Causal determination method for vfto-induced partial discharge based on timing characteristics
By collecting VFTO and UHF partial discharge data through integrated sensing devices, performing signal preprocessing and feature fusion verification, the problem of misjudging the cause and effect of VFTO-induced partial discharge was solved, and high-precision cause and effect determination was achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHONGQING ZHENYUAN ELECTRICAL CO LTD
- Filing Date
- 2026-04-10
- Publication Date
- 2026-05-29
AI Technical Summary
Existing technologies cannot effectively distinguish between partial discharge induced by VFTO and partial discharge caused by background interference or inherent defects in equipment, leading to misjudgment of cause and effect and making it difficult to accurately identify the VFTO-partial discharge positive feedback degradation mechanism, which exacerbates the risk, especially in ultra-high voltage projects.
VFTO and UHF partial discharge data are collected synchronously by an integrated sensing device. Signal preprocessing is performed, time-series features are extracted, causal relationships are determined based on correlation quantification criteria, the best-matching partial discharge feature units are selected, feature fusion verification is performed, and causal determination conclusions are output.
It enables accurate causal determination of VFTO-induced partial discharge, improves the accuracy and reliability of the determination, and supports accurate assessment of equipment insulation status.
Smart Images
Figure CN122109747A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the technical field of power equipment condition monitoring, and in particular to a causal determination method for VFTO-induced partial discharge based on time-series characteristics. Background Technology
[0002] In recent years, in the field of high-voltage power equipment operation and maintenance, there are related technologies for monitoring rapid transient overvoltage (VFTO) and detecting ultra-high frequency partial discharge. Through integrated sensing devices, multiple physical quantities such as VFTO and partial discharge can be collected simultaneously to assess the insulation status of high-voltage equipment such as GIS, switchgear, and circuit breakers.
[0003] However, existing analysis methods are mainly based on threshold judgment of a single physical quantity or shallow time series correlation analysis. They lack a deep understanding of the time series coupling and correlation of multiple physical quantities between VFTO and UHF partial discharge, as well as the ability to accurately match and quantify the cascade causal relationship between the two. They cannot effectively distinguish between VFTO-induced partial discharge and partial discharge caused by background interference or inherent defects in the equipment, which can easily lead to causal misjudgment and make it difficult to accurately identify the "VFTO-partial discharge" positive feedback degradation mechanism.
[0004] With the advancement of new power system construction and the large-scale commissioning of ultra-high voltage projects and high-altitude substations, the VFTO amplitude and bandwidth generated by the operation of GIS equipment disconnecting switches are higher and wider (reaching hundreds of MHz to GHz levels), significantly increasing the risk of inducing partial discharge at weak insulation points. The problem of insufficient accuracy in causal determination by existing technologies has become increasingly prominent, making it difficult to meet the core requirements of safe operation and accurate insulation status assessment of high-voltage power equipment. Summary of the Invention
[0005] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.
[0006] In view of the problems existing in the prior art, the present invention is proposed.
[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution: A causal determination method for VFTO-induced partial discharge based on time-series characteristics includes: Acquire VFTO detection dataset and UHF partial discharge detection dataset synchronously collected by integrated sensing device; wherein: the VFTO detection dataset includes multiple sets of VFTO collected data, and the UHF partial discharge detection dataset includes multiple sets of UHF partial discharge collected data. The VFTO acquisition data is preprocessed to obtain multiple VFTO timing feature structured units; and the UHF partial discharge acquisition data is preprocessed to obtain multiple effective partial discharge timing feature units. Based on the correlation quantification criterion, the causal correlation quantification value between each VFTO time-series feature structured unit and the corresponding partial discharge feature of the time period is determined; and based on the causal correlation quantification value, it is determined whether the collected data corresponding to each VFTO time-series feature structured unit is partial discharge induced source data, and the corresponding causal determination rule is determined. If the data is determined to be suspected induced source data, then based on the feature matching criterion, the effective partial discharge timing feature unit that best matches the VFTO timing feature structured unit is selected. Based on the causal determination rule, the optimally matched effective timing feature unit of partial discharge and the corresponding VFTO timing feature structured unit are subjected to feature fusion verification; and the collected data after each fusion verification are comprehensively analyzed to output the causal determination conclusion of VFTO-induced partial discharge.
[0008] As a preferred embodiment of the causal determination method for VFTO-induced partial discharge based on time-series features described in this invention, the method involves: preprocessing the VFTO acquisition data to obtain multiple VFTO time-series feature structured units, including: The VFTO-acquired data is demodulated and its features are extracted to obtain the corresponding time-series feature information; The time-series feature information is divided into time periods based on time-domain and frequency-domain features to obtain multiple feature units; Each of the aforementioned feature units is assigned a corresponding detection timestamp and signal quality characterization parameters to obtain multiple VFTO time-series feature structured units.
[0009] As a preferred embodiment of the causal determination method for VFTO-induced partial discharge based on time-series characteristics described in this invention, wherein: signal preprocessing is performed on each of the ultra-high frequency partial discharge acquisition data to obtain multiple effective time-series characteristic units of partial discharge, including: Calculate the inter-frame temporal feature deviation of each of the ultra-high frequency partial discharge acquisition data, and determine whether the corresponding partial discharge acquisition data has feature segmentation nodes based on the inter-frame temporal feature deviation; If so, the corresponding partial discharge acquisition data is segmented based on the feature segmentation nodes; the average signal amplitude and effective acquisition duration of each partial discharge acquisition data are calculated. Based on preset amplitude thresholds and preset duration thresholds, multiple sets of partial discharge acquisition data are filtered to obtain multiple effective temporal feature units of partial discharge.
[0010] As a preferred embodiment of the causal determination method for VFTO-induced partial discharge based on time-series features described in this invention, the causal correlation quantification value between each VFTO time-series feature structured unit and the corresponding time-segment partial discharge feature is determined based on the correlation quantification criterion, including: Based on each of the VFTO time-series feature structured units, the corresponding VFTO feature saliency and the proportion of key frequency domain components are determined. Based on the partial discharge acquisition data of each VFTO time-series structured unit corresponding to the time period, the corresponding partial discharge intensity, discharge phase concentration, and detection signal signal-to-noise ratio are determined. The causal correlation quantification value between VFTO and partial discharge is obtained by weighting the VFTO feature saliency, the proportion of key frequency domain components, the partial discharge intensity, the discharge phase concentration, and the signal-to-noise ratio of the detection signal.
[0011] As a preferred embodiment of the causal determination method for VFTO-induced partial discharge based on time-series features described in this invention, the method involves determining whether the collected data corresponding to each VFTO time-series feature structured unit is partial discharge induction source data based on the causal correlation quantization value, and simultaneously determining the corresponding causal determination rule, including determining whether the causal correlation quantization value is greater than or equal to a first preset threshold. If yes, the VFTO data is directly determined to be partial discharge induced source data; otherwise, it is determined to be suspected induced source data or non-induced source data. If the data is determined to be suspected induced source data, and the causal correlation quantification value is between the second preset threshold and the first preset threshold, then the core feature part of the VFTO time series feature structured unit is retained, and the remaining feature parts are fully fused and verified with the optimally matched effective time series feature unit of partial discharge. If the causal correlation quantification value is less than the second preset threshold, the VFTO collected data is determined to be non-inducing source data, and the VFTO time-series feature structured unit and the effective time-series feature unit of partial discharge are marked as having no causal correlation.
[0012] As a preferred embodiment of the causal determination method for VFTO-induced partial discharge based on time-series features described in this invention, the effective time-series feature units of partial discharge that best match the VFTO time-series feature structured unit are selected based on feature matching criteria, including: Calculate the timing feature coupling degree and the operating condition entity fit degree between the VFTO timing feature structured unit and each of the partial discharge effective timing feature units; Based on the timing feature coupling degree and the operating condition entity fit degree, the effective timing feature unit of partial discharge that best matches the VFTO timing feature structured unit is selected from multiple effective timing feature units of partial discharge.
[0013] As a preferred embodiment of the causal determination method for VFTO-induced partial discharge based on time-series features according to the present invention, the method includes: based on the causal determination rule, performing feature fusion verification between the optimally matched effective time-series feature unit of partial discharge and the corresponding VFTO time-series feature structured unit, including: Determine the timing coupling anchor point between the optimally matched effective timing feature unit of partial discharge and the VFTO timing feature structured unit; the timing coupling anchor point includes one of the VFTO signal peak frame, the partial discharge signal start frame, and the device operation trigger frame; Based on the timing coupling anchor point and the causal determination rule, the optimally matched effective timing feature unit of partial discharge and the corresponding VFTO timing feature structured unit are subjected to feature fusion verification.
[0014] As a preferred embodiment of the VFTO-induced partial discharge causality determination method based on timing features described in this invention, the method further includes, before performing signal preprocessing on each of the UHF partial discharge acquisition data to obtain multiple effective timing feature units of partial discharge, performing validity and anti-interference verification on each of the UHF partial discharge acquisition data, and removing noise frames, invalid acquisition frames and interference signal frames.
[0015] A system for determining the causal relationship of VFTO-induced partial discharge based on the above-mentioned time-series characteristics is provided. The system includes: a data acquisition module for acquiring VFTO detection datasets and UHF partial discharge detection datasets synchronously acquired by an integrated sensing device; wherein: the VFTO detection dataset includes multiple sets of VFTO acquisition data, and the UHF partial discharge detection dataset includes multiple sets of UHF partial discharge acquisition data. The signal preprocessing module is used to preprocess the VFTO acquisition data to obtain multiple VFTO timing feature structured units; and to preprocess the UHF partial discharge acquisition data to obtain multiple effective timing feature units of partial discharge. The correlation quantization module is used to determine the causal correlation quantization value between each VFTO time-series feature structured unit and the corresponding partial discharge feature of the time period based on the correlation quantization criterion; and to determine whether the collected data corresponding to each VFTO time-series feature structured unit is partial discharge induced source data based on the causal correlation quantization value, and at the same time determine the corresponding causal determination rule. The feature matching module is used to select the effective partial discharge timing feature unit that best matches each of the VFTO timing feature structured units if the data is determined to be suspected inducing source data, based on the feature matching criteria. The causal determination module is used to perform feature fusion verification between the optimally matched effective timing feature unit of partial discharge and the corresponding VFTO timing feature structured unit based on the causal determination rules; and to perform comprehensive analysis on the collected data after each fusion verification, and output the causal determination conclusion of VFTO-induced partial discharge.
[0016] The present invention also discloses a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described method for determining the causal relationship of VFTO-induced partial discharge based on timing characteristics.
[0017] The beneficial effects of this invention are as follows: After synchronously collecting VFTO and UHF partial discharge multi-physical quantity data through an integrated sensing device, the temporal coupling correlation between the two can be automatically explored, the inherent logic of insulation impact and discharge behavior can be analyzed, and end-to-end intelligent analysis can be realized from data acquisition to feature extraction, correlation quantification, causal verification and conclusion output, which effectively improves the accuracy and reliability of VFTO-induced partial discharge causal determination. Attached Figure Description
[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a schematic diagram of the overall process of the causal determination method for VFTO-induced partial discharge based on time-series characteristics proposed in this invention. Figure 2 This is a logical framework diagram of the causal determination method for VFTO-induced partial discharge based on time-series characteristics proposed in this invention. Detailed Implementation
[0019] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
[0020] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.
[0021] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.
[0022] Reference Figures 1-2 As an embodiment of the present invention, a causal determination method for VFTO-induced partial discharge based on time-series characteristics is provided. This method includes the following steps: Step 1: Obtain the VFTO detection dataset and the UHF partial discharge detection dataset synchronously collected by the integrated sensing device; wherein: the VFTO detection dataset includes multiple sets of VFTO collected data, and the UHF partial discharge detection dataset includes multiple sets of UHF partial discharge collected data. In the operation and maintenance of ultra-high voltage (UHV) power equipment, integrated sensing devices are typically used on-site, such as capacitive voltage divider VFTO detection units integrated with UHV antennas. These devices simultaneously collect raw detection data of multiple physical quantities during switch operation, covering multiple sets of VFTO signals and UHV partial discharge signals. VFTO data is used to characterize the amplitude, rise time, and frequency domain distribution of the rapid transient overvoltage generated by disconnector operation, serving as a core basis for assessing the insulation impact strength of the equipment. UHV partial discharge data is used to capture electromagnetic wave signals excited by weak points in the insulation, providing a key indicator for identifying early insulation defects in the equipment.
[0023] Step 2: Perform signal preprocessing on each VFTO acquisition data to obtain multiple VFTO timing feature structured units; and perform signal preprocessing on each UHF partial discharge acquisition data to obtain multiple effective partial discharge timing feature units; This step performs noise filtering, feature extraction, and validity screening on each group of VFTO and partial discharge signals in the original detection data, providing high-quality structured input for subsequent causal correlation analysis.
[0024] Step 3: Based on the correlation quantification criterion, determine the causal correlation quantification value between each VFTO time-series feature structured unit and the corresponding partial discharge characteristics; and based on the causal correlation quantification value, determine whether the collected data corresponding to each VFTO time-series feature structured unit is partial discharge induced source data, and at the same time determine the corresponding causal determination rule. This step calculates a causal correlation quantization value for each VFTO time-series structured feature unit based on a multi-index weighting criterion. This quantization value measures the induced coupling strength between the VFTO signal and the corresponding partial discharge signal. Subsequently, based on the causal correlation quantization value, it is determined whether the VFTO time-series structured feature unit is a suspected inducing source of partial discharge. If determined to be a suspected inducing source, the optimal matching feature unit is selected from multiple valid partial discharge time-series feature units for subsequent verification. Simultaneously, based on the interval division of the causal correlation quantization value, corresponding causal determination rules are determined. These rules include different processing methods such as retaining and verifying core feature segments and marking and excluding non-correlated segments.
[0025] Step 4: If the data is determined to be suspected induced source data, then based on the feature matching criterion, select the effective partial discharge timing feature unit that best matches the VFTO timing feature structured unit.
[0026] Step 5: Based on the causal determination rule, perform feature fusion verification between the optimally matched effective timing feature unit of partial discharge and the corresponding VFTO timing feature structured unit; and conduct comprehensive analysis on the collected data after each fusion verification to output the causal determination conclusion of VFTO-induced partial discharge.
[0027] Simultaneously complete signal noise reduction and completion, feature alignment and operation condition label association, organize VFTO time series features, partial discharge features and equipment operating parameters into a standardized analysis dataset, drive subsequent causal path tracing, degradation degree assessment and multi-dimensional result visualization, and output a causal determination report that can be directly used for equipment operation and maintenance decisions.
[0028] In the aforementioned method for determining the causal relationship of VFTO-induced partial discharge based on time-series characteristics, after simultaneously acquiring multiple physical quantity data of VFTO and UHF partial discharge through an integrated sensing device, the temporal coupling relationship between the two can be automatically mined, the inherent logic of insulation impact and discharge behavior can be analyzed, and end-to-end intelligent analysis can be realized from data acquisition to feature extraction, correlation quantification, causal verification, and conclusion output, effectively improving the accuracy and reliability of determining the causal relationship of VFTO-induced partial discharge.
[0029] In one embodiment, signal preprocessing is performed on the VFTO acquisition data to obtain multiple VFTO timing feature structured units, including: The VFTO-acquired data is demodulated and feature extracted to obtain the corresponding time-series feature information. Based on time-domain and frequency-domain features, time-series feature information is divided into time periods to obtain multiple feature units; Each feature unit is assigned a corresponding detection timestamp and signal quality characterization parameters to obtain multiple VFTO time-series feature structured units.
[0030] Specifically, in one particular embodiment, signal preprocessing of the data acquired by each VFTO includes the following steps: Step S101, VFTO signal preprocessing: For the high-frequency transient signal of the original VFTO acquisition data, an adaptive noise suppression algorithm based on wavelet transform is used to filter out interference, and the effective frequency band of the signal is constrained to the range of 1MHz-1GHz to improve the accuracy of feature extraction and the stability of subsequent time period division.
[0031] Step S102, signal demodulation and time-frequency domain feature alignment: Based on wavelet packet decomposition technology, the VFTO signal is demodulated and features are extracted, and the time-domain amplitude sequence, frequency-domain energy distribution and microsecond-level timestamps corresponding to the output signal are generated; at the same time, the amplitude stability coefficient and signal-to-noise ratio (SNR) of each feature segment are calculated.
[0032] Step S103, Feature Data Cleaning: Abnormal pulses with amplitudes exceeding the device's measurement range are removed. A sliding window averaging method is used to smooth signal spikes and correct frequency domain phase shifts caused by acquisition delays. Furthermore, the Empirical Mode Decomposition (EMD) algorithm is used to extract core transient features and remove redundant low-frequency components.
[0033] Step S104, temporal feature segmentation: Combining temporal amplitude abrupt changes and frequency domain energy peaks, the continuous feature sequence is segmented into time periods. Unit segmentation is performed according to the following rules: if the amplitude abrupt change exceeds three times the baseline, or the duration of the frequency domain energy peak is >50μs, it is used as a potential segmentation point; if the duration of the feature unit is >200μs, auxiliary segmentation is performed based on waveform similarity; if the unit duration is <20μs and is interference noise, it is merged with neighboring valid units. For example, one VFTO temporal feature structured unit is: {"unit_id":5,"start_time":0.123,"end_time":0.187,"amp_peak":235.6,"freq_peak":450e6,"stability":0.92,"snr":31.4}.
[0034] The fields are interpreted as follows: unit_id: unit ID; start_time: start time; end_time: end time; amp_peak: peak amplitude; freq_peak: peak frequency; stability: stability value; snr: signal-to-noise ratio. This structured unit serves as the input for subsequent causal correlation quantization, partial discharge feature matching, and fusion verification.
[0035] In one embodiment, signal preprocessing is performed on the ultra-high frequency partial discharge acquisition data to obtain multiple effective timing feature units of partial discharge, including: Calculate the inter-frame temporal feature deviation of each UHF partial discharge acquisition data, and determine whether the corresponding partial discharge acquisition data has feature segmentation nodes based on the inter-frame temporal feature deviation. If so, the corresponding partial discharge acquisition data is segmented based on the feature segmentation nodes; the average signal amplitude and effective acquisition duration of each partial discharge acquisition data are calculated. Based on preset amplitude thresholds and preset duration thresholds, multiple sets of partial discharge acquisition data are filtered to obtain multiple effective temporal feature units of partial discharge.
[0036] Step S201, Signal preprocessing: Band normalization is performed on the UHF partial discharge acquisition data to remove low-frequency interference below 300MHz and UHF noise components above 2GHz. At the same time, wavelet threshold denoising algorithm is used to smooth the signal and calibrate the amplitude reference to eliminate the acquisition link gain deviation.
[0037] Step S202, Partial Discharge Feature Segmentation and Effective Segment Selection: To extract representative effective temporal feature units of partial discharge, feature segmentation node detection and effective segment selection need to be performed. First, the inter-frame temporal feature deviation (amplitude gradient change and frequency domain energy shift) is calculated, feature segmentation points are marked, and the partial discharge acquisition data is segmented according to the segmentation points. Then, the average amplitude and energy concentration of each sub-segment are calculated, and sub-segments with amplitudes higher than the baseline and concentrated energy are selected (represented as effective discharge features). Finally, by limiting the duration of effective feature units to 50μs~200μs, effective temporal feature units of partial discharge are selected.
[0038] Timing characteristic deviation refers to a measure used during signal acquisition to identify the degree of amplitude and frequency domain energy change between adjacent sampling points. Feature segmentation nodes refer to sampling moments in the signal where amplitude changes abruptly or energy changes abruptly, typically corresponding to the start or end of a partial discharge. In partial discharge detection, timing characteristic deviation is a quantitative representation of the degree of abrupt change in the discharge pulse; feature segmentation nodes are used for the precise location of the boundaries of effective discharge events.
[0039] After partial discharge segments are filtered by timing feature deviation, they are then divided into several effective timing feature units of partial discharge between 50μs and 200μs by a timing correlation clustering algorithm according to the correlation of discharge pulses.
[0040] Step S203, generating structured labels for features: In order to support subsequent tasks such as time-series feature matching and operational condition entity consistency verification, each effective time-series feature unit of partial discharge needs to generate a structured label, including {unit ID, timestamp interval, peak amplitude, frequency domain center frequency, energy concentration, signal-to-noise ratio, operational condition label}. Step S204, Validity and Anti-interference Verification: Each effective timing feature unit of partial discharge must pass the pulse repetition rate verification and interference identification algorithm to eliminate periodic interference and noise spurious peaks, so as to ensure the reliability of the final causal determination result.
[0041] Through steps S201 to S204 above, each original UHF partial discharge acquisition data is standardized and structured with labels, providing a candidate feature basis for subsequent time-series feature matching, causal correlation quantification, and fusion verification.
[0042] In one embodiment, based on the correlation quantification criterion, the causal correlation quantification value between each VFTO time-series feature structured unit and the corresponding time-period partial discharge feature includes: Based on the structured units of each VFTO time-series feature, the saliency of the corresponding VFTO feature and the proportion of key frequency domain components are determined. Based on the partial discharge acquisition data of each VFTO time-series characteristic structured unit corresponding to the time period, the corresponding partial discharge intensity, discharge phase concentration, and detection signal signal-to-noise ratio are determined. The causal correlation between VFTO and partial discharge is quantified by weighting VFTO feature significance, key frequency domain component proportion, partial discharge intensity, discharge phase concentration, and signal-to-noise ratio of the detection signal.
[0043] The causal correlation quantification indicators in this application include VFTO feature significance, key frequency domain component proportion, partial discharge intensity, discharge phase concentration, and detection signal-to-noise ratio, as detailed in Table 1 below.
[0044] Table 1: Quantitative Indicators of Causal Relationships
[0045] Among them, the VFTO feature saliency and the proportion of key frequency domain components are determined based on the VFTO time-series feature structured units, the partial discharge intensity and discharge phase concentration are determined based on the partial discharge acquisition data of the corresponding time period, and the signal-to-noise ratio of the detection signal is determined based on the joint noise analysis of VFTO and partial discharge signals.
[0046] This application embodiment constructs a causal correlation quantification system of multiple physical quantities to achieve intelligent discrimination of the induced coupling relationship between VFTO and partial discharge. It integrates indicators such as impact intensity, high frequency components, discharge severity, phase correlation, and signal quality, and outputs causal correlation quantification values as the core basis for judgment, which can accurately distinguish between induced discharge and interference / inherent defect discharge.
[0047] In a specific embodiment, the causal correlation quantization value (CausalScore) of each VFTO temporal feature structured unit is calculated using the following formula: CausalScore = 0.25×Q1 + 0.2×Q2 + 0.25×Q3 + 0.2×Q4 + 0.1×Q5; where Q1 is the significance of VFTO features, Q2 is the proportion of key frequency domain components, Q3 is the partial discharge intensity, Q4 is the discharge phase concentration, and Q5 is the signal-to-noise ratio of the detection signal.
[0048] In one embodiment, the data collected by each VFTO time-series feature structured unit is determined as partial discharge induced source data based on the causal correlation quantization value, and the corresponding causal determination rule is determined as follows: determining whether the causal correlation quantization value is greater than or equal to a first preset threshold. If yes, the VFTO data is directly determined to be partial discharge induced source data; otherwise, it is determined to be suspected induced source data or non-induced source data. If the data is determined to be suspected induced source data, and the causal correlation quantification value is between the second preset threshold and the first preset threshold, then the core feature part of the VFTO time series feature structured unit is retained, and the remaining feature parts are fully fused and verified with the optimally matched effective time series feature unit of partial discharge. If the causal correlation quantification value is less than the second preset threshold, the VFTO collected data is determined to be non-inducing source data, and the VFTO time-series feature structured unit and the effective time-series feature unit of partial discharge are marked as having no causal correlation.
[0049] In a specific embodiment, a first preset threshold is set to 0.7 and a second preset threshold is set to 0.4. If CausalScore ≥ 0.7, it is directly determined as induced source data without additional verification and is directly included in the equipment insulation degradation early warning queue. Otherwise, the corresponding causal determination rule is as follows: if 0.4 ≤ CausalScore < 0.7, the core feature segments with the first 20% of the amplitude in the VFTO timing feature structured unit are retained, and the remaining parts are time-aligned and verified for entity consistency with the optimally matched effective timing feature unit of partial discharge; if CausalScore < 0.4, it is determined as non-induced source data, marked as interference or inherent defect discharge, and excluded from the degradation early warning queue.
[0050] Furthermore, if there are multiple consecutive highly correlated units (CausalScore ≥ 0.7), by default only the first unit is marked as the core triggering event, and subsequent units are marked as chain triggering events, merging them to generate a continuous degradation process chain to avoid duplicate warnings.
[0051] If there are multiple consecutive low-correlation units (CausalScore≤0.4), they are marked as non-induced events. The same type of interfering events are merged by using "temporal clustering + working condition label matching" to ensure the simplicity and readability of the causal determination results.
[0052] In one embodiment, based on feature matching criteria, the effective partial discharge timing feature units that best match the VFTO timing feature structured unit are selected, including: Calculate the coupling degree of the timing characteristics between the VFTO timing characteristic structured unit and each effective timing characteristic unit of partial discharge, as well as the fit degree of the operating condition entity. Based on the coupling degree of timing features and the fit of the operating conditions, the effective timing feature unit of partial discharge that best matches the VFTO timing feature structured unit is selected from multiple effective timing feature units of partial discharge.
[0053] The matching process is divided into two core stages. The first stage prioritizes the calculation of temporal feature coupling degree and completes the preliminary screening based on the correlation at the temporal level.
[0054] This stage constructs a unified multi-physical quantity temporal feature space, mapping VFTO transient features and partial discharge features to the same metric coordinate system, providing a quantitative basis for the temporal coupling analysis of the two.
[0055] For each target VFTO time-series feature structured unit, it needs to be aligned one by one with the set of candidate partial discharge effective time-series feature units. The specific processing flow is as follows: Temporal Feature Vectorization: A coding model combining wavelet packet decomposition and temporal convolutional networks is used to structure the target VFTO temporal features into units. Convert to a fixed-dimensional temporal feature vector This vector integrates the core features of VFTO, such as amplitude abrupt changes, frequency domain energy distribution, and time series continuity.
[0056] Discharge feature vectorization: effective temporal feature units for candidate partial discharges The key sampling points and feature segments are extracted and input into a time-frequency joint coding network to obtain the corresponding discharge feature vector. This vector characterizes the core properties of partial discharge, such as pulse intensity, phase distribution, and energy concentration.
[0057] Temporal coupling degree calculation: Measured using the Dynamic Time Warping (DTW) algorithm. and The temporal similarity between them is used to obtain the temporal coupling degree after normalizing the DTW distance. Select The first k candidate units with a value ≥0.7 constitute the initial matching set V.
[0058] The normalized calculation method for temporal coupling degree is as follows: Among them, DTW ( , )express and The dynamic time-warped distance between them, where DTWmax is the preset maximum tolerable distance threshold. ∈[0,1].
[0059] To avoid highly similar redundant feature units in the initial matching set V, this stage introduces the Maximum Diversity Relevance (MDR) index to select candidate units with better diversity while ensuring temporal coupling. The specific process is as follows: taking set V as input, initialize the selected matching unit group S, and iteratively select the unit with the highest MDR score from V to add to S until the preset number is met or all core features are covered.
[0060] For candidate units ∈V, the calculation logic for the MDR score is as follows: Among them, the weighting coefficients of α and β (α+β=1) Candidate unit With selected unit The cosine similarity between units is used to measure the degree of redundancy between units.
[0061] For example, by setting α=0.75 and β=0.25, this configuration can effectively improve the diversity of candidate units while ensuring temporal coupling, and avoid over-focusing on a single feature pattern.
[0062] After completing the initial matching and redundancy removal, the second stage of the working condition entity fit calculation is carried out. Based on the consistency between the equipment operating conditions and the physical scene, the final matching and screening is completed.
[0063] This phase establishes consistency constraints between VFTO and partial discharge characteristics to ensure that the matched discharge event and VFTO occur in the same equipment, under the same operating scenario, and at the same voltage level, thus avoiding misclassification of unrelated discharges under different operating conditions as induced discharges. For example, if the target VFTO originates from the tripping operation of a 500kV GIS disconnector, then partial discharge characteristic units of the same voltage level, equipment type, and operation type will be matched first.
[0064] First, VFTO terminal working condition entity recognition is carried out: for the target VFTO time-series feature structured unit, combined with the equipment operation log and collected configuration information, the associated working condition entity is extracted.
[0065] The input data includes equipment ledgers, operation records, and operating condition tags collected by VFTO. Through a rule engine and a lightweight classification model, the core operating condition entities associated with the VFTO unit are identified, covering categories such as equipment type, voltage level, operation type, installation location, and insulation defect type. The corresponding operating condition entity list is then output.
[0066] ; Where M represents the number of operating entities associated with the VFTO event, ei is the operating entity string (such as "500kVGIS" or "isolating switch tripping operation"), ci is the entity category (equipment, operation, location, defect, etc.), and si is the operating confidence of the entity, representing the strength of its association with the current VFTO event.
[0067] Subsequently, partial discharge end condition entity identification is carried out: for each effective temporal feature unit of partial discharge in the preliminary matching set, its associated end condition entity is extracted.
[0068] To accurately characterize the operating conditions of partial discharge feature units, this application constructs a three-level operating condition entity perception framework of "equipment-operation-physical quantity," identifying entity information associated with discharge events from multiple dimensions, specifically including: Equipment level identification: This is used to identify the equipment type and voltage level to which the partial discharge signal belongs. It adopts a classification network based on equipment tags and outputs the equipment category (such as GIS, switch cabinet, circuit breaker), voltage level value and corresponding confidence score.
[0069] Operation level identification: Used to identify the type of equipment operation when partial discharge occurs. Through the operation event clustering network, it outputs operation category labels (such as disconnecting switch opening, circuit breaker closing, grounding switch operation, etc.) to align with the operation type entity at the VFTO end, ensuring consistency of operating conditions.
[0070] Physical quantity hierarchical identification: used to extract the core physical quantity features of partial discharge. Through time-frequency domain feature extraction network, key parameters such as pulse peak amplitude, phase concentration, and high-frequency energy ratio are identified to construct high-order causal matching features (such as high-frequency VFTO corresponding to high-frequency partial discharge pulse) to avoid matching irrelevant discharge units with different physical quantity features.
[0071] The key sampling point sequence of each effective temporal characteristic unit of partial discharge is input into the above three-level working condition entity perception framework. Each sampling point contains multiple physical quantity data such as amplitude, frequency, and phase. The model outputs the working condition entity set of the corresponding unit. ; in, This indicates the number of operating entities associated with the partial discharge event. oj is the name of the operating entity (e.g., "500kV GIS" or "disconnector switch opening"), cj is the entity category (equipment, operation, physical quantity), wj is the degree of causal association between the entity and the target VFTO unit, and pj is the range of physical quantity parameters corresponding to the entity (e.g., amplitude range or phase range).
[0072] Based on the set of operating condition entities at the VFTO end and the partial discharge end, the degree of fit between the two operating condition entities is calculated. Select Units with a value ≥0.8 constitute the final matching set.
[0073] The specific calculation process is as follows: Collect the VFTO terminal working condition entities. Partial discharge terminal operating condition entity set Grouping by entity category c (such as equipment, operation, location, physical quantity) yields a subset of entities under each category. and .
[0074] For each type of entity c, calculate the matching degree Match(c) for that type: like and If there exists at least one semantically consistent entity pair (ei,oj) (e.g., “500kV GIS” and “500kV GIS”, “isolating switch opening” and “isolating switch opening operation”), then: Match(c) = si wj in which: si: The operating condition confidence level of entity ei in the middle; wj: The degree of causal relationship between the entities in the online judge (oj); If there is no semantically consistent pair of entities, then Match(c) = 0.
[0075] Weighted fusion yields the overall fit: based on preset category weights. (satisfy =1, where the operation and equipment categories, which are more closely related to the causal relationship, have higher weights. The total working condition entity fit is obtained by weighted summation of the matching degrees of each category: = Match(c) where ∈[0,1], the higher the value, the stronger the consistency of the working condition entity.
[0076] After completing the dual matching, the timing coupling verification stage begins: verifying whether the occurrence timing of the optimally matched partial discharge unit is within a 0-10ms window after the target VFTO unit, which corresponds to the typical delay range of VFTO-induced partial discharge. If the candidate unit timing exceeds this range, timing alignment is performed through feature interpolation or boundary truncation to ensure that the causal timing relationship between the two is valid.
[0077] This application embodiment achieves precise matching between VFTO and partial discharge feature units by constructing a triple matching constraint of "temporal feature coupling + operating condition entity matching + temporal coupling verification", effectively distinguishing between induced discharge and interference / inherent defect discharge, and significantly improving the accuracy and reliability of VFTO-induced partial discharge causal determination.
[0078] In one embodiment, based on the causal determination rule, the feature fusion verification of the optimally matched effective partial discharge timing feature unit and the corresponding VFTO timing feature structured unit includes: Determine the timing coupling anchor point between the optimally matched effective timing feature unit of partial discharge and the VFTO timing feature structured unit; the timing coupling anchor point includes one of the VFTO signal peak frame, the partial discharge signal start frame, and the device operation trigger frame; Based on the timing coupling anchor point and causal determination rule, the optimally matched effective timing feature unit of partial discharge is fused and verified with the corresponding VFTO timing feature structured unit.
[0079] To prevent timing misalignment, causal misjudgment, and interference signal confusion, this application introduces the following timing anchoring mechanism in feature fusion verification: Timing coupling anchor points are preferentially selected from frames with clear physical causal triggers and well-defined signal characteristics, avoiding strong electromagnetic interference segments and noise spurious peak segments, and eliminating abnormal amplitude jump points to ensure clear timing correlation and accurate causal determination. All timing coupling anchor points are preferentially aligned to the VFTO peak trigger, partial discharge initiation, and equipment operation command issuance time; if fusion verification needs to be performed midway, interference pulse segments and abnormal amplitude segments are avoided. The timing alignment logic uses signal amplitude mutations and frequency domain energy shifts to determine timing coupling anchor points, ensuring no misaligned causal correlations during fusion verification.
[0080] Fusion verification method control: Introducing timing alignment calibration and phase synchronization correction mechanism to achieve precise timing coupling of VFTO characteristics and partial discharge characteristics, thereby improving the reliability of causal determination and the accuracy of degradation analysis.
[0081] For various scenarios requiring feature fusion verification, appropriate temporal coupling anchor points are located based on the aforementioned temporal anchoring mechanism. After location, the fusion verification method is determined according to the correlation between VFTO feature units and partial discharge feature units and the causal determination rules. If the optimally matched effective timing feature unit of partial discharge is used as the "induction initiation verification", an anchor point pre-verification is adopted: the device operation trigger frame is calibrated first, and VFTO and partial discharge features are then synchronously aligned to enhance the traceability and continuity of causal triggering. The anchor point pre-verification method is to lock the operation trigger time first, and then align VFTO and discharge features. Technical effect: First, the initial trigger signal of device operation is locked, and then the subsequent VFTO impact and partial discharge response are aligned to construct a complete causal traceability chain of "operation → impact → discharge", strengthening the credibility of the induction relationship.
[0082] If the optimally matched effective timing feature unit of partial discharge is used as the "discharge response verification", an anchor point post-verification is adopted: first, the VFTO peak frame is locked, and then the partial discharge start frame is aligned to help delay the completion of causal verification; the anchor point post-verification method is that the VFTO peak feature is locked first, and the partial discharge feature is aligned subsequently. The system first captures the VFTO impact peak, and then aligns the subsequent discharge response signal. The technical effect is that the system first locates the moment of the strongest insulation impact, and then tracks the subsequent discharge event, smoothly connecting the causal link of "impact → response" and strengthening the continuity of the degradation mechanism.
[0083] If the optimally matched effective timing feature unit of partial discharge is used as "mid-segment interference investigation", the timing coupling anchor point is forcibly judged to be the starting frame of the partial discharge signal (by the dual judgment of discharge amplitude > 3 times the baseline + discharge phase concentrated in the power frequency peak segment).
[0084] This application automatically selects the fusion verification method based on the verification scenario of the optimally matched partial discharge unit, realizing a "temporal anchoring + feature fusion" causal verification strategy to improve the temporal accuracy and correlation reliability of causal determination. All temporal coupling anchor points must ensure that the starting frame of the partial discharge unit does not contain noise spurious peaks to avoid the causal jump that appears as if "the connection is broken immediately after the impact".
[0085] In one embodiment, before performing signal preprocessing on each UHF partial discharge acquisition data to obtain multiple effective timing feature units of partial discharge, the method further includes: performing validity and anti-interference verification on each UHF partial discharge acquisition data, and removing noise frames, invalid acquisition frames and interference signal frames.
[0086] Specifically, to ensure the accuracy and reliability of VFTO-induced partial discharge causality determination, each set of UHF partial discharge acquisition data must undergo multi-dimensional validity and anti-interference verification before entering the preprocessing process. First, signal quality verification is performed, detecting amplitude anomalies, noise baseline shifts, missed sampling points, and synchronization deviations in the sampled data. Next, interference feature identification is performed, using frequency domain spectrum analysis to identify periodic electromagnetic interference, white noise, and environmental noise components. Finally, spurious peak elimination is completed, distinguishing between genuine partial discharges and interference spurious peaks based on pulse morphology and phase characteristics. If no qualified partial discharge acquisition data is found after verification, a preset benchmark interference feature library is called for comparison, marking it as a non-induced interference event and excluding it from subsequent analysis. If the partial discharge acquisition data has problems such as sampling anomalies, noise overload, or interference aliasing, the corresponding sampling points are automatically removed or the entire set of acquisition data is directly deleted.
[0087] Furthermore, this embodiment also discloses a system for applying the aforementioned causal determination method for VFTO-induced partial discharge based on time-series features. The system includes: a data acquisition module for acquiring VFTO detection datasets and UHF partial discharge detection datasets synchronously acquired by an integrated sensing device; wherein the VFTO detection dataset includes multiple sets of VFTO acquisition data, and the UHF partial discharge detection dataset includes multiple sets of UHF partial discharge acquisition data; a signal preprocessing module for performing signal preprocessing on each VFTO acquisition data to obtain multiple VFTO time-series feature structured units; and performing signal preprocessing on each UHF partial discharge acquisition data to obtain multiple effective partial discharge time-series feature units.
[0088] The correlation quantization module is used to determine the causal correlation quantization value between each VFTO time-series feature structured unit and the corresponding partial discharge characteristics based on the correlation quantization criterion; and to determine whether the collected data corresponding to each VFTO time-series feature structured unit is partial discharge induced source data based on the causal correlation quantization value, while determining the corresponding causal determination rule; the feature matching module is used to select the effective partial discharge time-series feature unit that best matches each VFTO time-series feature structured unit if it is determined to be suspected induced source data, based on the feature matching criterion.
[0089] The causal determination module is used to perform feature fusion verification between the optimally matched effective timing feature unit of partial discharge and the corresponding VFTO timing feature structured unit based on the causal determination rules; and to perform comprehensive analysis on the collected data after each fusion verification, and output the causal determination conclusion of VFTO-induced partial discharge.
[0090] This embodiment also provides a storage medium storing a computer program. When executed by a processor, the program implements the causal determination method for VFTO-induced partial discharge based on timing characteristics as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.
[0091] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A causal determination method for VFTO-induced partial discharge based on time-series characteristics, characterized in that, include: Acquire VFTO detection dataset and UHF partial discharge detection dataset synchronously collected by integrated sensing device; wherein: the VFTO detection dataset includes multiple sets of VFTO collected data, and the UHF partial discharge detection dataset includes multiple sets of UHF partial discharge collected data. The VFTO acquisition data is preprocessed to obtain multiple VFTO timing feature structured units; and the UHF partial discharge acquisition data is preprocessed to obtain multiple effective partial discharge timing feature units. Based on the correlation quantification criterion, the causal correlation quantification value between each VFTO time-series feature structured unit and the corresponding partial discharge feature of the time period is determined; and based on the causal correlation quantification value, it is determined whether the collected data corresponding to each VFTO time-series feature structured unit is partial discharge induced source data, and the corresponding causal determination rule is determined. If the data is determined to be suspected induced source data, then based on the feature matching criterion, the effective partial discharge timing feature unit that best matches the VFTO timing feature structured unit is selected. Based on the causal determination rule, the optimally matched effective timing feature unit of partial discharge and the corresponding VFTO timing feature structured unit are subjected to feature fusion verification; and the collected data after each fusion verification are comprehensively analyzed to output the causal determination conclusion of VFTO-induced partial discharge.
2. The causal determination method for VFTO-induced partial discharge based on time-series characteristics according to claim 1, characterized in that: The VFTO acquisition data are preprocessed to obtain multiple VFTO time-series feature structured units, including: The VFTO-acquired data is demodulated and its features are extracted to obtain the corresponding time-series feature information; The time-series feature information is divided into time periods based on time-domain and frequency-domain features to obtain multiple feature units; Each of the aforementioned feature units is assigned a corresponding detection timestamp and signal quality characterization parameters to obtain multiple VFTO time-series feature structured units.
3. The causal determination method for VFTO-induced partial discharge based on time-series characteristics according to claim 1, characterized in that: Signal preprocessing is performed on the ultra-high frequency partial discharge acquisition data to obtain multiple effective temporal feature units of partial discharge, including: Calculate the inter-frame temporal feature deviation of each of the ultra-high frequency partial discharge acquisition data, and determine whether the corresponding partial discharge acquisition data has feature segmentation nodes based on the inter-frame temporal feature deviation; If so, the corresponding partial discharge acquisition data is segmented based on the feature segmentation nodes; the average signal amplitude and effective acquisition duration of each partial discharge acquisition data are calculated. Based on preset amplitude thresholds and preset duration thresholds, multiple sets of partial discharge acquisition data are filtered to obtain multiple effective temporal feature units of partial discharge.
4. The causal determination method for VFTO-induced partial discharge based on time-series characteristics according to claim 1, characterized in that: Based on the correlation quantification criterion, the causal correlation quantification values between each VFTO time-series feature structured unit and the corresponding time-period partial discharge characteristics are determined as follows: Based on each of the VFTO time-series feature structured units, the corresponding VFTO feature saliency and the proportion of key frequency domain components are determined. Based on the partial discharge acquisition data of each VFTO time-series structured unit corresponding to the time period, the corresponding partial discharge intensity, discharge phase concentration, and detection signal signal-to-noise ratio are determined. The causal correlation quantification value between VFTO and partial discharge is obtained by weighting the VFTO feature saliency, the proportion of key frequency domain components, the partial discharge intensity, the discharge phase concentration, and the signal-to-noise ratio of the detection signal.
5. The causal determination method for VFTO-induced partial discharge based on time-series characteristics according to claim 4, characterized in that: Based on the causal correlation quantification value, it is determined whether the collected data corresponding to each VFTO time-series feature structured unit is partial discharge induced source data, and the corresponding causal determination rule is determined, including: determining whether the causal correlation quantification value is greater than or equal to a first preset threshold. If yes, the VFTO data is directly determined to be partial discharge induced source data; otherwise, it is determined to be suspected induced source data or non-induced source data. If the data is determined to be suspected induced source data, and the causal correlation quantification value is between the second preset threshold and the first preset threshold, then the core feature part of the VFTO time series feature structured unit is retained, and the remaining feature parts are fully fused and verified with the optimally matched effective time series feature unit of partial discharge. If the causal correlation quantification value is less than the second preset threshold, the VFTO collected data is determined to be non-inducing source data, and the VFTO time-series feature structured unit and the effective time-series feature unit of partial discharge are marked as having no causal correlation.
6. The causal determination method for VFTO-induced partial discharge based on time-series characteristics according to claim 1, characterized in that: Based on the feature matching criterion, the effective partial discharge timing feature units that best match the VFTO timing feature structured unit include: Calculate the timing feature coupling degree and the operating condition entity fit degree between the VFTO timing feature structured unit and each of the partial discharge effective timing feature units; Based on the timing feature coupling degree and the operating condition entity fit degree, the effective timing feature unit of partial discharge that best matches the VFTO timing feature structured unit is selected from multiple effective timing feature units of partial discharge.
7. The causal determination method for VFTO-induced partial discharge based on time-series characteristics according to claim 5, characterized in that: Based on the aforementioned causal determination rule, the feature fusion verification of the optimally matched effective partial discharge timing feature unit and the corresponding VFTO timing feature structured unit includes: Determine the timing coupling anchor point between the optimally matched effective timing feature unit of partial discharge and the VFTO timing feature structured unit; the timing coupling anchor point includes one of the VFTO signal peak frame, the partial discharge signal start frame, and the device operation trigger frame; Based on the timing coupling anchor point and the causal determination rule, the optimally matched effective timing feature unit of partial discharge and the corresponding VFTO timing feature structured unit are subjected to feature fusion verification.
8. The causal determination method for VFTO-induced partial discharge based on time-series characteristics according to claim 3, characterized in that: Before performing signal preprocessing on each of the UHF partial discharge acquisition data to obtain multiple effective timing feature units of partial discharge, the method further includes: performing validity and anti-interference verification on each of the UHF partial discharge acquisition data, and removing noise frames, invalid acquisition frames and interference signal frames.
9. A determination system applied to the causal determination method for VFTO-induced partial discharge based on time-series characteristics as described in any one of claims 1 to 7, characterized in that: The system includes: a data acquisition module, used to acquire VFTO detection datasets and UHF partial discharge detection datasets synchronously collected by the integrated sensing device; wherein: the VFTO detection dataset includes multiple sets of VFTO acquired data, and the UHF partial discharge detection dataset includes multiple sets of UHF partial discharge acquired data; The signal preprocessing module is used to preprocess the VFTO acquisition data to obtain multiple VFTO timing feature structured units; and to preprocess the UHF partial discharge acquisition data to obtain multiple effective timing feature units of partial discharge. The correlation quantization module is used to determine the causal correlation quantization value between each VFTO time-series feature structured unit and the corresponding partial discharge feature of the time period based on the correlation quantization criterion; and to determine whether the collected data corresponding to each VFTO time-series feature structured unit is partial discharge induced source data based on the causal correlation quantization value, and at the same time determine the corresponding causal determination rule. The feature matching module is used to select the effective partial discharge timing feature unit that best matches each of the VFTO timing feature structured units if the data is determined to be suspected inducing source data, based on the feature matching criteria. The causal determination module is used to perform feature fusion verification between the optimally matched effective timing feature unit of partial discharge and the corresponding VFTO timing feature structured unit based on the causal determination rules; and to perform comprehensive analysis on the collected data after each fusion verification, and output the causal determination conclusion of VFTO-induced partial discharge.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the causal determination method for VFTO-induced partial discharge based on timing characteristics as described in any one of claims 1 to 8.