An acoustic-electric prpd fusion cable partial discharge evaluation decision method and system

Abstract

The application provides a kind of acoustic-electric PRPD fusion cable partial discharge evaluation decision method and system, it is related to partial discharge technical field, including: obtaining power frequency voltage reference signal, high-frequency electromagnetic pulse signal and ultrasonic pulse signal, according to phase scale registration acoustic-electric phase point, extract peak value, width and center frequency, determine acoustic-electric homologous event according to phase difference, amplitude order and width coincidence, construct joint phase atlas, input deep neural network output defect confidence value and severity score, determine treatment level in combination with continuous cycle number.The application, by unifying phase reference and homologous screening, includes two types of sensing signals in the same judgment chain, weakens electromagnetic interference, mechanical vibration and heterogenous pulse influence, maintains phase, amplitude, pulse width, frequency and time coincidence relationship, improves defect recognition and severity evaluation consistency, makes inspection cycle adjustment, live retest and outage maintenance have continuous event basis.

Description

Technical Field

[0001] This invention relates to the field of partial discharge technology, and in particular to a cable partial discharge assessment and decision-making method and system that integrates acoustic and electrical partial discharge detection (PRPD). Background Technology

[0002] The field of partial discharge technology involves online monitoring of cable joint insulation defects, pulse signal acquisition, phase spectrum construction, and status determination. It typically uses high-frequency current sensors, ultrasonic sensors, and power frequency voltage reference signals to acquire electromagnetic pulses, ultrasonic pulses, and phase references. Based on the pulse phase, amplitude, rise time, pulse width, and center frequency, a partial discharge detection (PRPD) spectrum or time-frequency spectrum is generated. Traditional cable partial discharge assessment and decision-making refers to the identification and maintenance determination of the partial discharge status of cable joints. This usually employs high-frequency current methods, ultra-high-frequency methods, and ultrasonic methods to acquire single-path signals. Based on PRPD spectra, S-transform time-frequency spectra, statistical characteristics, convolutional neural networks, long short-term memory networks, or expert experience, defect type identification, severity scoring, inspection cycle determination, live-line retesting, and power outage maintenance plan matching are completed.

[0003] Current cable partial discharge assessment methods mostly involve collecting and interpreting signals along a single path. High-frequency current method, ultra-high frequency method, and ultrasonic method each generate phase spectrum or time spectrum. Electromagnetic interference, mechanical vibration, and propagation attenuation are easily mixed into the criteria. The pulse source is difficult to verify within the same power frequency cycle. Amplitude, pulse width, frequency, and phase information are used separately. Defect type identification is easily affected by empirical thresholds and sample bias. There is a lack of continuous event constraints between the severity and operation and maintenance. The overall stability of retesting, inspection, and power outage maintenance decisions is insufficient. Summary of the Invention

[0004] The purpose of this invention is to provide a cable partial discharge assessment and decision-making method and system that integrates acoustic and electrical PRPD. By placing high-frequency electromagnetic pulse signals and ultrasonic pulse signals under a unified power frequency phase reference for feature registration, homogeneity determination, joint spectrum construction, and network diagnosis, and combining the occurrence of homogeneous events within a continuous power frequency cycle to form an operation and maintenance handling level, the cable partial discharge assessment can simultaneously utilize the phase, amplitude, duration, frequency, and time correlation of both acoustic and electrical detection information, providing mutually corroborating judgment criteria for inspection cycle adjustment, live retesting, and power outage maintenance.

[0005] To achieve the above objectives, the present invention provides a cable partial discharge assessment and decision-making method integrating acoustic and electrical partial discharge (PRPD) fusion, comprising the following steps: Acquire power frequency voltage reference signal, high frequency electromagnetic pulse signal and ultrasonic pulse signal, divide power frequency cycles according to the rising zero crossing time of the power frequency voltage reference signal, and divide each power frequency cycle into phase scale; Based on the high-frequency electromagnetic pulse signal and the ultrasonic pulse signal, the pulse peak value, rise point, duration and center frequency are determined respectively, and the electromagnetic phase point and ultrasonic phase point are registered according to the phase scale. Based on the phase difference between the electromagnetic phase point and the ultrasonic phase point, the amplitude sorting position, and the overlap between the duration width corresponding to the electromagnetic phase point and the duration width corresponding to the ultrasonic phase point, an acoustic-electric co-source discharge event is determined, and an acoustic-electric joint phase distribution map is constructed based on the acoustic-electric co-source discharge event. The acoustic-electric combined phase distribution map is input into a deep neural network trained from the sample phase distribution map to determine the defect type confidence value and insulation severity score. The partial discharge defect type is determined according to the largest term in the defect type confidence value. Based on the insulation severity score, the partial discharge defect type, and the number of times the acoustic-electric co-source discharge event occurs in the continuous power frequency cycles, the operation and maintenance handling level is determined, and the target handling instruction is determined and output from the inspection cycle adjustment instruction, the live retest instruction, and the power outage maintenance instruction according to the operation and maintenance handling level.

[0006] The present invention also provides a cable partial discharge assessment and decision system integrating acoustic and electrical PRPD, wherein the system is used to implement the above method: The signal acquisition and calibration module is used to acquire power frequency voltage reference signal, high frequency electromagnetic pulse signal and ultrasonic pulse signal, divide the power frequency cycle according to the rising zero crossing time of the power frequency voltage reference signal, divide each power frequency cycle into phase scale, and send the phase scale to the pulse feature registration module; The pulse feature registration module is used to determine the pulse peak value, rise start point, duration and center frequency based on the high-frequency electromagnetic pulse signal and the ultrasonic pulse signal, respectively, and register the electromagnetic phase point and the ultrasonic phase point according to the phase scale, and send the electromagnetic phase point and the ultrasonic phase point to the homology determination and map construction module. The homogeneity determination and spectrum construction module is used to determine the acoustic-electric homogeneous discharge event based on the phase difference between the electromagnetic phase point and the ultrasonic phase point, the amplitude sorting position, and the overlap state between the continuous width corresponding to the electromagnetic phase point and the continuous width corresponding to the ultrasonic phase point. The module then constructs an acoustic-electric joint phase distribution spectrum based on the acoustic-electric homogeneous discharge event and sends the acoustic-electric joint phase distribution spectrum to the deep network diagnostic module. The deep network diagnostic module is used to input the acoustic-electric combined phase distribution map into a deep neural network trained from the sample phase distribution map, determine the defect type confidence value and insulation severity score, determine the partial discharge defect type according to the largest item in the defect type confidence value, and send the defect type confidence value, the insulation severity score and the partial discharge defect type to the risk assessment and linkage module; The risk assessment and linkage module is used to determine the operation and maintenance handling level based on the insulation severity score, the partial discharge defect type, and the number of times the acoustic-electric co-source discharge event occurs in the continuous power frequency cycle, and to determine and output the target handling instruction from the inspection cycle adjustment instruction, the live retest instruction, and the power outage maintenance instruction according to the operation and maintenance handling level.

[0007] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: In this embodiment of the invention, the zero-crossing rise of the power frequency voltage is used as a unified phase reference. High-frequency electromagnetic pulses and ultrasonic pulses are mapped to the same power frequency cycle phase scale. The acoustic-electric co-source discharge events are screened by combining the phase difference, amplitude sorting position and the continuous width overlap state. The influence of environmental interference and heterogeneous disturbances on the criteria is compressed. The acoustic-electric joint phase distribution map retains the phase clustering, amplitude correlation, frequency response and time overlap relationship. The deep neural network synchronously outputs the defect type confidence value and insulation severity score under the constraint of the sample map. The number of consecutive cycles further constrains the state persistence, so that the inspection cycle adjustment, live retesting and power outage maintenance form a closed-loop decision supported by signal co-source, defect identification and risk level. Attached Figure Description

[0008] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts. Obviously, the drawings described below are merely some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0009] Figure 1 This is a flowchart of the cable partial discharge assessment and decision-making process based on the fusion of acoustic and electrical PRPD in this invention. Figure 2 This is a flowchart of the power frequency phase scale and acoustic-electric pulse feature registration process of the present invention; Figure 3 This is the logic diagram for determining the acoustic-electric co-source discharge event of the present invention; Figure 4 This invention provides a data flow graph for network input feature map generation and processing level determination. Figure 5This is a module collaboration diagram of the cable partial discharge assessment and decision-making system integrating acoustic and electrical PRPD of the present invention. Detailed Implementation

[0010] The following description, in conjunction with the technical solutions, content, and application documents of this invention, further illustrates the possible implementations of this invention. The power frequency voltage reference signal, high-frequency electromagnetic pulse signal, ultrasonic pulse signal, electromagnetic phase point, ultrasonic phase point, compensated ultrasonic phase point, acoustic-electric co-source discharge event, acoustic-electric joint phase distribution spectrum, network input feature map, defect type confidence value, insulation severity score, partial discharge defect type, operation and maintenance handling level, and target handling instruction described in the following embodiments are all determined to have their technical meanings according to the technical solutions described in the application documents. Since the application documents do not provide existing drawings and drawing numbers, the following embodiments refer to… Figures 1 to 5 All of these are newly added figures to illustrate the method flow, local data flow, same-source determination logic, network diagnostic data structure, and system module collaboration relationships; no undocumented entity component labels or structural assembly relationships are added to the figures.

[0011] Example 1

[0012] Please see Figures 1 to 4 This embodiment provides a cable partial discharge assessment and decision-making method integrating acoustic and electrical PRPD (partial discharge pulse detection) technologies. This method uses a power frequency voltage reference signal as a unified phase benchmark, incorporates high-frequency electromagnetic pulse signals and ultrasonic pulse signals into the same detection time window, and sequentially completes phase scale generation, acoustic and electrical pulse feature registration, determination of acoustic and electrical co-source discharge events, construction of an acoustic and electrical joint phase distribution map, deep neural network diagnosis, and output of operation and maintenance (O&M) handling level. The input sources of this method are the power frequency voltage reference signal, high-frequency electromagnetic pulse signal, and ultrasonic pulse signal within the same detection time window; intermediate states include phase scale sequences, electromagnetic phase points, ultrasonic phase points, compensated ultrasonic phase points, co-source candidate pairing records, acoustic and electrical co-source discharge events, and network input feature maps; the output results include defect type confidence values, insulation severity scores, partial discharge defect types, O&M handling levels, and target handling instructions.

[0013] S1 acquires the power frequency voltage reference signal, high-frequency electromagnetic pulse signal, and ultrasonic pulse signal. The power frequency cycle is divided according to the rising zero-crossing time of the power frequency voltage reference signal, and each power frequency cycle is divided into phase scales.

[0014] S1 executes on the same detection time window, including the power frequency voltage reference signal, the high-frequency electromagnetic pulse signal, and the ultrasonic pulse signal. The power frequency voltage reference signal provides a phase reference, while the high-frequency electromagnetic pulse signal and the ultrasonic pulse signal are used for subsequent pulse feature extraction. During execution, the power frequency voltage reference signal is first subjected to amplitude limiting and baseline offset subtraction to ensure that the determination of the zero-crossing moment is not affected by baseline drift. At the same time, sampling time markers are added to the high-frequency electromagnetic pulse signal and the ultrasonic pulse signal according to a unified sampling clock, so that subsequent phase registration, pulse pairing, and continuous power frequency cycle statistics have the same time reference.

[0015] S111: Acquire the power frequency voltage reference signal, high-frequency electromagnetic pulse signal, and ultrasonic pulse signal within the same detection time window. Perform amplitude limiting and baseline offset subtraction on the power frequency voltage reference signal, and add sampling time markers to the high-frequency electromagnetic pulse signal and ultrasonic pulse signal according to a unified sampling clock. After this processing, the power frequency voltage reference signal forms a reference sequence to be identified for zero crossing, and the high-frequency electromagnetic pulse signal and ultrasonic pulse signal form a detection sequence with sampling time markers.

[0016] S112: Based on the voltage polarity change and slope direction of continuous sampling points, select sampling positions that change from negative to positive polarity and have a slope greater than the zero-crossing slope threshold, and determine these sampling positions as rising zero-crossing moments. If continuous sampling points do not simultaneously meet the polarity change and slope direction conditions, then these sampling positions are not registered as rising zero-crossing moments to avoid using local fluctuations caused by noise disturbances as power frequency cycle boundaries.

[0017] S113. A power frequency cycle is determined by the time interval between two adjacent zero-crossing moments, and each power frequency cycle is divided into phase scales according to the equal-interval phase segmentation rule to generate a phase scale sequence. The phase scale sequence is used to receive subsequently registered electromagnetic phase points and ultrasonic phase points, and to provide a unified numbering relationship for the phase distribution alignment between different power frequency cycles.

[0018] S2, based on high-frequency electromagnetic pulse signals and ultrasonic pulse signals, determines the pulse peak value, rise start point, duration and center frequency respectively, and registers the electromagnetic phase point and ultrasonic phase point according to the phase scale.

[0019] S2 takes high-frequency electromagnetic pulse signals and ultrasonic pulse signals with sampling time stamps as inputs, and outputs electromagnetic phase points and ultrasonic phase points. The electromagnetic phase point is associated with at least its corresponding power frequency cycle number, phase scale number, pulse peak value, rise time, duration, and center frequency; the ultrasonic phase point is associated with at least its corresponding power frequency cycle number, phase scale number, pulse peak value, rise time, duration, and center frequency. These fields are all formed from signal segments within the same detection time window and are used in subsequent source pairing and network input feature map generation.

[0020] S211: Obtain the envelope sequences of the high-frequency electromagnetic pulse signal and the ultrasonic pulse signal within the same detection time window. Filter candidate pulse intervals based on the channel noise baseline and the pulse trigger threshold, and determine the sampling point with the largest amplitude within each candidate pulse interval as the pulse peak value. The channel noise baselines are derived from the high-frequency electromagnetic detection channel and the ultrasonic detection channel, respectively. The pulse trigger threshold is determined according to the average background amplitude and the background amplitude fluctuation range of the corresponding channel. The high-frequency electromagnetic detection channel and the ultrasonic detection channel each form their own candidate pulse intervals, and the two types of signals are not mixed during the triggering phase.

[0021] S212, the sampling positions that are continuously below the starting point judgment threshold before the pulse peak are retrieved in reverse time. The sampling position closest to the pulse peak and meeting the condition of continuous low amplitude is determined as the rising starting point, and the duration is determined based on the time interval between the rising starting point and the falling ending point. The falling ending point is the position where the pulse amplitude in the candidate pulse interval falls back to the termination judgment state. The duration is used to subsequently determine the overlap of the duration of the electromagnetic phase point and the ultrasonic phase point.

[0022] S213: Perform frequency domain energy statistics on the signal segments within the candidate pulse intervals, determine the center frequency based on the frequency position corresponding to the energy centroid, and associate the pulse peak value, rise point, duration, and center frequency with the phase scale number to generate electromagnetic phase points and ultrasonic phase points. If multiple candidate pulse intervals exist within the same phase scale, register the corresponding phase points respectively, and retain the amplitude sorting position of each phase point in the corresponding channel.

[0023] S214, the process of forming candidate pulse intervals includes: acquiring the average background amplitude and the background amplitude fluctuation range of the high-frequency electromagnetic detection channel and the ultrasonic detection channel during the pulse-free period, respectively; determining the pulse trigger threshold of the corresponding channel based on the average background amplitude and the background amplitude fluctuation range; when the envelope amplitude of consecutive sampling points is higher than the pulse trigger threshold of the corresponding channel and the duration reaches the minimum pulse holding time, registering the time range corresponding to the consecutive sampling points as candidate trigger intervals; merging candidate trigger intervals with adjacent intervals less than the pulse merging time threshold, and removing candidate trigger intervals with a duration exceeding the maximum pulse width threshold, thus obtaining candidate pulse intervals. The removed candidate trigger intervals do not enter the pulse peak, duration, and center frequency extraction process.

[0024] S3. Based on the phase difference between the electromagnetic phase point and the ultrasonic phase point, the amplitude sorting position, and the overlap state between the duration width corresponding to the electromagnetic phase point and the duration width corresponding to the ultrasonic phase point, the acoustic-electric co-source discharge event is determined, and an acoustic-electric joint phase distribution map is constructed based on the acoustic-electric co-source discharge event.

[0025] The input to S3 is an electromagnetic phase point and an ultrasonic phase point within the same power frequency cycle. The output is an acoustic-electric co-source discharge event and an acoustic-electric joint phase distribution map. Since the ultrasonic pulse needs time to propagate from the discharge location to the ultrasonic sensor, phase compensation is first performed on the ultrasonic phase point, and then it is paired with the electromagnetic phase point. Only when the phase difference threshold condition, amplitude order consistency condition, and duration width coincidence condition are all met is the paired event determined to be an acoustic-electric co-source discharge event.

[0026] S311: Obtain the electromagnetic phase point and ultrasonic phase point within the same power frequency cycle. Determine the ultrasonic propagation phase compensation amount based on the cable detection location, ultrasonic sensor installation location, and ultrasonic propagation velocity. Then, use the ultrasonic propagation phase compensation amount to correct the phase scale number of the ultrasonic phase point, thus obtaining the compensated ultrasonic phase point. The ultrasonic propagation phase compensation amount originates from the propagation path length, sound velocity configuration value, and phase shift relationship within the power frequency cycle, where the sound velocity configuration value corresponds to the cable sheath, insulation layer, or surrounding medium.

[0027] S312: Obtain the propagation path length between the ultrasonic sensor installation location and the cable detection location, as well as the sound velocity configuration value corresponding to the cable sheath, insulation layer, or surrounding medium. Determine the ultrasonic propagation time delay based on the propagation path length and sound velocity configuration value. Convert the ultrasonic propagation time delay into the corresponding phase offset within the power frequency cycle. When the phase offset exceeds half the width of a single phase scale, the phase scale number of the ultrasonic phase point is shifted forward to compensate, resulting in a compensated ultrasonic phase point. When the phase offset does not exceed half the width of a single phase scale, the phase scale number of the ultrasonic phase point remains unchanged, resulting in a compensated ultrasonic phase point. The compensated ultrasonic phase point retains its pulse peak value, rise time, duration, and center frequency fields.

[0028] S313, pair electromagnetic phase points with compensated ultrasonic phase points one by one, calculate the paired phase difference, compare the paired phase difference with the same-source phase difference threshold, and compare the amplitude ranking position and duration overlap of the paired points in their respective channels to generate same-source candidate pairing records. A same-source candidate pairing record must include at least the electromagnetic phase point identifier, the compensated ultrasonic phase point identifier, the paired phase difference, the amplitude ranking comparison result, the duration overlap state, and the same-source candidate state. If any condition is missing or a unique pairing relationship cannot be formed, the record will not be written into the joint phase registration table as an acoustic-electric same-source discharge event.

[0029] S314, when candidate pairing records from the same source simultaneously satisfy the phase difference threshold condition, amplitude sorting consistency condition, and duration width overlap condition, the corresponding electromagnetic phase point and the compensated ultrasonic phase point are identified as acoustic-electric co-source discharge events, and written into the joint phase registration table according to the power frequency cycle number and phase scale number, thus constructing an acoustic-electric joint phase distribution map. The acoustic-electric joint phase distribution map records the number of acoustic-electric co-source discharge events, electromagnetic pulse peak value statistics, ultrasonic pulse peak value statistics, electromagnetic center frequency statistics, and ultrasonic center frequency statistics in each phase scale, providing a unified phase distribution input for the deep neural network.

[0030] S4. Input the acoustic-electric combined phase distribution map into the deep neural network trained from the sample phase distribution map to determine the defect type confidence value and insulation severity score. Determine the partial discharge defect type according to the largest term in the defect type confidence value.

[0031] The input to S4 is the acoustic-electrical joint phase distribution map, and the output is the defect type confidence value, insulation severity score, and partial discharge defect type. The deep neural network is trained from the sample phase distribution map, and its network structure, training epochs, and sample size can be selected within the scope described in the application documents and are not specifically limited in this embodiment. The reproducible processing path of this step is as follows: the acoustic-electrical joint phase distribution map is parsed into layer data, a network input feature map is generated, which is input into the deep neural network, the defect type confidence value group and severity candidate score are read, and then the partial discharge defect type is determined according to the type determination confidence threshold.

[0032] S411 performs layer analysis on the acoustic-electric joint phase distribution map to obtain a phase density layer, an amplitude correlation layer, and a frequency distribution layer. The phase density layer, amplitude correlation layer, and frequency distribution layer are then aligned in phase scale and their amplitude scales are unified to generate the network input feature map. The phase density layer corresponds to the number of acoustic-electric co-source discharge events within each phase scale; the amplitude correlation layer corresponds to the statistical results of electromagnetic pulse peak values ​​and ultrasonic pulse peak values; and the frequency distribution layer corresponds to the statistical results of electromagnetic center frequency and ultrasonic center frequency.

[0033] S412: Obtain the number of acoustic-electric co-source discharge events, electromagnetic pulse peak value statistics, ultrasonic pulse peak value statistics, electromagnetic center frequency statistics, and ultrasonic center frequency statistics within each phase scale. Establish a fixed-order data arrangement relationship according to the phase scale number, and write the phase scale lacking acoustic-electric co-source discharge events into a zero-value placeholder mark. The zero-value placeholder mark is used to maintain the integrity of the phase scale sequence and is not included in the counting as a newly added discharge event.

[0034] S413, the amplitude and frequency data in the data arrangement are scaled according to the range boundaries of the corresponding channels, and the event quantity data, amplitude data, and frequency data are overlaid in channel order to generate a network input feature map. The network input feature map retains the correspondence between the phase scale order, the number of acoustic-electric discharge events, the amplitude statistics, and the center frequency statistics.

[0035] S414: The input feature map is input into the deep neural network, which outputs a defect type confidence value group and a severity candidate score. The confidence values ​​in the defect type confidence value group are then sorted to obtain the maximum confidence value and the corresponding defect category. When the maximum confidence value reaches the type determination confidence threshold, the defect category corresponding to the maximum confidence value is determined as a partial discharge defect type, and the severity candidate score is added to the insulation severity score. When the maximum confidence value does not reach the type determination confidence threshold, the partial discharge defect type is determined as a defect type to be reviewed, and the severity candidate score is added to the insulation severity score. Therefore, both the defect type identification result and the insulation severity score have clear input sources and output destinations.

[0036] S5 determines the maintenance and repair level based on the insulation severity score, the type of partial discharge defect, and the number of times the acoustic-electric discharge event occurs in the continuous power frequency cycle. Then, according to the maintenance and repair level, it determines and outputs the target repair instruction from the inspection cycle adjustment instruction, the live retest instruction, and the power outage maintenance instruction.

[0037] The S5 takes insulation severity score, partial discharge defect type, and number of consecutive occurrences as input, and outputs maintenance handling level and target handling instruction. The number of consecutive occurrences is the number of times the acoustic-electric discharge event occurs in a continuous power frequency cycle. A preset score range is used to form candidate handling levels. A defect type risk mapping table is used to calibrate the level according to the partial discharge defect type. The continuous discharge count threshold is used to determine whether the calibrated handling level needs to be increased.

[0038] S511: Obtain the insulation severity score, partial discharge defect type, and the number of occurrences of the acoustic-electric co-source discharge event in continuous power frequency cycles. Determine the number of occurrences of the acoustic-electric co-source discharge event in continuous power frequency cycles as the consecutive occurrence count. Compare the insulation severity score with a preset score interval to determine candidate treatment levels. If the insulation severity score is at the boundary of the score interval, determine the candidate treatment level according to the boundary assignment rules corresponding to the preset score interval. If the insulation severity score is missing, do not directly generate the final treatment level, but use this status as a review input status.

[0039] S512: Read the type risk level corresponding to the partial discharge defect type from the defect type risk mapping table, and perform level calibration on the candidate treatment level based on the type risk level to obtain the calibrated treatment level. If the partial discharge defect type is a defect type to be reviewed, the calibration process retains the review status and uses this status as the input basis for the live retest command or review-type treatment.

[0040] S513: When the number of consecutive occurrences reaches the continuous discharge count threshold, the calibration handling level is upgraded to an adjacent higher level to obtain the operation and maintenance handling level; when the number of consecutive occurrences does not reach the continuous discharge count threshold, the calibration handling level is determined as the operation and maintenance handling level. The determination of the number of consecutive occurrences is based on acoustic-electric discharge events from the same source, rather than single-channel candidate pulses, to maintain consistency between the handling level and the acoustic-electric discharge event screening results.

[0041] S514 matches the target handling instruction from the inspection cycle adjustment instruction, live-line retest instruction, and power outage maintenance instruction according to the operation and maintenance handling level, generates a handling instruction group, and outputs it. The target handling instruction is received by the operation and maintenance handling process; when an inspection cycle adjustment instruction is output, the handling instruction group carries the inspection cycle adjustment requirement; when a live-line retest instruction is output, the handling instruction group carries the retest trigger requirement; when a power outage maintenance instruction is output, the handling instruction group carries the maintenance handling requirement. If there is inconsistency between candidate handling levels, type risk levels, or consecutive occurrence counts, the final output is formed according to the processing order of level calibration and continuous discharge count threshold, and candidate handling levels that have not completed calibration are not directly used as the basis for the target handling instruction.

[0042] In the above method, the power frequency voltage reference signal is used to form a phase scale sequence, the high-frequency electromagnetic pulse signal and the ultrasonic pulse signal form electromagnetic phase points and ultrasonic phase points, respectively. The compensated ultrasonic phase points are used to compare the phase difference, amplitude sorting position and duration overlap with the electromagnetic phase points. The acoustic-electric co-source discharge event is used to construct an acoustic-electric joint phase distribution map. The acoustic-electric joint phase distribution map is input into a deep neural network via a network feature map, and the output result, together with the number of consecutive occurrences, forms the operation and maintenance handling level. Each intermediate state in this chain is formed by the previous processing link and called by the next processing link, thus completing the closed processing process from detection input to handling command output.

[0043] Example 2

[0044] Please see Figure 1 and Figure 5This embodiment provides a cable partial discharge assessment and decision-making system integrating acoustic and electrical PRPD. The difference between this embodiment and Embodiment 1 is that this embodiment implements the method described in Embodiment 1 in a system module collaborative manner. Signal acquisition, phase scale generation, pulse feature registration, homogeneity determination, spectrum construction, deep network diagnosis, and treatment linkage are respectively configured into corresponding functional modules, and the modules form a system-level processing chain according to the data sending, data receiving, and result calling relationship.

[0045] The system comprises a signal acquisition and calibration module, a pulse feature registration module, a homogeneous discharge determination and spectrum construction module, a deep network diagnostic module, and a risk assessment and linkage module. The signal acquisition and calibration module takes as input a power frequency voltage reference signal, a high-frequency electromagnetic pulse signal, and an ultrasonic pulse signal, and outputs a phase calibration sequence and a detection sequence with additional sampling time stamps. The pulse feature registration module receives the phase calibration sequence, the high-frequency electromagnetic pulse signal, and the ultrasonic pulse signal, and outputs electromagnetic phase points and ultrasonic phase points. The homogeneous discharge determination and spectrum construction module receives the electromagnetic phase points and ultrasonic phase points, and outputs acoustic-electric homogeneous discharge events and an acoustic-electric joint phase distribution spectrum. The deep network diagnostic module receives the acoustic-electric joint phase distribution spectrum and outputs a defect type confidence value, insulation severity score, and partial discharge defect type. The risk assessment and linkage module receives the defect type confidence value, insulation severity score, partial discharge defect type, and the number of occurrences of acoustic-electric homogeneous discharge events in continuous power frequency cycles, and outputs the operation and maintenance handling level and target handling instructions.

[0046] The signal acquisition and calibration module acquires power frequency voltage reference signals, high-frequency electromagnetic pulse signals, and ultrasonic pulse signals. It divides the power frequency cycle according to the rising zero-crossing time of the power frequency voltage reference signal, segments each power frequency cycle into phase scales, and sends the phase scales to the pulse feature registration module. When generating the phase scales, the signal acquisition and calibration module performs amplitude limiting, baseline offset subtraction, rising zero-crossing identification, power frequency cycle division, and phase scale sequence generation, and passes the sampling time stamp under a unified sampling clock to subsequent modules.

[0047] The pulse feature registration module is used to determine the pulse peak value, rise point, duration, and center frequency based on high-frequency electromagnetic pulse signals and ultrasonic pulse signals, respectively. It then registers electromagnetic and ultrasonic phase points according to the phase scale and sends these points to the homology determination and spectrum construction module. Upon receiving the phase scale sequence, this module filters candidate pulse intervals based on the channel noise baseline and pulse trigger threshold, and performs peak location, rise point location, duration determination, and center frequency determination on the candidate pulse intervals. The output electromagnetic and ultrasonic phase points each carry a power frequency cycle number, a phase scale number, and corresponding pulse feature fields.

[0048] The co-source determination and spectrum construction module is used to determine acoustic-electric co-source discharge events based on the phase difference between electromagnetic and ultrasonic phase points, amplitude ranking positions, and the overlap of the duration widths corresponding to the electromagnetic and ultrasonic phase points. It then constructs an acoustic-electric joint phase distribution spectrum based on these events and sends it to the deep network diagnostic module. This module first generates an ultrasonic propagation phase compensation value based on the cable detection location, ultrasonic sensor installation location, and ultrasonic propagation velocity. Then, it generates compensated ultrasonic phase points based on this compensation value. Subsequently, it pairs the electromagnetic phase points and compensated ultrasonic phase points item by item to generate co-source candidate pairing records. When the phase difference threshold, amplitude ranking consistency, and duration width overlap conditions are simultaneously met, the acoustic-electric co-source discharge event is registered.

[0049] The deep network diagnostic module inputs the acoustic-electric joint phase distribution map into a deep neural network trained from the sample phase distribution maps. It determines the defect type confidence value and insulation severity score, identifies the partial discharge defect type based on the largest term in the defect type confidence value, and sends the defect type confidence value, insulation severity score, and partial discharge defect type to the risk assessment and linkage module. This module performs phase density, amplitude correlation, and frequency distribution layer analysis on the acoustic-electric joint phase distribution map to generate a network input feature map. It then sorts the defect type confidence value groups to obtain the largest confidence value. When the largest confidence value reaches the type determination confidence threshold, it outputs the corresponding defect category; when the largest confidence value does not reach the type determination confidence threshold, it outputs the defect type to be reviewed.

[0050] The risk assessment and linkage module determines the maintenance and repair (O&M) handling level based on insulation severity scores, partial discharge defect types, and the frequency of occurrence of acoustic-electrical co-source discharge events within continuous power frequency cycles. It then identifies and outputs target handling instructions from inspection cycle adjustment instructions, live-line retest instructions, and power outage maintenance instructions according to the O&M handling level. This module first compares the insulation severity score with a preset score range to form candidate handling levels. Then, it reads the type risk level from the defect type risk mapping table and performs level calibration. When the number of consecutive occurrences reaches the continuous discharge count threshold, it outputs an upgraded O&M handling level; otherwise, it outputs the O&M handling level corresponding to the calibrated handling level. After the O&M handling level is determined, the risk assessment and linkage module matches the target handling instruction among the inspection cycle adjustment instructions, live-line retest instructions, and power outage maintenance instructions, and generates a handling instruction group.

[0051] The system described in Example 2 maintains consistency with the method described in Example 1 in terms of data and result items: the signal acquisition and calibration module corresponds to S1, the pulse feature registration module corresponds to S2, the homology determination and spectrum construction module corresponds to S3, the deep network diagnosis module corresponds to S4, and the risk assessment and linkage module corresponds to S5. The data transmission direction between the modules is consistent with the processing order in Example 1, with the output of the previous module serving as the input of the next module. Finally, the risk assessment and linkage module outputs the target treatment instruction, thereby realizing the technical solution described in this invention through system module collaboration.

[0052] The specific processes, field states, judgment conditions, processing order, parameter sources, module collaboration relationships, and disposal output relationships described in the above embodiments are all used to illustrate the feasible implementations of the present invention and should not limit the present invention to the specific embodiments listed. Without departing from the technical solutions described in the present invention and the scope of the application documents, any equivalent substitutions, equivalent modifications, equivalent combinations, order adjustments, corresponding module replacements, equivalent changes in field names, equivalent acceptance of the executing entity, or equivalent changes in the carrier form that can be conceived by those skilled in the art should fall within the scope of protection of this patent; however, they should not be extended to unclaimed topics, nor should the substantive correspondence of the technical objects be changed through changes in reference numerals.

Claims

1. A cable partial discharge assessment and decision-making method integrating acoustic and electrical PRPD, characterized in that, include: Acquire power frequency voltage reference signal, high frequency electromagnetic pulse signal and ultrasonic pulse signal, divide power frequency cycles according to the rising zero crossing time of the power frequency voltage reference signal, and divide each power frequency cycle into phase scale; Based on the high-frequency electromagnetic pulse signal and the ultrasonic pulse signal, the pulse peak value, rise point, duration and center frequency are determined respectively, and the electromagnetic phase point and ultrasonic phase point are registered according to the phase scale. Based on the phase difference between the electromagnetic phase point and the ultrasonic phase point, the amplitude sorting position, and the overlap between the duration width corresponding to the electromagnetic phase point and the duration width corresponding to the ultrasonic phase point, an acoustic-electric co-source discharge event is determined, and an acoustic-electric joint phase distribution map is constructed based on the acoustic-electric co-source discharge event. The acoustic-electric combined phase distribution map is input into a deep neural network trained from the sample phase distribution map to determine the defect type confidence value and insulation severity score. The partial discharge defect type is determined according to the largest term in the defect type confidence value. Based on the insulation severity score, the partial discharge defect type, and the number of times the acoustic-electric co-source discharge event occurs in the continuous power frequency cycles, the operation and maintenance handling level is determined, and the target handling instruction is determined and output from the inspection cycle adjustment instruction, the live retest instruction, and the power outage maintenance instruction according to the operation and maintenance handling level.

2. The cable partial discharge assessment and decision-making method integrating acoustic and electrical PRPD according to claim 1, characterized in that, The step of dividing the power frequency cycle according to the rising zero-crossing time of the power frequency voltage reference signal, and dividing each power frequency cycle into phase scales, includes: The power frequency voltage reference signal, the high-frequency electromagnetic pulse signal, and the ultrasonic pulse signal within the same detection time window are acquired. The power frequency voltage reference signal is subjected to amplitude limiting and baseline offset subtraction. Sampling time markers are added to the high-frequency electromagnetic pulse signal and the ultrasonic pulse signal according to a unified sampling clock. Based on the voltage polarity change and slope direction of continuous sampling points, sampling positions that change from negative polarity to positive polarity and have a slope greater than the zero-crossing slope threshold are selected, and the sampling positions are determined as the rising zero-crossing moment. The power frequency cycle is determined by the time interval between two adjacent rising zero crossing times, and each power frequency cycle is divided into phase scales according to the equal-interval phase segmentation rule to generate a phase scale sequence.

3. The cable partial discharge assessment and decision-making method integrating acoustic and electrical PRPD according to claim 1, characterized in that, The steps of determining the pulse peak value, rise point, duration, and center frequency based on the high-frequency electromagnetic pulse signal and the ultrasonic pulse signal, and registering the electromagnetic phase point and ultrasonic phase point according to the phase scale, include: The envelope sequences of the high-frequency electromagnetic pulse signal and the ultrasonic pulse signal within the same detection time window are obtained. Candidate pulse intervals are screened based on the channel noise baseline and the pulse trigger threshold. The sampling point with the largest amplitude in each candidate pulse interval is determined as the pulse peak value. The sampling positions that are continuously lower than the starting point determination threshold before the pulse peak are retrieved in reverse time. The sampling position that is closest to the pulse peak and meets the condition of continuous low amplitude is determined as the rising starting point. The continuous width is determined according to the time interval between the rising starting point and the falling end point. Frequency domain energy statistics are performed on the signal segments within the candidate pulse interval. The center frequency is determined based on the frequency position corresponding to the energy centroid. The pulse peak value, the rising start point, the duration, and the center frequency are associated with the phase scale number to generate the electromagnetic phase point and the ultrasonic phase point.

4. The cable partial discharge assessment and decision-making method integrating acoustic and electrical PRPD according to claim 3, characterized in that, The screening process for the candidate pulse interval includes: The average background amplitude and the fluctuation range of the background amplitude in the non-pulse period are obtained for the high-frequency electromagnetic detection channel and the ultrasonic detection channel, respectively. The pulse trigger threshold of the corresponding channel is determined based on the average background amplitude and the fluctuation range of the background amplitude. When the envelope amplitude of consecutive sampling points is higher than the pulse trigger threshold of the corresponding channel and the duration reaches the minimum pulse hold time, the time range corresponding to the consecutive sampling points is registered as a candidate trigger interval. Candidate trigger intervals with adjacent intervals less than the pulse merging time threshold are merged, and candidate trigger intervals with durations exceeding the maximum pulse width threshold are removed to obtain the candidate pulse intervals.

5. The cable partial discharge assessment and decision-making method integrating acoustic and electrical PRPD according to claim 1, characterized in that, The steps of identifying acoustic-electric co-source discharge events and constructing an acoustic-electric joint phase distribution map based on the acoustic-electric co-source discharge events include: The electromagnetic phase point and the ultrasonic phase point within the same power frequency cycle are obtained. The ultrasonic propagation phase compensation amount is determined based on the cable detection position, the ultrasonic sensor installation position and the ultrasonic propagation speed. The phase scale number of the ultrasonic phase point is corrected using the ultrasonic propagation phase compensation amount to obtain the compensated ultrasonic phase point. The electromagnetic phase points and the compensated ultrasonic phase points are paired one by one, the pairing phase difference is calculated, the pairing phase difference is compared with the same source phase difference threshold, and the amplitude sorting position and the continuous width overlap state of the paired parties in their respective channels are compared to generate same source candidate pairing records. When the candidate pairing records of the same source simultaneously meet the phase difference threshold condition, the amplitude sorting consistency condition, and the duration width coincidence condition, the corresponding electromagnetic phase point and the compensated ultrasonic phase point are determined as the acoustic-electric co-source discharge event, and written into the joint phase registration table according to the power frequency cycle number and phase scale number to construct the acoustic-electric joint phase distribution map.

6. The cable partial discharge assessment and decision-making method integrating acoustic and electrical PRPD according to claim 5, characterized in that, The process of determining the ultrasonic propagation phase compensation amount includes: The propagation path length between the installation position of the ultrasonic sensor and the detection position of the cable is obtained, as well as the sound velocity configuration value corresponding to the cable sheath, insulation layer or surrounding medium. The ultrasonic propagation time delay is determined based on the propagation path length and the sound velocity configuration value. The ultrasonic propagation time delay is converted into the corresponding phase offset within the power frequency cycle. When the phase offset exceeds half the width of a single phase scale, the phase scale number of the ultrasonic phase point is shifted forward to compensate for the loss, thus obtaining the compensated ultrasonic phase point. When the phase offset does not exceed half the width of a single phase scale, the phase scale number of the ultrasonic phase point remains unchanged, and the compensated ultrasonic phase point is obtained.

7. The cable partial discharge assessment and decision-making method integrating acoustic and electrical PRPD according to claim 1, characterized in that, The steps of determining the defect type confidence value and insulation severity score, and determining the partial discharge defect type according to the largest item among the defect type confidence values, include: The acoustic-electric combined phase distribution map is analyzed to obtain a phase density layer, an amplitude correlation layer, and a frequency distribution layer. The phase density layer, the amplitude correlation layer, and the frequency distribution layer are aligned in phase scale and unified in amplitude scale to generate a network input feature map. The network input feature map is input into the deep neural network, which outputs a defect type confidence value group and a severity candidate score, and sorts the confidence values ​​in the defect type confidence value group to obtain the maximum confidence value and the defect category corresponding to the maximum confidence value. When the maximum confidence value reaches the type determination confidence threshold, the defect category corresponding to the maximum confidence value is determined as the partial discharge defect type, and the severity candidate score is written into the insulation severity score; When the maximum confidence value does not reach the type determination confidence threshold, the partial discharge defect type is determined as a defect type to be reviewed, and the severity candidate score is written into the insulation severity score.

8. The cable partial discharge assessment and decision-making method integrating acoustic and electrical PRPD according to claim 7, characterized in that, The process of generating the network input feature map includes: The number of acoustic-electric co-source discharge events, the statistical results of electromagnetic pulse peak values, the statistical results of ultrasonic pulse peak values, the statistical results of electromagnetic center frequency, and the statistical results of ultrasonic center frequency are obtained within each phase scale. A data arrangement relationship with a fixed order is established according to the phase scale number, and the phase scale that lacks the acoustic-electric co-source discharge event is written into a zero-value placeholder mark. The amplitude data and frequency data in the data arrangement relationship are scaled according to the range boundaries of the corresponding channels, and the event quantity data, amplitude data and frequency data are stacked in channel order to generate the network input feature map.

9. The cable partial discharge assessment and decision-making method integrating acoustic and electrical PRPD according to claim 1, characterized in that, The steps for determining the maintenance and repair level, and then selecting and outputting the target repair instruction from the inspection cycle adjustment instruction, live-line retest instruction, and power outage maintenance instruction according to the maintenance and repair level include: The insulation severity score, the partial discharge defect type, and the number of occurrences of the acoustic-electric co-source discharge event in the continuous power frequency cycles are obtained. The number of occurrences of the acoustic-electric co-source discharge event in the continuous power frequency cycles is determined as the number of consecutive occurrences. The insulation severity score is compared with a preset score interval to determine the candidate treatment level. The type risk level corresponding to the partial discharge defect type is read from the defect type risk mapping table, and the candidate treatment level is calibrated based on the type risk level to obtain the calibrated treatment level; When the number of consecutive occurrences reaches the threshold for the number of continuous discharges, the calibration handling level is upgraded to an adjacent higher level to obtain the operation and maintenance handling level; When the number of consecutive occurrences does not reach the threshold of the number of continuous discharges, the calibration handling level is determined as the operation and maintenance handling level, and the target handling instruction is matched from the inspection cycle adjustment instruction, the live retest instruction and the power outage maintenance instruction according to the operation and maintenance handling level to generate a handling instruction group.

10. A cable partial discharge assessment and decision-making system integrating acoustic and electrical PRPD, characterized in that, The system is used to implement the method according to any one of claims 1-9, comprising: The signal acquisition and calibration module is used to acquire power frequency voltage reference signal, high frequency electromagnetic pulse signal and ultrasonic pulse signal, divide the power frequency cycle according to the rising zero crossing time of the power frequency voltage reference signal, divide each power frequency cycle into phase scale, and send the phase scale to the pulse feature registration module; The pulse feature registration module is used to determine the pulse peak value, rise start point, duration and center frequency based on the high-frequency electromagnetic pulse signal and the ultrasonic pulse signal, respectively, and register the electromagnetic phase point and the ultrasonic phase point according to the phase scale, and send the electromagnetic phase point and the ultrasonic phase point to the homology determination and map construction module. The homogeneity determination and spectrum construction module is used to determine the acoustic-electric homogeneous discharge event based on the phase difference between the electromagnetic phase point and the ultrasonic phase point, the amplitude sorting position, and the overlap state between the continuous width corresponding to the electromagnetic phase point and the continuous width corresponding to the ultrasonic phase point. The module then constructs an acoustic-electric joint phase distribution spectrum based on the acoustic-electric homogeneous discharge event and sends the acoustic-electric joint phase distribution spectrum to the deep network diagnostic module. The deep network diagnostic module is used to input the acoustic-electric combined phase distribution map into a deep neural network trained from the sample phase distribution map, determine the defect type confidence value and insulation severity score, determine the partial discharge defect type according to the largest item in the defect type confidence value, and send the defect type confidence value, the insulation severity score and the partial discharge defect type to the risk assessment and linkage module; The risk assessment and linkage module is used to determine the operation and maintenance handling level based on the insulation severity score, the partial discharge defect type, and the number of times the acoustic-electric co-source discharge event occurs in the continuous power frequency cycle, and to determine and output the target handling instruction from the inspection cycle adjustment instruction, the live retest instruction, and the power outage maintenance instruction according to the operation and maintenance handling level.

Images