is extracted as the main frequency amplitude feature, and the actual peak frequency corresponding to the local peaks is extracted; the main frequency offset of the actual peak frequency relative to the theoretical fundamental frequency is calculated; based on the main frequency offset, an exponential correction model is used to compensate the main frequency amplitude feature to obtain the compensated vibration characteristic value; the compensated vibration characteristic value is correlated with the synchronously acquired transformer operation reference value to output the transformer operation status assessment result.
[0101] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, etc.) containing computer-usable program code.
[0102] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0103] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0104] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0105] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.
[0106] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.
[0107] It should be understood that the above description is for illustrative purposes and not for limitation. Many embodiments and applications beyond the provided examples will become apparent to those skilled in the art upon reading the above description.
Claims
1. A non-contact multi-point vibration monitoring method for substation transformers based on millimeter-wave radar, characterized in that, include: The complex intermediate frequency signal is obtained by transmitting a frequency-modulated continuous wave signal to a transformer via millimeter-wave radar and receiving the echo. A range-dimensional fast Fourier transform is performed on the complex intermediate frequency signal to obtain a one-dimensional range image and the complex echo sequence corresponding to each range gate; Based on the number of local energy peaks and full width at half maximum (FWHM) in a one-dimensional range image, vibration displacement sequences of at least two vibration response points are decomposed from the complex echo sequence. Spectral analysis was performed on the vibration displacement sequence at each vibration response point to obtain the amplitude spectrum of each vibration response point. In the amplitude spectrum of each vibration response point, local peaks are searched in a preset neighborhood with the theoretical fundamental frequency as the reference center. The amplitude of the local peaks is extracted as the main frequency amplitude feature, and the actual peak frequency corresponding to the local peaks is extracted. Calculate the frequency offset of the actual peak frequency relative to the theoretical fundamental frequency; Based on the main frequency offset, an exponential correction model is used to compensate for the main frequency amplitude characteristics, and the compensated vibration characteristic values are obtained. The compensated vibration characteristic values are correlated with the synchronously acquired transformer operating reference values to output the transformer operating status assessment results.
2. The non-contact multi-point vibration monitoring method for substation transformers based on millimeter-wave radar according to claim 1, characterized in that, Based on the number of local energy peaks and full width at half maximum (FWHM) in the one-dimensional range image, a vibration displacement sequence at least two vibration response points is decomposed from the complex echo sequence, including: When there are multiple independent energy peaks in a one-dimensional range image and the full width at half maximum (FWHM) of each independent energy peak is less than or equal to twice the system range resolution, an incoherent energy accumulation is performed on the one-dimensional range image along the slow time dimension to construct a statistical energy curve. Extract the original complex echo sequences corresponding to the distance gates of each independent energy peak; The original complex echo sequence is unwrapped and differentiated to generate a radially separable vibration displacement sequence for each vibration response point; When a single broadening peak exists in a one-dimensional range image and the full width at half maximum (FWHM) of the single broadening peak is greater than twice the system range resolution, a near-field steering vector is constructed based on the multi-channel original complex echo signal within the target range layer where the single broadening peak is located. Using the near-field steering vector, Bartlett beamforming spatial spectrum focusing is performed to search for multiple local peaks in the spatial spectrum and separate vibration response points at different angles in the same layer aliasing scene; The complex focusing signal of each vibration response point is extracted and subjected to phase demodulation and displacement conversion to generate the vibration displacement sequence of each vibration response point.
3. The non-contact multi-point vibration monitoring method for substation transformers based on millimeter-wave radar according to claim 2, characterized in that, Bartlett beamforming spatial spectrum focusing using near-field steering vectors includes: The search range of candidate points is limited to the spatial surface determined by the radial distance value of the target distance layer; The radial distance coordinates of each candidate point are fixed as the radial distance values corresponding to the target distance layer; The azimuth and elevation angles of each candidate point are traversed and searched to calculate the Bartlett spatial spectrum of the corresponding location.
4. The non-contact multi-point vibration monitoring method for substation transformers based on millimeter-wave radar according to claim 2, characterized in that, After constructing a statistical energy curve by incoherent energy accumulation of the one-dimensional range image along the slow time dimension, the following steps are also included: Extract multiple local peak points from the statistical energy curve; Multiple local peak points are filtered by applying minimum gate spacing constraints in spatial order to determine multiple target distance gates that are independent of each other in the radial direction. The minimum gate spacing constraint threshold is 2 to 5 times the system distance resolution. The process of extracting the original complex echo sequence of each independent energy peak corresponding to the range gate is performed based on multiple target range gates.
5. The non-contact multi-point vibration monitoring method for substation transformers based on millimeter-wave radar according to claim 1, characterized in that, An exponential correction model is used to compensate for the amplitude characteristics of the main frequency, including the following calculation formula: in, F The compensated vibration characteristic value, A Main frequency amplitude characteristics, This is the main frequency offset. c For compensation coefficient, These are the first-order frequency compensation coefficients. These are second-order frequency compensation coefficients, and c , and All system parameters were determined through offline joint optimization and calibration.
6. The non-contact multi-point vibration monitoring method for substation transformers based on millimeter-wave radar according to claim 5, characterized in that, c , and The offline joint optimization calibration process includes: Multiple sets of vibration signals were collected under various load rate conditions of the transformer. Obtain the main frequency amplitude characteristics, main frequency offset, and load rate per unit value under various load rate conditions; Introducing a normalization factor Build Minimum The objective function is , where The characteristic of the main frequency amplitude under the i-th operating condition is... Let be the main frequency offset under the i-th operating condition. Let i be the per-unit value of the load rate under the i-th operating condition; Simultaneously solve the problem using numerical optimization algorithms. , c , and The optimal value; Discard the optimal value The optimal values of the remaining parameters are solidified into the compensation coefficients, first-order frequency compensation coefficients, and second-order frequency compensation coefficients of the exponential correction model.
7. The non-contact multi-point vibration monitoring method for substation transformers based on millimeter-wave radar according to claim 1, characterized in that, Before transmitting the frequency-modulated continuous wave signal through the millimeter-wave radar front end, the following is also included: The millimeter-wave radar front end is activated according to the preset monitoring schedule to perform data acquisition operations on the transmitted signals; During non-acquisition periods after data acquisition operations are completed, the millimeter-wave radar front end is placed in a low-power sleep mode; Record the cumulative runtime and power consumption data of the monitoring process, and calculate the cumulative carbon emissions based on the power grid carbon emission factor and power consumption data.
8. A non-contact multi-point vibration monitoring system for substation transformers based on millimeter-wave radar, based on the method of any one of claims 1-7, characterized in that, Includes a millimeter-wave radar front-end module (100); The communication end of the millimeter-wave radar front-end module (100) is connected to a data acquisition and preprocessing module (200), the output end of the data acquisition and preprocessing module (200) is connected to a multi-point vibration separation module (300), and the output end of the multi-point vibration separation module (300) is connected to a vibration feature extraction and state assessment module (400). The power supply terminals of the millimeter-wave radar front-end module (100), the data acquisition and preprocessing module (200), the multi-point vibration separation module (300), and the vibration feature extraction and state assessment module (400) are all connected to the communication and power supply module (500). The control terminal of the millimeter-wave radar front-end module (100) is connected to a green management module (600). The millimeter-wave radar front-end module (100) includes virtual array elements arranged in a two-dimensional array, and the millimeter-wave radar front-end module (100) is located outside the transformer box and has an insulating distance gap with the surface of the transformer box. The main beam direction of the millimeter-wave radar front-end module (100) forms an angle of less than or equal to 8° with the normal direction of the surface of the transformer box.
9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the non-contact multi-point vibration monitoring method for substation transformers based on millimeter-wave radar as described in any one of claims 1 to 7.
10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the non-contact multi-point vibration monitoring method for substation transformers based on millimeter-wave radar as described in any one of claims 1 to 7.
Citation Information
Patent Citations
RD plane weak target detection tracking method and device based on speed square filter
CN110161494A
