Ultra-high frequency partial discharge positioning system based on optical synchronization time difference of arrival

Abstract

The application relates to the technical field of discharge positioning, and discloses a very-high-frequency partial discharge positioning system based on optical synchronization time difference, which is applied to a gas insulated enclosed device provided with a very-high-frequency sensor and comprises the following steps: a sensor array and a three-dimensional coordinate system representing a propagable space and a medium boundary are established; a calibration excitation is applied to pre-acquire a frequency domain calibration reference; when a discharge is triggered, original very-high-frequency waveforms are synchronously collected, a first-arriving frequency spectrum segment and an original arrival time are extracted, the original arrival time is corrected by extracting a dispersion time delay correction amount from the frequency domain calibration reference, a target time difference is obtained; candidate discharge coordinates are calculated based on internal propagation constraint paths of the device, the candidate discharge coordinates are mapped to a device surface with a discharge occurrence condition to obtain actual discharge coordinates, and a partial discharge source positioning result is output. The application can compensate for systematic deviation of a first-arriving time caused by internal components of the device, invalid positioning points can be avoided, and the positioning result can serve on-site maintenance.

Description

Technical Field

[0001] This invention relates to the field of discharge location technology, and more specifically, to an ultra-high frequency partial discharge location system based on optical synchronous arrival time difference. Background Technology

[0002] Gas-insulated enclosures are widely used in high-voltage power transmission and transformation systems, and their internal insulation condition directly affects operational reliability. Ultra-high frequency (UHF) partial discharge detection is a common method for assessing the insulation condition of such equipment: multiple UHF sensors are installed on the equipment casing to receive discharge pulses, and the spatial location of the discharge source is calculated based on the arrival time difference of the signals received by each sensor. In engineering practice, optical synchronization is widely used to unify the acquisition time of each sensor, effectively suppressing clock drift between channels, and time synchronization accuracy is no longer the main source of positioning error. However, during the propagation of the discharge pulse from the discharge source to each sensor, it needs to pass through insulators, bending nodes, branch nodes, and breaking components. Different frequency components and propagation modes exhibit significant differences in attenuation, reflection, and group delay. The spectral composition of the initial signal received by each sensor differs from the initial discharge signal. Existing positioning methods generally approximate the propagation inside the equipment as a straight-line propagation in a uniform medium, taking the first peak of the waveform or the energy inflection point as the first arrival time, without considering the changes in the spectral composition of the first arrival signal caused by internal components, thus introducing a systematic arrival time deviation. After clock errors are suppressed by optical synchronization, these deviations become the main source of residual errors in time difference positioning, which cannot be eliminated by conventional time synchronization calibration and link calibration. In addition, existing spatial calculations mostly use linear distance differences, and the calculation results often fall inside metal conductors or areas that electromagnetic waves cannot reach, which do not correspond to the specific components that can be repaired, and on-site troubleshooting still requires relying on experience. Summary of the Invention

[0003] This invention provides an ultra-high frequency partial discharge positioning system based on optical synchronous time difference of arrival, which solves the technical problems mentioned in the background art.

[0004] To achieve the above-mentioned objectives, the present invention provides the following technical solution.

[0005] A UHF partial discharge location system based on optical synchronous time difference of arrival is applied to a gas-insulated enclosure equipped with multiple UHF sensors that synchronously acquire data under a unified time reference, and is configured to perform: A three-dimensional structural coordinate system for the sensing array and the device is established based on multiple ultra-high frequency sensors. The three-dimensional structural coordinate system of the device represents the propagable space and medium boundary inside the gas-insulated sealed device. A calibration excitation is applied to the gas-insulated enclosure to obtain a frequency domain calibration reference generated by the calibration excitation and characterizing the dispersion delay characteristics caused by the internal structure of the gas-insulated enclosure. When the partial discharge pulse triggering condition is met, the original ultra-high frequency waveform is acquired through the sensor array. Extract the first-reach spectrum segment of the original UHF waveform; The original arrival time of the original UHF waveform is extracted, and the structural dispersion delay correction is extracted using the frequency domain calibration reference and the first arrival spectrum segment. The original arrival time is corrected with the structural dispersion delay correction to obtain the target arrival time difference. Candidate discharge coordinates are obtained based on the internal propagation constraint path of the device, which is constrained by the propagable space and the medium boundary. The candidate discharge coordinates are mapped to the physical surface of the device that meets the conditions for discharge to occur, thus obtaining the actual discharge physical coordinates. The output includes the localization result of the partial discharge source, which contains the actual physical coordinates of the discharge.

[0006] Compared with the prior art, the present invention has the following substantial features and significant progress: This invention, based on establishing a sensor array and a three-dimensional coordinate system characterizing the propagable space and medium boundary, first applies a calibration excitation to the device to obtain a frequency domain calibration reference in advance. After discharge triggering, the original UHF waveforms of each sensor are synchronously acquired. The dispersion delay correction is extracted by combining the first arrival spectrum segment with the frequency domain calibration reference. After correcting the original arrival time, the target arrival time difference is obtained. Then, the candidate discharge coordinates are calculated along the internal propagation constraint path constrained by the propagable space and medium boundary, and mapped onto the device surface with discharge conditions to obtain the actual discharge coordinates. Using this invention, the systematic deviations caused by insulators, bending nodes, and breaking components in the arrival time difference are compensated, and the residual positioning error is reduced accordingly. The constraint path calculation avoids invalid positioning points caused by linear difference. The positioning results directly correspond to the equipment surface that can be repaired, and on-site processing can be arranged according to the relevant section. The entire process is completed by sensors and fiber optic links installed outside the equipment, without the need to install active detection devices inside the equipment or perform high-precision manual mapping. Attached Figure Description

[0007] Figure 1 This is a system module diagram of the present invention; Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0008] The following description is provided in conjunction with the accompanying drawings. It should be understood that the following embodiments are for illustrative purposes only and are not intended to limit the invention; those skilled in the art can make equivalent substitutions or combinations for the specific implementations.

[0009] This specific embodiment relates to an ultra-high frequency partial discharge location system based on optical synchronous time difference of arrival. The system is applied to a gas-insulated enclosure equipped with multiple ultra-high frequency sensors that are synchronously acquired under a unified time reference, and is used to spatially locate the partial discharge source occurring inside the gas-insulated enclosure. Figure 1 This is a system architecture diagram. Figure 2 This is a flowchart of the positioning method. In this embodiment, unless otherwise specified, the values ​​of time-related physical quantities are in nanoseconds, the values ​​of spatial distance-related physical quantities are in meters, the values ​​of frequency-related physical quantities are in megahertz, the values ​​of angle-related physical quantities are in degrees, and the values ​​of voltage-related physical quantities are in millivolts.

[0010] The gas-insulated enclosure equipment consists of a metal-enclosed coaxial chamber formed by components such as a busbar, basin insulators, three-dielectric joint edges, bent connection nodes, shunt nodes, and disconnecting switch contacts. When a partial discharge pulse propagates within this chamber, the different frequency components and propagation modes of the UHF signal exhibit significant differentiation in attenuation, reflection, and group delay. The spectral composition of the initial signal received by each UHF sensor differs from the initial discharge signal, causing a systematic deviation in arrival time, i.e., spectral timescale deflection error. Although optical synchronization technology effectively suppresses clock drift between acquisition nodes, it cannot eliminate this systematic deviation caused by the propagation mechanism within the equipment. This deviation constitutes the main source of residual error in arrival time difference positioning after optical synchronization. On the other hand, existing spatial calculations are mostly performed using linear distance differential methods, and the calculation results often fall inside the metal conductor or in non-propagating areas where electromagnetic waves cannot reach, making it difficult to correspond to the surface of equipment components available for maintenance. This system extracts the structural dispersion delay correction by comparing the pre-acquired frequency domain calibration benchmark with the first arrival spectrum segment, compensates for the original arrival time, performs spatial calculation based on the internal propagation constraint path of the device constrained by the propagable space and medium boundary, and then maps the obtained candidate discharge coordinates to the physical surface of the device that has the conditions for discharge to occur, thereby solving the above two problems at the same time.

[0011] See Figure 1The system includes a gas-insulated enclosure device 111, an ultra-high frequency sensor group 112, a calibration excitation source 113, an optical synchronous acquisition unit 114, a frequency domain calibration reference unit 115, a first-arrival spectrum extraction unit 116, a dispersion delay correction and arrival time difference calculation unit 117, a constraint path coordinate calculation and surface mapping unit 118, and a positioning result output unit 119. The UHF sensor group 112 consists of multiple UHF sensors, which are installed at different monitoring ports on the outer shell of the gas-insulated enclosure 111, forming the sensing front end of the sensor array. The calibration excitation source 113 is located on the outside of the gas-insulated enclosure 111 and is used to apply calibration excitation to the gas-insulated enclosure 111. The optical synchronization acquisition unit 114 includes a synchronization time master station and multiple acquisition nodes. Each acquisition node is physically connected to the radio frequency interface behind a corresponding UHF sensor. Each acquisition node is connected to the synchronization time master station through optical fiber to complete optical synchronization locking. The frequency domain calibration reference unit 115, the first-arrival spectrum extraction unit 116, the dispersion delay correction and arrival time difference calculation unit 117, the constraint path coordinate calculation and surface mapping unit 118, and the positioning result output unit 119 are connected in sequence to form the positioning calculation and output link.

[0012] See Figure 2 The positioning process of this system includes steps S211 to S218. Step S211: Establish a three-dimensional structural coordinate system between the sensor array and the device. Step S212: Apply a verification excitation to obtain a frequency domain calibration reference. Step S213: Determine whether the partial discharge pulse triggering condition is met. If the result is negative, return to continuous monitoring and continue judging. If the result is positive, proceed to step S214. Step S214: Acquire the original UHF waveform. Step S215: Extract the first arrival spectrum segment. Step S216: Extract the structural dispersion delay correction amount and correct the original arrival time to obtain the target arrival time difference. Step S217: Calculate candidate discharge coordinates based on the propagation constraint path inside the device. Step S218: Map the coordinates to the physical surface of the device that meets the conditions for discharge occurrence and output the partial discharge source positioning result.

[0013] In step S211, combined with Figure 1 The UHF sensor array 112 and optical synchronous acquisition unit 114 shown in the diagram establish a sensing array and a three-dimensional structural coordinate system for the equipment based on multiple UHF sensors. The sensing array refers to an array-based detection system composed of multiple UHF sensors and their corresponding acquisition nodes, synchronously acquiring data under a unified time reference. The three-dimensional structural coordinate system of the equipment serves as a three-dimensional spatial reference characterizing the propagable space and medium boundaries within the gas-insulated enclosed equipment, providing a unified spatial reference for solving subsequent propagation constraint paths within the equipment and mapping candidate discharge coordinates to the physical surface of the equipment.

[0014] In one implementation, the three-dimensional structural coordinate system of the equipment is established using manufacturing drawings of the gas-insulated enclosure as the data source. Specifically, the manufacturing drawings of the gas-insulated enclosure are acquired, and these drawings include the real-valued geometric and physical dimensions of the equipment's outer shell center axis, inner conductor centerline, and each branch node. Using the geometric center of the flange connection section at one end of the gas-insulated enclosure as the three-dimensional absolute zero point, an axial mileage scale is established along the direction extending from the center axis of the main busbar cylinder to the other end. The direct extension distance from each internal solid component, medium partition plate, and diameter-changing node to the three-dimensional absolute zero point is extracted and recorded as the axial mileage. The upper limit of the axial mileage range depends on the total length of the equipment being measured, and the distance measurement accuracy is fixed at 0.01. Using the central axis of the main busbar as the pole, a circumferential polar scale is established on the normal plane perpendicular to the central axis of the main busbar. The vertically upward direction is set as the starting 0 degrees. The relative angular deflection values ​​of each external observation window and branch downlead are extracted and recorded in a clockwise direction as the radial azimuth. The range of the radial azimuth is truncated and restricted to a continuous closed interval from 0 degrees to 360 degrees. The radial spatial straight-line distance span from the outer surface of the internal high-voltage center conductor to the inner surface shielding boundary of the external grounded metal shell is measured as the cavity radius. The radius constant of the inner conductor's outer surface ranges from 0.05 to 0.08, and the radius constant of the metal shell's inner surface ranges from 0.2 to 0.4. Tensor binding is performed on the axial mileage distance sequence, the radial azimuth angle sequence, and the cavity radius dimension sequence to construct and generate three-dimensional spatial coordinate data. In another alternative implementation, the equipment's three-dimensional structural coordinate system can also be established based on spatial data obtained through on-site measurement modeling or three-dimensional scanning. The essence of both is to obtain a three-dimensional coordinate reference that characterizes the boundary of the propagable space and medium. Those skilled in the art can select the appropriate data source according to the completeness of the equipment file.

[0015] The establishment of the sensor array includes two aspects: spatial point calibration and time reference unification. For spatial point calibration, multiple UHF sensors for sensing spatial electromagnetic waves are arranged at the outer flange of the gas-insulated enclosure equipment. The minimum threshold for the number of UHF sensors deployed is four. The set of spatial coordinates of the equivalent geometric origin point when the internal microwave antenna structure of each UHF sensor receives electromagnetic wave energy in a specific frequency band is obtained and set as the phase center. Using the normal projection calculation rule based on the minimum Euclidean straight-line distance, the two-dimensional coordinate intersection point projected from the phase center along the shortest straight-line spatial path onto the surface of the internal cavity of the equipment's metal shell is calculated. The geometric boundary range of the gas barrier insulation plane where this two-dimensional coordinate intersection point is located is the corresponding installation port section. The three-dimensional spatial coordinate parameters of the phase center are transferred and attached to the corresponding installation port section through a coordinate translation transformation operator, thereby calibrating the true coupling starting position between the UHF sensor and the coaxial electromagnetic wave propagation space inside the equipment. For time reference unification, a broadband analog-to-digital converter (ADC) electronic hardware unit physically connected to the RF interface of each UHF sensor is used as a data acquisition node, with each UHF sensor corresponding to one data acquisition node. A synchronization time master station capable of continuously outputting highly stable reference clock optical pulses is configured in the measurement and control area. All data acquisition nodes and the synchronization time master station are connected to form a star topology network via optical fiber, using quartz glass fiber. The synchronization time master station continuously sends time base calibration optical pulses to each data acquisition node via optical fiber. The photoelectric clock module inside each data acquisition node adjusts the trigger count value of its local ADC sampling hardware based on the received time base calibration optical pulses and transmits feedback optical pulses to the synchronization time master station along the original optical fiber. After the system's embedded loopback bidirectional delay measurement and delay compensation addition and subtraction calculation operations, the trigger timestamp variable of the local hardware clock of all data acquisition nodes and the absolute time base deviation of the synchronization time master station are stabilized within a very small range where the absolute value is less than or equal to 1, achieving optical synchronization locking. Thus, multiple UHF sensors, after spatial point coordinate anchoring mapping and time reference unification binding processing, constitute a sensor array.

[0016] Furthermore, based on the three-dimensional spatial coordinate data, a device structure coordinate table describing the propagation space of the metal cavity, the medium boundary, and the phase centers of multiple UHF sensors is generated as the device's three-dimensional structural coordinate system. Specifically, based on the three-dimensional spatial coordinate data that has been reconstructed, a three-dimensional spatial connected lattice domain is extracted inside the gas-insulated enclosure device. This domain is completely filled with insulating gas medium, not occupied by the metal inner conductor or grounding shield, and allows UHF electromagnetic waves to propagate in refraction and reflection modes. This domain is then marked and extracted as the propagation space of the metal cavity. The three-dimensional geometric interface coordinate distribution matrix of non-gas solid insulating shielding components that generate high-frequency electromagnetic wave energy reflection and cause differentiation and attenuation of group delay characteristics is identified and extracted. The three-dimensional contour coordinate values ​​of the physical blocking surface envelope of the basin insulator, the three-medium junction edge region, the right-angle bending connection node, the three-way shunt node, and the internal disconnecting switch metal break are extracted. The spatial set of these physical blocking surface envelope three-dimensional contour coordinate values ​​is defined as the medium boundary. The three-dimensional coordinate point set of the propagation space of the metal cavity, the three-dimensional coordinate point set of the medium boundary, and the three-dimensional spatial reference coordinate data of multiple phase centers after projection calibration are uniformly loaded into memory. Multi-dimensional floating-point data matrix splicing and topology graph formatting combination and concatenation operations are performed, and encapsulated into a device structural coordinate table. This device structural coordinate table is configured in the overall system architecture as the global three-dimensional spatial reference, i.e., the device's three-dimensional structural coordinate system, for subsequent computation modules to perform spatial discharge path distance optimization iteration and solve the receiver point arrival time difference positioning equation.

[0017] Step S212 is the calibration preparation stage. See [link / reference] Figure 1 The verification excitation is applied to the gas-insulated enclosure 111 by the verification excitation source 113, and the frequency domain calibration reference is generated by the frequency domain calibration reference unit 115. The frequency domain calibration reference, as prior data characterizing the dispersion delay characteristics of the internal structure of the gas-insulated enclosure, is invoked when extracting the structural dispersion delay correction in step S216.

[0018] In one embodiment, the calibration excitation source 113 is an external pulse source. The output of the external pulse source is connected to the external monitoring medium window of the gas-insulated enclosure. The external monitoring medium window is either a non-metallic epoxy resin observation port pre-cast and sealed by the manufacturer on the metal shell surface of the gas-insulated enclosure, or a dedicated electromagnetic calibration port that does not change the original electric field distribution of the internal cavity. The rise time constant of the electromagnetic pulse output by the external pulse source is set to be limited to between 0.5 and 1, and the signal amplitude of the output high-frequency pulse voltage is forcibly limited to the output range of 50 to 100. This amplitude range ensures that the electromagnetic energy physically excited by the pulse is always lower than the initial physical triggering threshold that could cause a real electric field breakdown or abnormal discharge event in the high-voltage insulating gas inside the device under test, thus ensuring that the calibration injection will not induce a real discharge. Using a mechanically limiting coupling clamp with a specific microwave impedance matching structure, the output of the external pulse source is directly and rigidly pressed and attached to the outer physical barrier surface of the external monitoring medium window, establishing a lossless physical coupling channel path for the external calibration electromagnetic pulse to penetrate and radiate into the coaxial waveguide space inside the enclosed metal device.

[0019] With the synchronization time master station locked, multiple sets of broadband verification pulses are continuously injected as verification excitation. Specifically, the dynamic hardware trigger count values ​​of the local photoelectric conversion clock modules of all acquisition nodes in the sensor array are continuously scanned and extracted. The dynamic hardware trigger count value of each acquisition node is cyclically compared with the absolute timestamp reference parameter of the reference clock optical pulse sent by the synchronization time master station to the fiber optic port of each node. When the absolute value of the real-time local hardware trigger time alignment time difference offset of all acquisition nodes continuously converges within the target fault tolerance range of less than or equal to 1, and this stable alignment operation is maintained for a time span of up to 10,000,000,000, the system is determined to be in the synchronization time master station locked state. The automated software control logic program calls the transmitter control data interface of an external pulse source, causing it to continuously emit a sequence of electromagnetic pulses covering the effective frequency range of the ultra-high frequency broadband working detection into the coaxial medium of the equipment cavity according to a constant high-frequency repetitive cycle. This sequence serves as broadband verification pulses. The injection time interval between two adjacent broadband verification pulses is set to 20,000,000, corresponding to a broadband verification pulse emission frequency constant of 50. A total of 100 independent broadband pulses are continuously and uninterruptedly emitted. During the injection calibration operation, the system's built-in pulse amplitude threshold monitoring program continuously scans background state characteristics to eliminate random continuous discharge pulses spontaneously generated by defects in the internal insulation material or surface-attached particles within the cavity of the tested equipment. This prevents temporal crosstalk between the actual partial discharge characteristic physical pulse waveform and the verification reference waveform introduced into the system.

[0020] After each set of broadband verification pulses is injected, the control program instructs each acquisition node to synchronously perform the freeze-latch processing of the voltage amplitude digital sampling point matrix data segment intercepted by the local underlying analog-to-digital conversion module, based on the absolute optical timestamp variable that has been uniformly phase-locked across the entire network. This process assembles and generates a two-dimensional discrete floating-point time series numerical array containing the horizontal axis dimension of time flow advancement and the vertical axis dimension of electromagnetic wave voltage induction acquisition amplitude, and stores it as the calibration waveform of the corresponding acquisition node. The calibration waveform data array stored in the system memory space is retrieved. A segment of random low-amplitude background noise data is extracted from the left side of the continuously shifting time axis, before the starting point of the physical energy inflection point where the UHF pulse energy shows a significant increase and reversal. The arithmetic mean and average of all discrete voltage amplitude samples contained in this segment are calculated to obtain the DC bias parameter inherent to this specific hardware acquisition channel. The DC bias parameter is then uniformly subtracted from the voltage floating-point real value sequence of all signal sampling points within the total recording length of the calibration waveform, and a background baseline zeroing subtraction operation is performed. The envelope of the clean, unbiased baseline voltage discrete time domain sequence waveform, which exhibits a symmetrical oscillation and decay distribution around the absolute zero voltage point after baseline subtraction, is extracted and retained as the corresponding calibration impulse response.

[0021] The frequency domain calibration reference unit 115 performs frequency domain analysis on the calibration impulse response, extracting the calibration spectrum centroid and calibration group time delay gradient as the frequency domain calibration reference. Specifically, it extracts the voltage discrete time sequence data segment of the calibration impulse response of each acquisition channel, performs a discrete Fourier frequency domain mapping transformation, and converts the waveform sequence data representing the one-dimensional time domain amplitude into frequency domain energy distribution characteristic complex data. From the converted frequency domain energy distribution characteristic complex data, it extracts the square power of the absolute value of the complex characteristic amplitude at each frequency calculation coordinate step point as the real value of the voltage square spectral density distribution characteristic, and simultaneously extracts the corresponding complex characteristic angle offset physical quantity as the phase parameter. Within a continuous closed frequency range bounded by 300 MHz to 1500 MHz, scalar multiplication is performed on the frequency independent variable and the corresponding real value of the voltage square spectral density distribution characteristic. The results of all single-step scalar multiplications are then continuously summed forward along the frequency variation axis to obtain the first integral characteristic scalar result used as the numerator parameter. Within the same frequency range, the real value of the voltage square spectral density distribution characteristic is directly summed along the frequency variation axis to obtain the second integral characteristic scalar result used as the denominator parameter. The first integral characteristic scalar result is divided by the second integral characteristic scalar result to obtain the calibrated spectral centroid. ; in, The effective frequency band lower limit constant is 300 MHz. The upper limit of the effective frequency band is a constant of 1500 MHz. To calibrate the real-valued characteristics of the voltage squared spectral density distribution of the impulse response. This is for calibrating the spectral centroid. The calibrated spectral centroid physically reflects the absolute offset of the geometric center of the calibrated RF signal's broadband energy concentration within the channel node.

[0022] Furthermore, all phase parameters covering the continuous frequency band range are extracted from the output of the Discrete Fourier Frequency Domain Mapping Transform. The first-order derivative is numerically differentiated relative to the coordinates of the independent variable at the step frequency. The real value of the slope of the derivative is multiplied by -1 to reverse its polarity. Then, the product value after polarity reversal is divided by the product of a constant 2 and pi. This outputs the channel calibration group time delay parameter, which can physically characterize the spatial morphology of the nonlinear delay distribution law of the physical propagation penetration time of various frequency components within different frequency bands of the radio frequency electromagnetic wave group during medium propagation. ; in, For phase parameters, The group delay parameter is used to calibrate the channel.

[0023] Subsequently, bias calculations are performed within the continuous frequency integration interval bounded by 300 MHz to 1500 MHz: the addressing extracts the real value of the reference delay constant at the corresponding point on the vertical axis of the frequency corresponding to the centroid of the calibration spectrum, representing the channel calibration group delay parameter; the bandwidth subtraction difference offset feature term generated by subtracting the value of the abscissa parameter of any frequency in the interval from the value of the centroid of the calibration spectrum is extracted, and the group delay difference offset feature term generated by subtracting the real value of the reference delay constant at the point is extracted simultaneously; the two difference offset feature terms and the real value of the voltage squared spectral density distribution feature at the corresponding frequency sampling point are then subjected to ternary continuous multiplication calculation. The result of the ternary multiplication is subjected to discrete definite integral multiplication along the entire sampling frequency coordinate axis to obtain the normalized partial numerator gradient feature real value. Simultaneously, the basic subtraction difference feature term generated by subtracting the frequency abscissa parameter from the value of the calibration spectrum centroid is itself subjected to quadratic power calculation. The squared result is multiplied in place with the voltage square spectral density distribution feature real value at the corresponding frequency sampling point. The multiplication result is then subjected to discrete definite integral summation along the entire sampling frequency coordinate axis to obtain the attenuation feature real value used as the denominator scaling constraint. The normalized partial numerator gradient feature real value is divided by the attenuation feature real value to obtain the calibration group delay gradient. ; in, To calibrate the group delay gradient, This represents the real value of the point reference delay constant for the channel calibration group delay parameter at the centroid of the calibration spectrum. The calibration group delay gradient characterizes the rate of change of the discrete slope of the time-domain wavefront dispersion of each branch of the electromagnetic wave composite component due to the physical blocking effect of the complex metal inner partition and the supporting insulating solid dielectric structure of the equipment cavity.

[0024] Finally, the calibration spectrum centroid and calibration group delay gradient of each acquisition channel are numerically concatenated using a multidimensional tensor variable sequence and encapsulated in the system's underlying memory to obtain the frequency domain calibration benchmark. The frequency domain calibration benchmark is an independent two-dimensional prior data benchmark for each acquisition channel, characterizing the dispersion delay characteristics caused by the propagation of ultra-high frequency signals from the internal structure of the gas-insulated enclosure equipment, and serving as a static reference for subsequent extraction of structural dispersion delay correction.

[0025] In addition, during the calibration phase, the system synchronously records the fixed delay of the synchronization link measured by the optical link loopback. Specifically, in the central control and processing core of the synchronization time master station, the absolute photoelectric timestamp of the moment when the optical pulse of the synchronization pair is transmitted to the output port of the optical fiber link of the acquisition node is continuously extracted in a loop. This is compared with the absolute photoelectric timestamp of the arrival time of the optical pulse of the loopback pair received through the reverse optical fiber from the physical feedback of the photoelectric transceiver module of the acquisition node. The real value of the time difference between the two absolute photoelectric timestamps is calculated. This real value of the time difference is divided by a fixed constant of 2 to obtain the physical time consumption of unidirectional optical pulse propagation. The physical time consumption of unidirectional optical pulse propagation is superimposed with the preset delay scalar of the fixed constant of the signal sensing response of the analog-to-digital conversion trigger circuit of each acquisition front end. The results are then calculated and combined to obtain the fixed delay of the synchronization link of the acquisition channel, which is the sum of the node hardware sensing response delay and the optical fiber transmission fixed delay, and stored.

[0026] After entering the monitoring stage, the optical synchronous acquisition unit 114 executes step S213 to determine whether the partial discharge pulse triggering conditions are met. Specifically, all acquisition nodes are controlled to continuously cache monitoring data under a unified time reference: The system acquires the time synchronization pulse command issued by the pre-configured star fiber optic topology, forcibly locking and aligning the local internal trigger clocks of all acquisition nodes to maintain a unified time reference with an absolute error value less than or equal to 1. Within this reference framework, the analog-to-digital converter hardware array distributed across each monitoring channel is driven to perform continuous electromagnetic wave spatial voltage signal acquisition and quantization. The fixed physical sampling frequency of the analog-to-digital converter hardware array is set to 5,000,000 times per second, and the constant range of the analog radio frequency signal detection bandwidth is limited to between 300 MHz and 1500 MHz. A first-in-first-out (FIFO) circular queue space with a fixed length of 20,000 floating-point data storage units is independently allocated in the system memory for each physical communication channel. The discrete digital quantized voltage data points output by the hardware acquisition and conversion are pushed into the corresponding circular queue space one by one according to the order of physical time elapsed. When new acquired data fills the 20,000 floating-point data storage units and an address overflow occurs, a memory address pointer reset loop mechanism is used to overwrite the oldest data block initially written with the latest sampled data, achieving rolling updates.

[0027] During the parallel execution of continuous waveform acquisition and rolling write, the underlying voltage amplitude threshold comparison and monitoring program is configured: a constant for the absolute value of voltage fluctuation extremes used to determine the starting point of abnormal insulation discharge events is set and fixed at 15. The physical logic judgment action of continuously breaking through and exceeding the numerical extreme value of 15 in the real-time acquired and converted voltage digital quantity is set as the partial discharge pulse trigger condition. The real-time input data of all channels is continuously scanned through high-speed parallel logic judgment gate circuits. When the real-time voltage absolute value on any UHF sensor monitoring link exceeds 15, it is determined that the partial discharge pulse trigger condition is met. If the judgment result is not met, the system maintains the continuous monitoring state and returns to continue scanning and judgment; if the judgment result is met, the acquisition node that first generates the trigger alarm judgment action broadcasts a global hardware freeze command with a unique event identifier to the entire network through the fiber optic bus, and proceeds to step S214.

[0028] In step S214, after receiving the global hardware freeze command, the system's underlying control logic starts a time delay write counter with a total of 2000 sampling period constants to continue recording the corresponding delay continuation waveform, with a corresponding duration of 400. After the count value returns to zero one by one, the hardware circuit forcibly blocks the address pointer push operation of all underlying analog-to-digital converters writing new data to the circular queue space, and marks the continuous digital voltage dot matrix array containing the state before and after the sudden change in the memory as synchronous cache data containing the current discharge event.

[0029] Subsequently, the voltage time series, hardware trigger timestamp, and sampling interval parameters of each acquisition node are extracted. Specifically, from the locked and frozen address blocks of each channel's storage queue, the continuously stored discrete digital voltage values ​​are read according to the first-in-first-out timing sequence. These values ​​are then arranged in the order of recording to construct a one-dimensional data vector combination entity containing 20,000 constant floating-point amplitude components, which is defined as the voltage time series of the corresponding acquisition node. The internal time digital register of the underlying optical timing clock control module of each channel is called to extract the floating-point real number of the clock reading physically latched and saved at the moment the trigger judgment threshold takes effect, which is defined as the hardware trigger timestamp representing the offset parameter of the network-wide relative zero-point burst time of the current event. The previously set sampling frequency is called, and the reciprocal of the sampling frequency is taken to obtain the time interval between adjacent sampling points, which is converted into a value of 0.2 in nanoseconds and defined as the sampling interval parameter.

[0030] Furthermore, the voltage time series containing the pre-trigger background noise segment, the first arrival waveform segment, and the first reflection segment is packaged into an event data packet to obtain the original UHF waveform. Specifically, a segmentation mapping scan is performed on the voltage time series with a total length of 20,000 points: a pure random low-amplitude fluctuation data segment of 5,000 sampling points is extracted forward from the coordinate point corresponding to the hardware trigger action as the endpoint, which is defined as the background noise segment before triggering. This segment covers the cavity noise level before discharge occurs. The initial coordinate base point is the physical abrupt point where energy initially rises upward, and the initial pulse interval point array is extracted and defined as the first arrival waveform segment. This segment covers the pure components that arrive along the fastest constrained geometric path of the cavity. After the end point of the first arrival waveform segment, the scan continues to extend backward, and the interval point array sequence of the secondary composite energy pulse oscillation rising wave packet caused by the obstruction, refraction and reflection of the equipment cavity solid end cover and the internal metal conductor dielectric partition is defined as the first reflection wave segment. The discrete sequence data of the voltage time series of the three segments, together with the physical space number of the UHF sensor, the floating-point real value of the hardware trigger timestamp, and the constant value of the sampling interval parameter, are injected into a dictionary-style hierarchical multidimensional composite data structure encapsulation with formatted file header identification words and tail check mark information. The encapsulation and combination are performed to package the original UHF waveform and output it.

[0031] After obtaining the original UHF waveform, the first-reach spectrum extraction unit 116 executes step S215 to extract the first-reach spectrum segment of the original UHF waveform. The extraction process includes three steps in sequence: waveform preprocessing, first-reach time window truncation, and frequency domain transformation.

[0032] Waveform preprocessing: The background noise segment before triggering is extracted from the pre-cached event data packet, specifically an array of 5000 discrete floating-point real-value voltage sampling points preceding the coordinate point corresponding to the hardware trigger action. Each of these 5000 discrete floating-point real-values ​​is summed using scalar addition, divided by the total sample size constant of 5000, to calculate the arithmetic mean real value representing the channel's baseline zero-point drift error. This value is then marked and extracted as the DC bias parameter for the corresponding acquisition node. Subsequently, the original UHF waveform covering the complete discharge event cycle is invoked, and a subtraction calculation is performed on each of the 20000 constant sampling nodes to remove the DC bias parameter. This aligns and corrects the background baseline with random positive and negative drift to a physical horizontal reference axis with an absolute value constant of 0. Furthermore, zero-phase digital bandpass filtering is performed on the waveform after bias removal according to the effective frequency band range of the UHF sensor: the upper and lower boundary frequency constant parameters of the effective frequency band range are read, namely the numerical constant of 300 MHz limiting the lower limit of the cutoff operating frequency region and the numerical constant of 1500 MHz limiting the upper limit of the cutoff operating frequency region; a Butterworth digital bandpass filter network model parameter matrix set with a fixed order of 4 is instantiated in the signal processing kernel; the waveform after bias removal is subjected to the first discrete convolution filtering along the time-increasing positive coordinate axis, and the first intermediate pass is output. The filter sequence array is flipped and rearranged in reverse order along the discrete time coordinate axis. A second discrete convolution filter is performed on the flipped and rearranged data using the same Butterworth digital bandpass filter network model parameter matrix set, outputting a second intermediate over-filtered sequence array. The second intermediate over-filtered sequence array is then flipped and restored along the spatial axis. Through symmetrical forward and reverse convolution mapping calculations, the group phase time delay distortion difference generated during the analog hardware filtering process is mathematically canceled out, resulting in a phase-fidelity and frequency-clean time-domain signal.

[0033] The initial arrival time window interception stage: The initial arrival time window is intercepted within the energy rise interval of the filtered waveform. Specifically, a short-time sliding energy integral calculation engine based on a constant step size of 10 discrete sampling points is activated. The engine continuously calculates the sum of the squares of the absolute values ​​of 10 consecutive voltage values ​​within the sliding cursor frame as time progresses. The difference points calculated at adjacent step times are compared and identified using first-order derivative difference and second-order derivative extreme value mapping. When the first data-related coordinate mapping point where the second-order derivative mathematical evaluation result abruptly jumps from a constant of 0 to a positive maximum peak is identified, this physical coordinate mapping point is marked as the first energy rise point of the filtered waveform. This base point coordinate is then used to establish the initial arrival point of the UHF wavefront reaching the receiving probe. To prevent secondary reflection and superposition pollution caused by the internal high-voltage bus conductor support structure and the basin insulator boundary within the cavity of the insulating gas equipment, the one-dimensional time envelope data of the calibration impulse response calibrated under the same channel is used. The coordinates of the first direct energy peak with the highest absolute amplitude are searched. Continuing along the time axis, the search continues for the location of the second local extremum peak node that appears more than 20 sampling points away from it. This location coordinate is then bound to the time corresponding to the first reflection peak in the calibration impulse response. A time frame is defined extending backward from the established initial arrival point. A threshold function for the coordinate minimum value is forcibly set to block and discard multipath fading and mixed waveform segments after crossing the limited coordinate time. A pure, independent one-dimensional floating-point sequence with a length ranging from 2 to 10 is extracted and set as the first arrival time window. The end time of the first arrival time window does not exceed the time corresponding to the first reflection peak in the calibration impulse response, thus ensuring that the window only contains the initial arrival direct wavefront component.

[0034] Frequency domain transformation stage: Frequency domain transformation is performed on the waveform data within the first arrival time window to obtain the first arrival spectrum segment. Specifically, the corresponding pure one-dimensional discrete voltage sequence combination is extracted from the first arrival time window and marked as waveform data. Its length covers 10 to 50 discrete sampling sequence points, corresponding to a duration of 2 to 10. To meet the binary addressing length requirements of the underlying discrete Fourier calculation operator module and to improve the smooth mapping and refinement resolution of the frequency calculation grid output distribution, real values ​​with constant 0 are continuously filled at the array positions at the end of the waveform data sequence. The length of the one-dimensional tensor is stretched and solidified to a constant size of 1024 standard sampling coordinate points. The expanded and recombined sequence is processed by fast Fourier integral arithmetic matrix transformation mapping, which deconstructs and recombines the one-dimensional time domain data entity that reflects voltage amplitude fluctuations into a high-dimensional frequency domain composite matrix set containing frequency point abscissas and corresponding independent real and imaginary real part distribution components. This set is solidified and defined as the first arrival spectrum segment at the output level. The first spectrum segment is represented in the frequency domain, which only characterizes the pattern of how the characteristics of the front-end transmission path are affected by the physical structure of the space component. It provides an independent feature segment for subsequent inverse compensation and recovery of the time offset extreme point on the frequency domain scale.

[0035] After obtaining the first arrival spectrum segment, the dispersion delay correction and arrival time difference calculation unit 117 executes step S216: extracting the original arrival time from the waveform corresponding to the first arrival time window, and extracting the structural dispersion delay correction amount using the frequency domain calibration reference and the first arrival spectrum segment. After correcting the original arrival time with the structural dispersion delay correction amount, the time offset between multiple corrected arrival times is analyzed to obtain the target arrival time difference. The inputs to this step are the first arrival time window, the first arrival spectrum segment, the frequency domain calibration reference, and the fixed delay of the synchronization link for each acquisition channel, and the output is the target arrival time difference component for each non-reference channel.

[0036] Regarding the extraction of the original arrival time: The discrete voltage digital array sequence within the first arrival time window is extracted from the pre-processing steps. A sliding time aperture length constant is set in the computing memory and assigned a value of 2. At the starting position of the discrete time coordinate axis corresponding to the first arrival time window, a local truncation data frame is established extending backward with the sliding time aperture length constant as the span. For each voltage floating-point real number falling within the range of this local truncation data frame, a square power operation is performed to extract the voltage square amplitude parameter. All voltage square amplitude parameters within the range of the sliding time aperture length constant are summed. As the local truncation data frame moves backward point by point along the discrete time coordinate axis within the first arrival time window, the corresponding accumulated summed real number result is continuously output at each step clock node. These accumulated summed real number results, arranged continuously in chronological order, are merged and spliced ​​to form the sliding energy integral curve. Extract the continuously discrete numerical values ​​from the sliding energy integral curve, perform differential differentiation along the horizontal axis of the time variable to adjacent points, and output a one-dimensional first-order derivative sequence. Perform the same differential differentiation again on the first-order derivative sequence to output a one-dimensional second-order derivative sequence. Within the time span bounded by the first arrival time window, traverse and scan from the start point to the end point to find specific data array index position mapping points in the second-order derivative sequence where the numerical values ​​undergo a zero-crossing transition from positive real numbers to negative real numbers, and the positive amplitude of the position where the transition occurs reaches the global absolute maximum value. Extract the absolute timestamp floating-point real number scale value associated with and aligned with the data array index position mapping point, and directly assign it as the original arrival time.

[0037] Regarding the extraction of structural dispersion delay correction, the current spectral centroid of the first-reach spectrum segment is extracted. Specifically, the absolute value parameters of complex characteristic amplitudes distributed at each frequency calculation step point in the first-reach spectrum segment are extracted. For each parameter, a square root calculation is performed, and the obtained square constant is set as the real value of the voltage square spectral density distribution characteristic reflecting the energy concentration density of a physical point at a specific frequency. The lower limit constant 300 and the upper limit constant 1500 of the effective frequency band are obtained. Within the continuous frequency value closed interval enclosed by constants 300 and 1500, the real value of the frequency horizontal axis step is compared with the real value of the voltage square spectral density distribution characteristic at the same position coordinate. The values ​​are subjected to bitwise scalar multiplication. For all single-step scalar multiplication results within the defined closed interval, the floating-point real numbers are continuously summed along the horizontal axis of frequency variation using definite integrals to obtain the first definite integral characteristic scalar result used as the numerator parameter. Within the same closed interval, all voltage squared spectral density distribution characteristic real values ​​are directly extracted, and independent definite integrals are continuously summed along the horizontal axis of frequency variation to obtain the second definite integral characteristic scalar result used as the denominator parameter scaling constraint. The first definite integral characteristic scalar result is divided by the second definite integral characteristic scalar result to obtain the current spectral centroid. ; in, The real-valued characteristic of the voltage squared spectral density distribution of the first spectral segment. This represents the current spectral centroid. The current spectral centroid physically represents the coordinates of the geometric center of the first effective energy frequency distribution of the partial discharge composite wavefront inside the coaxial cavity.

[0038] Furthermore, based on the mapping relationship between the current spectral centroid, the calibration spectral centroid, and the calibration group delay gradient, a structural dispersion delay correction amount is generated to compensate for signal structure propagation distortion. Specifically, the system retrieves the floating-point real frequency constant of the calibration spectrum centroid and the corresponding floating-point real slope constant of the calibration group delay gradient, which were independently mapped and stored in the early calibration stage of the current acquisition channel, from the system's underlying parameter calibration matrix table. It then extracts the floating-point real frequency constant value of the current spectrum centroid, which was output by the dynamic real-time integration calculation in the previous step. The value of the current spectrum centroid and the value of the calibration spectrum centroid are loaded into the subtraction processor for scalar subtraction, outputting a real value of the frequency offset difference with positive or negative numerical polarity. This real value of the frequency offset difference is then multiplied by the floating-point real slope constant of the calibration group delay gradient. The transverse deflection of the energy centroid caused by the high-order propagation mode conversion and differentiation attenuation of the high-frequency leading wavefront of partial discharge within the gas-insulated cavity due to obstruction by a solid insulating partition is linearly mapped to the longitudinal physical time advance or lag deviation extreme value on the absolute optical alignment time coordinate axis using the prior-obtained slope attribute of the inherent group delay discrete change of the propagation medium. This generates the structural dispersion delay correction quantity. ; in, It is a structural dispersion delay correction quantity, with the dimension of time, used to compensate for the arrival time deviation caused by dispersion when the signal propagates through the internal structure of the equipment.

[0039] In terms of deriving the target arrival time difference: the fixed delay of the synchronization link measured by the optical link loopback is combined with the structural dispersion delay correction to compensate for the original arrival time, thereby generating the node arrival time corresponding to multiple UHF sensors one by one. Specifically, the real constant value of the fixed delay of the synchronous link, unique to the current acquisition channel, is retrieved from the optical network time matching mapping register. The floating-point real time value corresponding to the original arrival time is directly subtracted from the real constant value of the fixed delay of the synchronous link, performing the first subtraction operation to offset the static system constant time scale delay deviation caused by deducting the fixed length delay of the optical fiber transmission path and the inherent start-up trigger electrical lag of the high-speed analog-to-digital conversion circuit. Based on the remaining real result of the first subtraction operation, the floating-point real value parameter corresponding to the structural dispersion delay correction is subtracted, performing the second subtraction operation. The initial observation timestamp, which is recorded by the combined distortion of multipath superposition contamination of the physical components of the equipment and the conversion delay of the underlying optoelectronic network, is reverse-derived and mapped back to the theoretical starting time scale point where the real electromagnetic physical wavefront initially touches the spatial detector phase surface in the ideal homogenized propagation medium, and the node arrival time is registered and generated. ; in, The original arrival time, For the synchronization link, a fixed delay is required. This represents the arrival time of the node.

[0040] Furthermore, the time offsets between the arrival times of multiple nodes are analyzed to obtain the target arrival time difference. Specifically, the floating-point real-valued arrival times of all independently generated nodes participating in monitoring this discharge event are extracted. A linear ascending sort scan is performed in ascending order from smallest to largest within the system memory core time sorting array module. The node that is first in the sorted results—that is, the node with the earliest signal sensing and recording action relative to the overall spatial reference coordinates in the absolute time series of the outbreak—and has the smallest timestamp reading, is designated as the reference node. The floating-point real-valued arrival time of this reference node, bound to it, is extracted as the fixed value for performing the global distance offset differential comparison subtraction operation. The subtraction reference constant; for all other non-reference acquisition nodes, the main control scheduler initiates a loop traversal retrieval action, extracting the floating-point real value of the node arrival time bound to each node as the minuend, and sequentially sending it to the differential subtraction operation comparison core array, subtracting the floating-point real value of the reference node's node arrival time respectively, outputting a floating-point real value difference array representing the discrete relative time distance offset lag interval caused by the UHF electromagnetic energy space radio frequency wave array reaching the measured environment at different geometric positions of each acquisition end, performing multi-dimensional packaging and merging and assigning values ​​to obtain the target arrival time difference: ; in, The arrival times of nodes that are not reference nodes. The arrival time of the reference node. This is the component in the time difference of arrival of the target that corresponds to the non-reference node.

[0041] After obtaining the target arrival time difference, the constraint path coordinate calculation and surface mapping unit 118 executes step S217 to obtain candidate discharge coordinates based on the internal propagation constraint path of the device, which is constrained by the propagable space and the medium boundary. The internal propagation constraint path of the device refers to the propagation path within the three-dimensional structural coordinate system of the device, constrained by the propagable space and the medium boundary, and bypassing the non-propagable region. If the spatial calculation is performed using straight-line distance instead of the internal propagation constraint path of the device, the theoretical propagation time difference will not match the actual propagation process of the signal inside the device. The calculated coordinates are likely to fall inside the metal conductor or the non-propagable region, making it difficult to correspond to the equipment components available for maintenance.

[0042] Specifically, in the three-dimensional structural coordinate system of the equipment, a coordinate grid resolution iteration constant is set and fixed at 0.01. A continuous three-dimensional grid discretization traversal search point set is generated within the defined closed three-dimensional bounding box of the accessible space filled with insulating gas. Any uncalculated independent three-dimensional grid intersection point is extracted from this set and mapped and substituted as the current space point to be evaluated. The space obstacle avoidance shortest path pathfinding component loaded with the coordinate data of the physical space obstacle avoidance boundary is called. Taking the space point to be evaluated as the starting position, the phase center extending from the starting position to each UHF sensor is calculated. The total length of the shortest effective geometric connection polyline between coordinate points; during the trajectory point calculation process, the control engine program's judgment rules implement spatial boundary interception, forcing the polyline to bypass the three-dimensional insurmountable shielding and impenetrable electromagnetic wave medium distribution area occupied by the solid high-voltage internal conductive shaft core and the grounded metal flange shielding body. The real cumulative scalar calculation result of the path length along the surface of the continuous physical isomorphic ultra-high frequency electromagnetic wave topology connection propagation detour natural propagation in the insulating gas channel inside the equipment's cavity, after going through tortuous obstacle avoidance and path finding, is used as the internal propagation constraint path length of the equipment.

[0043] By independently dividing the calculated internal propagation constraint path length value for each acquisition channel by a pre-given wave velocity reference constant, the theoretical one-sided time delay of the electromagnetic pulse traveling alone from the spatial point to be evaluated to the corresponding channel coordinate point is directly obtained: ; in, For the current spatial point to be evaluated, From the spatial point to be evaluated to the first The internal propagation constraint path length of the phase center of an ultra-high frequency sensor. The wave velocity reference constant is a reference constant value that characterizes the average propagation rate of the equivalent propagation rate of the leading edge wave packet of the UHF electromagnetic signal body inside the coaxial shielded tube. In this embodiment, it is taken as 295000000.

[0044] For each non-reference receiving channel, the independently calculated one-sided theoretical delay is subtracted from the one-sided theoretical delay calculated by the reference channel to derive the theoretical propagation time difference encompassing the multi-dimensional pairing comparison components: ; in, The theoretical time delay of the reference channel is 1 / 2. For the theoretical propagation time difference and the first The component corresponding to each non-reference receiving channel.

[0045] In the core processor unit for optimization matching, the theoretical propagation time difference calculated for each channel is compared with the target arrival time difference extracted from the actual measurement under the same channel. The comparison module performs a bilateral scalar subtraction comparison to obtain a single-dimensional residual data item entity that measures the fluctuation of the difference between the measured and predicted values. For each single-dimensional residual data item entity, the quadratic power calculation unit independently performs its own square product divergence operation. The square divergence results of all channel branches are summarized and merged, then fed into the addition operator for arithmetic horizontal summation and convergence calculation. The output is the residual square sum real number representing the overall positioning error cost index level of the current spatial point to be evaluated. ; in, This represents the total number of non-reference receiving channels. This represents the sum of squared residuals corresponding to the spatial points to be evaluated.

[0046] Finally, the automatic scheduling scanning module is activated to continuously traverse and deduce the discretized three-dimensional mesh, search for all remaining coordinate meshes within the spatial set of search points, calculate the residual sum of squares corresponding to the mesh points, and store them uniformly in the memory comparison sequence array stack to perform real-time dynamic extreme value comparison and refresh, replacing the size sorting and arrangement. After completing the mesh point exploration, calculation, comparison and retrieval process of all limited effective finite closed gas volume spatial regions, the coordinate landing point that makes the residual sum of squares reach the global minimum value is found, and its horizontal and vertical axis three-dimensional coordinate space reference offset parameter real number array is fixed as a whole to generate candidate discharge coordinates.

[0047] After obtaining the candidate discharge coordinates, the constraint path coordinate calculation and surface mapping unit 118 executes the mapping process in step S218 to map the candidate discharge coordinates to the physical surface of the device that meets the conditions for discharge to occur, thereby obtaining the actual discharge physical coordinates. The positioning result output unit 119 then outputs the local discharge source positioning result containing the actual discharge physical coordinates. The mapping process includes two steps: spatial coordinate comparison and reverse search projection.

[0048] Spatial coordinate comparison step: The three-dimensional spatial lattice floating-point coordinate array output from the previous calculation is obtained as candidate discharge coordinates. The spatial geometric envelope surface and spatial volume set that meet the conditions for partial discharge and conform to the physical defect generation mechanism are retrieved from the three-dimensional structural coordinate system data structure of the equipment. This set of regions with partial discharge conditions is defined as the set of regions with partial discharge conditions. Specifically, this set of regions with partial discharge conditions includes the three-dimensional coordinate group of the three-dielectric joint edge of the basin insulator, the three-dimensional coordinate group of the insulator surface, the three-dimensional coordinate group of the area on the surface of the internal high-voltage conductor suspected of having burrs, the three-dimensional coordinate group near the physical break of the disconnecting switch, the three-dimensional coordinate group of the chamber section at the bottom of the busbar drum where there may be floating metal particles, and the three-dimensional coordinate group of the stress concentration area of ​​the cable terminal. The vector parameter composed of the real numbers of the horizontal, vertical, and triangular coordinates of the candidate discharge coordinates is used to perform spatial Euclidean linear distance difference calculation and position comparison with the coordinates of all legal 3D mesh nodes contained within the region set with partial discharge conditions. A spatial fitting tolerance constant with a value of 0.005 is set: when the absolute value of the spatial Euclidean linear distance between the candidate discharge coordinates and any legal 3D mesh node in the region set is less than or equal to the spatial fitting tolerance constant, it is determined that the coordinate point hits the legal discharge entity position, and the candidate discharge coordinates are used as the landing point to directly enter the subsequent binding process; otherwise, if the absolute value of the spatial Euclidean linear distance between the candidate discharge coordinates and all mesh nodes in the region set is greater than the spatial fitting tolerance constant, it is determined that the candidate discharge coordinates are in the space of the non-discharge entity region, and the reverse search projection stage is entered.

[0049] In the reverse search projection stage: when the comparison determines that the candidate discharge coordinates fall inside the envelope boundary of the high-voltage current-carrying center conductive cylinder (i.e., inside the metal conductor), or fall outside the outline boundary of the physical gas chamber isolated and enclosed by the grounded metal shell (i.e., outside the closed gas chamber), the coordinate boundary out-of-bounds anomaly handling logic program is triggered. Specifically, the discrete spatial polyline three-dimensional coordinate point array data sequence generated in the previous extreme value optimization step, which points from the candidate discharge coordinates to the phase center of each UHF sensor, is extracted as the internal propagation constraint path of the device; taking the erroneous landing point in the three-dimensional space where the candidate discharge coordinates are located as the starting point of physical tracing, and following the direction of physical propagation delay backtracking, the point array sequence of each internal propagation constraint path of the device is traced in reverse, and the progressive reverse step displacement and position sampling calculation is performed with a fixed spatial step constant of 0.01; at each discrete sampling coordinate point that is backtracked in reverse, the current point coordinates are extracted in real time and the spatial geometric surface collision detection operator module is called to determine whether the sampling coordinate point forms a geometric tangential intersection with the legal entity surface defined by the set of regions with partial discharge conditions or whether the point penetrates and hits. Once the collision detection operator module returns the logical truth response result indicating intersection at a specific spatial reverse stepping node, the subsequent backtracking stepping loop in that path direction is stopped. The three-dimensional real-valued landing point coordinates of the horizontal, vertical, and longitudinal coordinates of this intersection position are extracted. For the coordinate landing points of multiple successful intersection results generated by multiple acquisition channels, the cumulative displacement scalar length real value of each return along the spatial polyline is calculated. The three-dimensional landing point coordinates corresponding to the smallest value among the cumulative displacement scalar length real values ​​are selected and determined as the projection position on the physical surface of the device that satisfies the shortest distance constraint. To avoid confusion between the mesh layout iteration in the coordinate solution stage and the reverse backtracking stepping in the surface mapping stage, the fixed spatial stepping constant and the coordinate mesh resolution iteration constant are both 0.01, but are defined and assigned values ​​independently.

[0050] Furthermore, the candidate discharge coordinates or projected position coordinates are bound to the corresponding equipment segment and radial orientation to generate the actual discharge physical coordinates. Specifically, the candidate discharge coordinates, which are used as the landing point when the spatial coordinate comparison is successful, or the projected position coordinates obtained by projecting them onto the physical surface of the equipment that meets the shortest distance constraint when the spatial coordinate comparison is unsuccessful, are loaded into the drawing structure index dictionary matrix associated with the equipment's three-dimensional structural coordinate system. Based on the inclusion-attribution mapping algorithm relationship of the spatial coordinates, the geometric bounding box entity region with a named number that the coordinate point falls into is found, and the name information of the physical component corresponding to the region is extracted as the equipment segment. Taking the central axis of the main generatrix of the equipment segment as the origin of the polar coordinate circle, the real value of the circumferential angle formed between the tangential line connecting the landing point in the cross section plane and the reference vertical upward zero-degree central axis is calculated and assigned as the radial orientation. The equipment segment, radial orientation, and the real values ​​of the horizontal coordinate, vertical coordinate, and vertical axis coordinate of the landing point are combined and formatted into a packet to generate the actual discharge physical coordinates, so that the positioning result is located on the physical surface of the equipment that meets the conditions for discharge to occur.

[0051] In one implementation, the positioning result output unit 119 converts the actual discharge physical coordinates into the drawing standard name and three-dimensional spatial coordinate data of the associated equipment section: it independently extracts the data array set of floating real numbers of horizontal geometric coordinates, floating real numbers of vertical azimuth coordinates, and floating real numbers of vertical elevation axis coordinates from the actual discharge physical coordinates, and formats it as three-dimensional spatial coordinate data; it retrieves the manufacturer's basic design database pre-loaded in the system's read-only memory, extracts the scalar parameter of the equipment section number attached to and bound inside the actual discharge physical coordinates as the retrieval input key, performs a traversal query, and after a successful match, reads the statically associated continuous Chinese character sequence entity describing the physical name of the physical asset, and defines it as the drawing standard name. Furthermore, the relative axial mileage is calculated using the geometric starting point of the equipment section where the discharge occurs as the reference zero point: the three-dimensional spatial starting endpoint coordinate vector at the geometric center of the flange docking section along the extension direction of the main busbar metal cavity of the equipment section is extracted and statically fixed as the reference zero point; the target landing point matrix coordinate vector corresponding to the three-dimensional spatial coordinate data is obtained, and vector subtraction is performed with the reference zero point to generate a three-dimensional offset vector entity; the three-dimensional spatial direction unit reference vector corresponding to the direction of the high voltage conductor axis inside the equipment section is extracted, and the three-dimensional offset vector entity and the three-dimensional spatial direction unit reference vector are input into the inner product calculation logic operator module to perform a dot product operation; the output projection length floating-point real number calculation result is set as the relative axial mileage, and the numerical precision is forcibly rounded to 3 decimal places. Finally, the standard drawing name, relative axial mileage, and actual discharge physical coordinates are formatted and spliced ​​together: a one-dimensional multi-type data storage sequence structure is instantiated in the data packet output buffer, and the continuous Chinese character sequence entity of the standard drawing name, the constant floating-point real number result of the relative axial mileage, and the three independent real number component array data of the horizontal, vertical, and horizontal coordinates of the actual discharge physical coordinates are pushed in sequentially according to the hierarchical progression from macroscopic orientation to microscopic coordinates; the character stream and data stream are spliced ​​together and the basic operator component performs seamless serial splicing and calculation encapsulation to form a coherent combined sequence packet data body, which is sent out through the system external communication interface bus protocol or the graphic digital display interface window module to output the partial discharge power supply positioning result.

[0052] Example 1:

[0053] The following is an example of inputting complete numerical values.

[0054] Assuming the total length of the gas-insulated enclosure under test is 30, the radius constant of the outer surface of the inner conductor is taken as 0.06, falling within the range of 0.05 to 0.08, and the radius constant of the inner surface of the metal shell is taken as 0.3, falling within the range of 0.2 to 0.4. Using the geometric center of the flange connection section at one end of the equipment as the three-dimensional absolute zero point, with a distance measurement accuracy of 0.01, four UHF sensors are arranged along the central axis of the main busbar, reaching the minimum deployment threshold of four. These are designated as UHF sensor 1 to UHF sensor 4. Their phase centers are projected onto the corresponding installed port section via the minimum Euclidean linear distance normal, with axial mileages of 6, 8, 10, and 12, and radial azimuths of 0 degrees, 90 degrees, 180 degrees, and 270 degrees, respectively. The propagation space and medium boundary of the metal cavity are extracted and encapsulated along with the four phase centers to form a structural coordinate table for the equipment, serving as the equipment's three-dimensional structural coordinate system. Four acquisition nodes, each corresponding to one of the four UHF sensors, are connected to the synchronization time master station in a star topology via quartz glass optical fiber. After loopback bidirectional delay calculation and time delay compensation, the absolute time base deviation between the hardware trigger timestamp of each acquisition node and the synchronization time master station is stabilized at an absolute value less than or equal to 1, thus achieving optical synchronization locking.

[0055] Calibration Phase: With the master station locked in sync time, the external pulse source is pressed onto the external monitoring medium window via a mechanically limited coupling clamp. The rise time constant of the broadband verification pulse is set to 0.8 and the amplitude to 80, both within the specified range, and the amplitude is lower than the actual discharge initiation trigger threshold. The injection time interval between two adjacent broadband verification pulses is 20,000,000, and the transmission frequency constant is 50. 100 injections are performed continuously, and the background is continuously scanned during the injection to eliminate crosstalk from the actual discharge pulses. Each acquisition node latches the calibration waveform. The DC bias parameter is obtained by taking the arithmetic mean of 5,000 sampling points in the background noise segment before triggering. The DC bias parameter is subtracted from each of the 20,000 sampling points of the calibration waveform to extract the calibration impulse response. The fixed delay of the synchronization link measured by the optical link loopback is recorded synchronously. Taking the first acquisition channel as an example, the fixed delay of the synchronization link is measured to be 0.5. A Discrete Fourier Transform (DFT) is performed on the calibration impulse response within the 300 MHz to 1500 MHz range. Taking the first acquisition channel as an example, the calibration spectrum centroid is 90°, and the calibration group delay gradient is 0.00001. The second to fourth acquisition channels are processed using the same procedure, calculating their respective calibration spectrum centroids, calibration group delay gradients, and synchronization link fixed delays. The calibration spectrum centroids and calibration group delay gradients from all channels are concatenated and encapsulated to form a frequency domain calibration reference.

[0056] Monitoring Phase: The system continuously acquires data at a sampling frequency of 500,000,000 times per second. The analog radio frequency signal detection bandwidth is limited to between 300 MHz and 1500 MHz. A circular queue of 20,000 floating-point units is opened for rolling over each channel. When the real-time absolute value of the voltage on the monitoring link of the second UHF sensor exceeds 15, it is determined that the partial discharge pulse triggering condition is met. The second acquisition node broadcasts a global hardware freeze command through the fiber optic bus, records for 2000 sampling cycles, and then forcibly blocks the writing, latching the synchronous buffer data containing the current discharge event. The system reads 20,000 floating-point amplitude values ​​from each channel in a first-in-first-out manner to form a voltage time series. The hardware trigger timestamp and sampling interval parameter 0.2 are extracted. The voltage time series containing the background noise segment before triggering, the first arrival waveform segment, and the first reflection segment are packaged into an event data packet to obtain the original UHF waveform.

[0057] Waveform processing stage: The DC bias parameter is obtained by taking the arithmetic mean of 5000 sampling points of the background noise segment before triggering. The DC bias parameter is subtracted point by point from the original 20000 sampling points of the UHF waveform. Zero-phase digital bandpass filtering is performed by convolving the 4th order Butterworth digital bandpass filter twice with 300 MHz and 1500 MHz as the boundary. The first energy rise point is identified by sliding the energy integral with a step size of 10 sampling points and making the second derivative extremum. The second local extremum with an interval of more than 20 sampling points after the first direct energy peak in the calibrated impulse response is the time corresponding to the first reflection peak. A clean one-dimensional sequence with a duration of 6 is extracted as the first arrival time window, and its end time does not exceed the time corresponding to the first reflection peak. The waveform data in the first arrival time window is padded with zeros to 1024 points and a fast Fourier transform is performed to obtain the first arrival spectrum segment.

[0058] Time difference correction stage: Using the sliding time aperture length constant 2 as the constant, the sliding energy integral curve is obtained by sliding the square of the voltage within the first arrival time window. After second-order difference, the extremum of the second derivative, which crosses zero positively and has the largest global positive amplitude, is taken, and its timestamp is assigned as the original arrival time; taking the first acquisition channel as an example, the original arrival time is 1200. The current spectral centroid of the first arrival spectrum segment is extracted; taking the first acquisition channel as an example, the current spectral centroid is 920. The structural dispersion delay correction amount is generated: the current spectral centroid 920 is subtracted from the calibration spectral centroid 900, and the difference 20 is multiplied by the calibration group delay gradient 0.00001, resulting in a structural dispersion delay correction amount of 0.0002 for the first acquisition channel. Subtracting the fixed delay of the synchronization link (0.5) and the structural dispersion delay correction (0.0002) from the original arrival time of 1200, we obtain the node arrival time of the first acquisition channel as 1199.4998. The node arrival times of the second to fourth acquisition channels are calculated using the same process. All node arrival times are sorted in ascending order, and the acquisition channel corresponding to the smallest timestamp is anchored as the reference node. Let the first acquisition channel be the reference node, with an arrival time of 1199.4998. Subtracting the arrival time of the reference node from the arrival time of each non-reference node yields the target arrival time difference. The node arrival time of the second acquisition channel is 1199.6000, with a corresponding component of 0.1002; the node arrival time of the third acquisition channel is 1199.8000, with a corresponding component of 0.3002; and the node arrival time of the fourth acquisition channel is 1199.9000, with a corresponding component of 0.4002.

[0059] Coordinate calculation stage: Points are placed within the passable bounding box in the three-dimensional structural coordinate system of the device with a coordinate grid resolution iteration constant of 0.01. The internal propagation constraint path lengths from a certain spatial point to be evaluated to the phase center of the first UHF sensor to the fourth UHF sensor are calculated as 8, 8.03, 8.09, and 8.12 respectively after obstacle avoidance and pathfinding. These are divided by the wave velocity reference constant 295000000 and converted to nanoseconds, respectively, to obtain the theoretical one-sided time delays of each channel as 27.1186, 27.2203, 27.4237, and 27.5254. The theoretical propagation time difference components obtained by differential analysis with the first channel as the reference channel are 0.1017, 0.3051, and 0.4068 respectively. Subtracting the theoretical propagation time difference components from the target arrival time difference components (0.1002, 0.3002, 0.4002) and summing the squares, the residual sum of squares for the spatial point to be evaluated is 0.0000698. For comparison, for another spatial point to be evaluated, whose internal propagation constraint path lengths are 8.5, 8.44, 8.60, and 8.90 respectively, the residual sum of squares calculated using the same procedure is 1.0070, which is greater than the residual sum of squares for the previous spatial point. After traversing all grid points, the spatial point whose residual sum of squares reaches the global minimum is fixed as the candidate discharge coordinate.

[0060] Surface mapping and output stage: The candidate discharge coordinates are compared with the set of regions with partial discharge conditions. Let the three-dimensional coordinates of the candidate discharge coordinates be (9.000, 0.030, 0.100). The spatial Euclidean distance between the candidate discharge coordinates and a legal three-dimensional grid node in the three-dimensional coordinate group of the three-dielectric junction edge of the basin insulator is 0.003, which is less than the spatial fit tolerance constant of 0.005. It is determined that the legal discharge entity position is hit, and the candidate discharge coordinates are used as the landing point to generate the actual discharge physical coordinates. If the comparison result is not hit, the device backtracks along the internal propagation constraint path of the candidate discharge coordinates to each UHF sensor with a fixed spatial step constant of 0.01. The collision detection stops when the legal entity surface is hit for the first time. The multi-channel landing point with the smallest cumulative backtracking displacement is selected and projected onto the physical surface of the device that meets the shortest distance constraint condition to generate the actual discharge physical coordinates. Finally, the actual discharge physical coordinates are converted into the standard drawing name and three-dimensional spatial coordinate data of the associated equipment section. The geometric starting point coordinates of the equipment section where the discharge occurred are set to (8.000, 0, 0). Using this as the reference zero point, the offset vector of the target landing point relative to the reference zero point is (1.000, 0.030, 0.100). Projecting this vector onto the unit vector (1, 0, 0) of the high-voltage conductor axis at the center of the equipment section yields a relative axial mileage of 1.000. The numerical code of the associated equipment section is extracted and retrieved from the manufacturer's basic design database to obtain the standard drawing name. The standard drawing name, relative axial mileage, and actual discharge physical coordinates are then formatted and spliced ​​together to output the partial discharge location result, completing the entire process of this partial discharge location.

[0061] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and all such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A UHF partial discharge positioning system based on optical synchronous time difference of arrival, applied to a gas-insulated enclosure equipped with multiple UHF sensors that synchronously acquire data under a unified time reference, characterized in that, Configured for execution: A three-dimensional structural coordinate system for the sensing array and the device is established based on multiple ultra-high frequency sensors. The three-dimensional structural coordinate system of the device represents the propagable space and medium boundary inside the gas-insulated sealed device. A calibration excitation is applied to the gas-insulated enclosure to obtain a frequency domain calibration reference generated by the calibration excitation and characterizing the dispersion delay characteristics caused by the internal structure of the gas-insulated enclosure. When the partial discharge pulse triggering condition is met, the original ultra-high frequency waveform is acquired through the sensor array. Extract the first-reach spectrum segment of the original UHF waveform; The original arrival time of the original UHF waveform is extracted, and the structural dispersion delay correction is extracted using the frequency domain calibration reference and the first arrival spectrum segment. The original arrival time is corrected with the structural dispersion delay correction to obtain the target arrival time difference. Candidate discharge coordinates are obtained based on the internal propagation constraint path of the device, which is constrained by the propagable space and the medium boundary. The candidate discharge coordinates are mapped to the physical surface of the device that meets the conditions for discharge to occur, thus obtaining the actual discharge physical coordinates. The output includes the localization result of the partial discharge source, which contains the actual physical coordinates of the discharge.

2. The UHF partial discharge positioning system based on optical synchronous time difference of arrival as described in claim 1, characterized in that, Establishing a sensing array based on multiple ultra-high frequency sensors includes: Each of the multiple UHF sensors corresponds to a data acquisition node, and each data acquisition node is connected to a synchronization time master station via optical fiber to complete optical synchronization locking.

3. The UHF partial discharge positioning system based on optical synchronous time difference of arrival as described in claim 1, characterized in that, Establishing a three-dimensional structural coordinate system for the device, characterizing the propagable space and medium boundary within the gas-insulated enclosure, includes: The manufacturing drawings of the gas-insulated enclosure are converted into three-dimensional spatial coordinate data, and a device structure coordinate table is generated describing the propagation space of the metal cavity, the medium boundary, and the phase center of the multiple ultra-high frequency sensors, which serves as the three-dimensional structural coordinate system of the device.

4. The UHF partial discharge positioning system based on optical synchronous time difference of arrival as described in claim 1, characterized in that, The verification excitation is a broadband verification pulse. Obtaining the frequency domain calibration reference includes: recording the calibration waveform and extracting the calibration impulse response from it, and extracting the calibration spectrum centroid and calibration group delay gradient of the calibration impulse response as the frequency domain calibration reference.

5. The UHF partial discharge positioning system based on optical synchronous time difference of arrival as described in claim 1, characterized in that, Extracting the first-arrival spectral segment of the original UHF waveform includes: The first arrival time window is extracted from the energy rise range of the original UHF waveform, and frequency domain transformation is performed on the waveform within the first arrival time window to obtain the first arrival spectrum segment.

6. The UHF partial discharge positioning system based on optical synchronous time difference of arrival as described in claim 4, characterized in that, The frequency domain calibration reference is used to extract the structural dispersion delay correction: Extract the current spectral centroid of the first-arrival spectral segment, and based on the mapping relationship between the current spectral centroid, the calibration spectral centroid, and the calibration group delay gradient, generate the structural dispersion delay correction amount used to compensate for signal structure propagation distortion.

7. The UHF partial discharge positioning system based on optical synchronous time difference of arrival as described in claim 1, characterized in that, The original arrival time is corrected using the structural dispersion delay correction amount to obtain the target arrival time difference, including: The original arrival time is compensated by combining the fixed delay of the synchronization link measured by the optical link loopback with the structure dispersion delay correction amount, generating the node arrival time corresponding to each of the multiple ultra-high frequency sensors, analyzing the time offset between the multiple node arrival times, and obtaining the target arrival time difference.

8. The UHF partial discharge positioning system based on optical synchronous time difference of arrival as described in claim 1, characterized in that, The candidate discharge coordinates are obtained based on the internal propagation constraint path of the device, which is constrained by the propagable space and the medium boundary, including: The spatial points in three-dimensional space that minimize the sum of squared residuals between the target arrival time difference and the theoretical propagation time difference are used as the candidate discharge coordinates. The theoretical propagation time difference is determined based on the propagation constraint path length and wave speed reference constant inside the device.

9. The UHF partial discharge positioning system based on optical synchronous time difference of arrival as described in claim 1, characterized in that, Mapping the candidate discharge coordinates to the physical surface of a device that meets the conditions for discharge occurrence includes: The candidate discharge coordinates are compared with the set of regions with partial discharge conditions set in the three-dimensional structural coordinate system of the device. When the candidate discharge coordinates do not match the set of regions with partial discharge conditions, the device propagation constraint path from the candidate discharge coordinates to each of the UHF sensors is reversed and projected onto the physical surface of the device that satisfies the shortest distance constraint condition to obtain the actual discharge physical coordinates.

Images