Partial discharge monitoring method and device based on multi-cycle data superposition

Abstract

The application relates to the technical field of power equipment monitoring, and in particular to a partial discharge monitoring method and device based on multi-cycle data superposition. The method comprises the following steps: periodically collecting partial discharge monitoring signals and operating voltage signals of power equipment; for each partial discharge monitoring point, taking the voltage zero-crossing point of the operating voltage signal as a reference, the partial discharge monitoring signals corresponding to each collection cycle are respectively phase-aligned; for each partial discharge monitoring point, the phase-aligned partial discharge monitoring signals of the multiple collection cycles are superposed; based on the superposed partial discharge monitoring signals, a partial discharge phase distribution diagram corresponding to each partial discharge monitoring point is constructed; and the partial discharge phase diagram is input into a pre-trained classification and recognition model to obtain a partial discharge recognition result. Through superposition of the data of multiple collection cycles of the same partial discharge monitoring point, the effective signal capture time is prolonged from the second level of a single cycle to the hour level of multiple cycles, and the detection rate of early sporadic partial discharge signals is significantly improved.

Description

Technical Field

[0001] This application relates to the field of power equipment monitoring technology, and in particular to a partial discharge monitoring method and device based on multi-cycle data superposition. Background Technology

[0002] Partial discharge (PD) refers to a non-penetrating discharge phenomenon that occurs inside or on the surface of the insulation of power equipment. It is an important characteristic signal reflecting the deterioration of the insulation of power equipment. By monitoring PD signals online, early insulation defects of equipment can be detected in time, thus avoiding serious failures.

[0003] In existing online partial discharge (PD) monitoring technologies, to balance power consumption and standby time, the sensor's acquisition interval is typically long (e.g., 5 min, 10 min, 30 min, 1 hour or more), while the acquisition window of a single acquisition cycle is extremely short (generally only about 1 second). This acquisition mode has significant drawbacks: early PD signals are intermittent and short-lived, and the long acquisition interval and extremely short acquisition window lead to the missed detection of most intermittent PD signals. Therefore, existing online PD monitoring technologies suffer from the problem of missing early intermittent PD signals. Summary of the Invention

[0004] This application provides a partial discharge monitoring method and apparatus based on multi-cycle data superposition to solve the problem of missed detection of early intermittent partial discharge signals.

[0005] In a first aspect, embodiments of this application provide a partial discharge monitoring method based on multi-cycle data superposition. The method includes: periodically acquiring partial discharge monitoring signals and operating voltage signals of power equipment; wherein, the partial discharge sampling points of the partial discharge monitoring signals and the voltage sampling points of the operating voltage signals are synchronously acquired based on the same preset sampling rate; the power equipment includes at least one partial discharge monitoring point, and the partial discharge monitoring signal is obtained by monitoring the partial discharge monitoring point; for each partial discharge monitoring point, the partial discharge monitoring signals corresponding to each acquisition cycle are phase aligned based on the voltage zero-crossing point of the operating voltage signal; for each partial discharge monitoring point, the phase-aligned partial discharge monitoring signals of multiple acquisition cycles are superimposed; based on the superimposed partial discharge monitoring signals, a partial discharge phase distribution map corresponding to each partial discharge monitoring point is constructed; and the partial discharge phase map is input into a pre-trained classification and recognition model to obtain the partial discharge recognition result.

[0006] In one possible implementation, the step of aligning the phase of the partial discharge monitoring signal corresponding to each acquisition cycle, using the zero-crossing point of the operating voltage signal as a reference, includes: extracting the reference voltage zero-crossing point of the operating voltage signal within each acquisition cycle, wherein, in each acquisition cycle, when the voltage amplitude of one voltage sampling point is negative and the voltage amplitude of its adjacent next voltage sampling point is positive, the first voltage sampling point with a positive voltage amplitude is the reference voltage zero-crossing point; determining the sampling sequence number of each reference voltage zero-crossing point as the zero-crossing point sequence number corresponding to each acquisition cycle; determining the phase alignment point of the partial discharge monitoring signal corresponding to each acquisition cycle based on the zero-crossing point sequence number corresponding to each acquisition cycle; wherein, the phase alignment point is the partial discharge sampling point in the partial discharge monitoring signal whose sampling point sequence number is the same as the zero-crossing point sequence number; for the partial discharge monitoring signal corresponding to each acquisition cycle, truncating the partial discharge sampling points before the phase alignment point and splicing them to the end to obtain the phase-aligned partial discharge monitoring signal corresponding to each acquisition cycle.

[0007] In one possible implementation, before extracting the reference voltage zero-crossing point of the operating voltage signal in each acquisition cycle, the method further includes: filtering the operating voltage signal in each acquisition cycle, wherein the filtering process includes one or a combination of 50Hz notch filtering and low-pass filtering.

[0008] In one possible implementation, superimposing the partial discharge monitoring signals of multiple acquisition cycles after phase alignment for each partial discharge monitoring point includes: superimposing the amplitudes of all partial discharge sampling points in the same phase interval for the phase-aligned partial discharge monitoring signals corresponding to each acquisition cycle of each partial discharge monitoring point; wherein, the phase interval is pre-divided based on the voltage phase range; the amplitude superposition includes any one of accumulation, maximum value superposition, or average value superposition.

[0009] In one possible implementation, based on the superimposed partial discharge monitoring signal, a partial discharge phase distribution map corresponding to each partial discharge monitoring point is constructed, including: for each partial discharge monitoring point, a phase-resolved partial discharge PRPD map is constructed with voltage phase as the abscissa and the amplitude of the superimposed partial discharge monitoring signal as the ordinate.

[0010] In one possible implementation, the partial discharge identification result includes whether the power equipment has partial discharge and the type of partial discharge, and the partial discharge type includes at least one of particle discharge, surface discharge, suspension discharge, corona discharge, and air gap discharge.

[0011] In one possible implementation, the partial discharge monitoring signal and the operating voltage signal correspond to the same acquisition period and acquisition window; the length of the acquisition window is shorter than the length of the acquisition period.

[0012] In one possible implementation, the training steps of the classification and recognition model include: collecting PRPD map samples of different types of partial discharges to construct a training dataset; performing data augmentation on the training dataset, including at least one of rotation, scaling, and noise addition; and inputting the data-augmented training dataset into a convolutional neural network for training to obtain the classification and recognition model.

[0013] In one possible implementation, the partial discharge monitoring signal is acquired by an ultrasonic sensor or an ultra-high frequency sensor, and the operating voltage signal is acquired by a voltage transformer; the power equipment is at least one of a transformer, a gas-insulated switchgear, a switch cabinet, and a cable.

[0014] Secondly, this application provides a partial discharge monitoring device based on multi-period data superposition. The device includes: a data acquisition module configured to periodically acquire partial discharge monitoring signals and operating voltage signals of power equipment; wherein, the partial discharge sampling points of the partial discharge monitoring signals and the voltage sampling points of the operating voltage signals are synchronously acquired based on the same preset sampling rate; the power equipment includes at least one partial discharge monitoring point, and the partial discharge monitoring signal is obtained by monitoring the partial discharge monitoring point; The phase alignment module is configured to: for each partial discharge monitoring point, perform phase alignment on the partial discharge monitoring signals corresponding to each acquisition cycle, based on the zero-crossing point of the operating voltage signal; the overlay module is configured to: overlay the phase-aligned partial discharge monitoring signals from multiple acquisition cycles for each partial discharge monitoring point; the plotting module is configured to: construct a partial discharge phase distribution map corresponding to each partial discharge monitoring point based on the overlaid partial discharge monitoring signals; and the identification module is configured to: input the partial discharge phase map into a pre-trained classification and identification model to obtain the partial discharge identification result.

[0015] As can be seen from the above, this application provides a method and apparatus for partial discharge monitoring based on multi-cycle data superposition. The method includes: periodically acquiring partial discharge monitoring signals and operating voltage signals from power equipment; wherein the partial discharge sampling points of the partial discharge monitoring signals and the voltage sampling points of the operating voltage signals are synchronously acquired based on the same preset sampling rate; the power equipment includes at least one partial discharge monitoring point, and the partial discharge monitoring signal is obtained by monitoring the partial discharge monitoring point; for each partial discharge monitoring point, using the voltage zero-crossing point of the operating voltage signal as a reference, the phase alignment of the partial discharge monitoring signals corresponding to each acquisition cycle is performed; for each partial discharge monitoring point, the phase-aligned partial discharge monitoring signals from multiple acquisition cycles are superimposed; based on the superimposed partial discharge monitoring signals, a partial discharge phase distribution map corresponding to each partial discharge monitoring point is constructed; the partial discharge phase map is input into a pre-trained classification and recognition model to obtain the partial discharge recognition result. It can be seen that this application, by superimposing data from multiple acquisition cycles of the same monitoring point, extends the effective signal capture time from the seconds of a single cycle to the hours of multiple cycles, significantly improving the detection rate of early intermittent partial discharge signals. Attached Figure Description

[0016] Figure 1 A flowchart illustrating the partial discharge monitoring method based on multi-period data overlay provided in this application embodiment; Figure 2 This is a schematic diagram illustrating the acquisition of partial discharge monitoring signals in an embodiment of this application; Figure 3 A schematic diagram illustrating phase alignment of partial discharge monitoring signals provided in an embodiment of this application; Figure 4 A schematic diagram of the PRPD provided for an embodiment of this application; Figure 5 This is a schematic diagram of the partial discharge monitoring device based on multi-period data superposition provided in an embodiment of this application. Detailed Implementation

[0017] To enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of this application.

[0018] First, the terminology used in the embodiments of this application will be introduced by way of example.

[0019] ① Partial Discharge (PD): This refers to a non-penetrating discharge phenomenon that occurs inside or on the surface of the insulation of power equipment. It is a key characteristic signal reflecting the deterioration of the equipment's insulation. When there are insulation defects inside power equipment, local electric field breakdown will occur at the defect under the action of operating voltage, generating a weak discharge pulse signal. Common types of PD include particulate discharge, surface discharge, suspended discharge, corona discharge, and air gap discharge. Different types of PD have different phase distribution characteristics and waveform characteristics.

[0020] ② Phase angle: The electrical angle of a sinusoidal alternating current at a certain moment, reflecting the instantaneous position of the voltage / current signal within the period, and is measured in degrees. It is a key parameter for analyzing the phase correlation between partial discharge and voltage.

[0021] ③ Power frequency cycle: The voltage cycle corresponding to the 50Hz frequency of the power grid has a duration of 20ms. That is to say, when the power grid frequency is 50Hz, the corresponding power frequency cycle is 20ms, which means that there are 50 complete power frequency cycles in 1 second. Each cycle corresponds to a phase angle change of 0°-360°, that is, the voltage phase range is 0°-360°.

[0022] ④ Phase Resolved Partial Discharge Pattern (PRPD) Diagram: Also known as a partial discharge phase distribution diagram, the PRPD diagram uses the operating voltage phase (0°-360°) of the power equipment as the horizontal axis and the amplitude of the partial discharge signal as the vertical axis, visually presenting the correlation between the partial discharge signal and the voltage phase. The PRPD diagram is the core analytical tool for partial discharge feature identification and fault diagnosis. Different types of partial discharge (such as surface discharge and particulate discharge) exhibit different phase distribution patterns in the PRPD diagram.

[0023] ⑤ Voltage zero-crossing point: refers to the critical point in a 50Hz sinusoidal voltage circuit where the amplitude changes from negative to positive (positive zero-crossing point) or from positive to negative (negative zero-crossing point). In one complete cycle (360°) of a sinusoidal alternating current, there are two zero-crossing points: the positive zero-crossing point at 0° (changing from negative to positive) and the negative zero-crossing point at 180° (changing from positive to negative).

[0024] ⑥ Convolutional Neural Network (CNN): A deep learning model that excels at image feature extraction and classification. CNN automatically extracts local features from images through convolutional layers, reduces feature dimensionality through pooling layers, and performs classification through fully connected layers.

[0025] To address the issue of missed detection of early intermittent partial discharge signals in existing online partial discharge monitoring technologies, this application provides a partial discharge monitoring method and apparatus based on multi-cycle data superposition. This method, without altering the existing sensor acquisition mode (i.e., without increasing device power consumption or shortening standby time), constructs a "virtual long window" (i.e., an equivalent long-duration acquisition window formed through data processing, rather than an actual extension of the acquisition time) by superimposing multi-cycle data. This extends the effective signal capture time from 1 second in a single cycle to the cumulative duration of multiple cycles (e.g., 24 hours), fundamentally solving the problem of missed detection of early intermittent partial discharge signals due to an excessively short capture window.

[0026] Figure 1 This is a flowchart illustrating the partial discharge monitoring method based on multi-period data overlay provided in an embodiment of this application.

[0027] like Figure 1 As shown, the partial discharge monitoring method based on multi-period data superposition provided in this application includes the following steps S100-S200.

[0028] S100: Periodically collects partial discharge monitoring signals and operating voltage signals of power equipment; wherein, the partial discharge sampling points of the partial discharge monitoring signals and the voltage sampling points of the operating voltage signals are collected synchronously based on the same preset sampling rate.

[0029] In addition, the power equipment includes at least one partial discharge monitoring point, and the partial discharge monitoring signal is obtained by monitoring the partial discharge monitoring point.

[0030] The partial discharge (PD) monitoring signal is acquired by a PD sensor, specifically an ultrasonic sensor or an ultra-high frequency (UHF) sensor. The operating voltage signal is acquired by a voltage transformer; the power equipment includes at least one of the following: transformer, gas-insulated switchgear, switchgear, and cable. It is understood that if a non-penetrating discharge phenomenon occurs inside or on the surface of the power equipment insulation, i.e., partial discharge, this discharge pulse can be acquired by the PD sensor, forming a PD monitoring signal. Specifically, the acquired data value is the ultrasonic or UHF signal intensity, expressed in dBmV or mV, characterizing the instantaneous energy of the discharge pulse. When constructing the PRPD diagram, these signal intensity values ​​correspond to the amplitude on the vertical axis.

[0031] Understandably, the operating voltage signal is a 50Hz sinusoidal alternating current, which can be used to determine voltage phase information.

[0032] For example, when the power equipment is a switchgear, the partial discharge sensor can be installed on the surface or inside the switchgear cabinet to acquire partial discharge monitoring signals, and the voltage transformer can be located on the incoming side or in the busbar room of the switchgear to acquire operating voltage signals.

[0033] Furthermore, the interior of power equipment can be divided into multiple partial discharge monitoring points, each equipped with a partial discharge sensor. For example, switchgear is typically divided into multiple independent compartments such as busbar compartments, circuit breaker compartments, and cable compartments, and partial discharge may occur in any of these compartments. To comprehensively monitor the overall insulation status of the switchgear, embodiments of this application can arrange partial discharge sensors in different compartments, thereby forming multiple partial discharge monitoring points. Each partial discharge sensor independently collects the partial discharge monitoring signal from its corresponding compartment (partial discharge monitoring point), and performs synchronous acquisition and subsequent phase alignment and superposition processing with the operating voltage signal, thereby independently assessing the insulation status of each compartment.

[0034] Furthermore, during the periodic acquisition of partial discharge monitoring signals and operating voltage signals, the partial discharge monitoring signals and operating voltage signals can correspond to acquisition periods and acquisition windows of the same length; the length of the acquisition window is shorter than the length of the acquisition period. For example, the acquisition period can be 5 minutes, 10 minutes, preferably 10 minutes, and the acquisition window can be 0.5 seconds or 1 second.

[0035] Figure 2 This is a schematic diagram of collecting partial discharge monitoring signals in an embodiment of this application.

[0036] like Figure 2 As shown, the preset sampling rate can be 3000Hz, 3600Hz, or other values, meaning that 60, 72, or other preset number of sampling points are collected within each power frequency cycle (20ms). Taking 72 sampling points per power frequency cycle as an example, this means 72 × 50 = 3600 sampling points per second.

[0037] For example, taking a 110kV gas-insulated switchgear (GIS) as an example, a sensor is installed on the surface or inside the GIS housing to capture partial discharge signals generated during equipment operation.

[0038] The data acquisition interval is set to 10 minutes, meaning data acquisition begins every 10 minutes. The acquisition window duration for a single acquisition cycle is set to 1 second. The power grid frequency is 50Hz, and the 1-second acquisition window contains 50 complete power frequency cycles. 72 sampling points are acquired within each power frequency cycle (20ms), therefore, 72 × 50 = 3600 partial discharge monitoring values ​​(partial discharge sampling points) are acquired per second, forming a one-dimensional array containing 3600 sampling points, denoted as: Pk=[pk,1,pk,2,…,pk,3600]; Where Pk represents the sampling point sequence corresponding to the partial discharge monitoring signal, pk represents a partial discharge sampling point, k represents the kth acquisition period (k=1,2,…,N), and N is the total number of acquisition periods.

[0039] In addition, the operating voltage signal of the power equipment is synchronously acquired through a voltage transformer. The operating voltage is a 50Hz sinusoidal alternating current, and the acquisition frequency is consistent with the acquisition frequency of the partial discharge monitoring signal, i.e., 3600 voltage sampling points per second, forming a one-dimensional array containing 3600 sampling points, denoted as: Vk=[vk,1,vk,2,…,vk,3600]; Where Vk represents the sampling point sequence corresponding to the operating voltage signal, vk represents a voltage sampling point, "vk,1" represents the first voltage sampling point, k represents the kth acquisition period (k=1,2,…,N), and N is the total number of acquisition periods.

[0040] In this embodiment of the application, the length of the acquisition window (1 second) is much shorter than the length of the acquisition cycle (10 minutes). This acquisition mode can effectively reduce device power consumption and extend standby time.

[0041] Furthermore, embodiments of this application can continuously collect data for multiple cycles, acquiring raw partial discharge monitoring signal data and corresponding operating voltage signal data for multiple acquisition cycles. For example, embodiments of this application can continuously collect data for 24 hours, acquiring raw partial discharge monitoring signal data and corresponding operating voltage signal data for a total of 144 acquisition cycles. That is, N=144, with each cycle containing 3600 sampling points. It is understood that the data acquisition process can be performed continuously in a loop, not limited to 24 hours.

[0042] S200: For each partial discharge monitoring point, the phase alignment of the partial discharge monitoring signal in each acquisition cycle is performed based on the zero-crossing point of the operating voltage signal.

[0043] It is understandable that, since partial discharge in power equipment is clearly related to voltage phase (e.g., surface discharge is more likely to occur at 90° and 270° phases), ideally, the partial discharge sensor should start acquiring data from phase 0 (0°) to ensure consistent starting phases for each acquisition cycle, facilitating subsequent multi-cycle data superposition. However, in actual acquisition, a fixed starting phase cannot be guaranteed, and different acquisition cycles may start from different phases. Directly superimposing multiple sets of data would lead to chaotic phase distribution and misrepresentation of partial discharge characteristics. In this embodiment, the operating voltage signal and the partial discharge monitoring signal are acquired synchronously in each acquisition cycle. The zero-crossing point is determined by analyzing the voltage waveform, and the phase of the partial discharge monitoring signal is aligned using the zero-crossing point as a reference. This ensures that the starting point of the partial discharge monitoring signal for all acquisition cycles corresponds to the zero-crossing point of the operating voltage signal, providing a consistent phase reference for subsequent superposition.

[0044] It is understood that when there are multiple partial discharge monitoring points, the embodiments of this application can perform phase alignment on the partial discharge monitoring signal corresponding to each partial discharge monitoring point.

[0045] As can be seen, in this embodiment of the application, after phase alignment based on the voltage zero-crossing point for the same partial discharge monitoring point, the partial discharge signals corresponding to the same phase position in each acquisition cycle can achieve accurate matching, ensuring that the same type of partial discharge features accurately overlap in the PRPD diagram when multiple cycles are superimposed, avoiding feature dispersion or false partial discharge features due to phase shift, thereby significantly improving the detection rate and identification accuracy of early intermittent partial discharge signals.

[0046] It is understood that, for a single partial discharge monitoring point, the embodiments of this application can perform phase alignment on the partial discharge monitoring signals within multiple (e.g., 144) acquisition cycles. In practical applications, at a fixed time point (e.g., midnight or 10:00 AM daily), the partial discharge monitoring signals from all acquisition cycles in the previous time period can be uniformly phase aligned, amplitude superimposed, and PRPD diagram constructed to output complete partial discharge monitoring results at regular intervals.

[0047] Specifically, for a partial discharge monitoring point, step S200 may include the following steps S201-S204.

[0048] S201: Extract the reference voltage zero-crossing point of the operating voltage signal in each acquisition cycle; wherein, in each acquisition cycle, when the voltage amplitude of one voltage sampling point is negative and the voltage amplitude of its adjacent next voltage sampling point is positive, the first voltage sampling point with a positive voltage amplitude is the reference voltage zero-crossing point.

[0049] For the same partial discharge monitoring point, for the operating voltage signal collected in each acquisition cycle, this embodiment of the application can use the method of "adjacent sampling point amplitude sign judgment" to calculate the reference voltage zero-crossing point corresponding to each acquisition cycle. Specifically, the smoothed operating voltage signal is traversed, and the amplitude sign of two adjacent voltage sampling points is judged in turn. When the voltage amplitude of a certain voltage sampling point is negative and the voltage amplitude of its next adjacent voltage sampling point is positive, it is determined that there is a positive zero-crossing point between the adjacent points, and the next voltage sampling point (i.e., the first voltage sampling point with a positive voltage amplitude) is determined as the voltage zero-crossing point (i.e., the reference voltage zero-crossing point). The sequence number X of the voltage sampling point in the sequence is recorded, and the sequence number X is the zero-crossing point number.

[0050] Taking a single data acquisition time of 1 second and a power grid frequency of 50Hz as an example, theoretically there are 50 positive zero-crossing points within 1 second. In this embodiment, the voltage zero-crossing point specifically refers to the first positive zero-crossing point. Since the initial phase of the acquisition window is not fixed, the position X of the first positive zero-crossing point in the sequence (operating voltage signal) also varies randomly and may be any value between 1 and 3600.

[0051] For example, if the first positive zero crossing occurs at the 180th voltage sampling point, then X = 180.

[0052] S202: Determine the sampling sequence number of each reference voltage zero-crossing point as the zero-crossing point sequence number corresponding to each acquisition cycle.

[0053] Specifically, for the same partial discharge monitoring point, the embodiments of this application can determine the zero-crossing index corresponding to each acquisition cycle. It is understood that this zero-crossing index indicates the index of the critical position of the voltage signal from negative to positive within each acquisition cycle in the sampling sequence, providing a unified reference for the phase alignment of subsequent partial discharge signals. The embodiments of this application can determine the zero-crossing index corresponding to each acquisition cycle.

[0054] S203: Based on the zero-crossing index corresponding to each acquisition cycle, determine the phase alignment point of the partial discharge monitoring signal corresponding to each acquisition cycle; wherein, the phase alignment point is the partial discharge sampling point in the partial discharge monitoring signal whose sampling point index is the same as the zero-crossing index.

[0055] Since the partial discharge monitoring signal and the operating voltage signal are synchronously acquired using the same preset sampling rate, there is a one-to-one correspondence between their sampling point numbers. Therefore, the partial discharge sampling point corresponding to the zero-crossing point number X is the phase alignment point. The sampling point number of this phase alignment point is the same as the zero-crossing point number. For the same partial discharge monitoring point, based on the zero-crossing point number corresponding to each acquisition cycle, the embodiments of this application can determine the phase alignment point corresponding to each acquisition cycle separately.

[0056] S204: For the partial discharge monitoring signal corresponding to each acquisition cycle, extract the partial discharge sampling points before the phase alignment point and splice them to the end to obtain the phase-aligned partial discharge monitoring signal corresponding to each acquisition cycle.

[0057] This step is used to perform cyclic shifting of the data, making the phase alignment point the starting point of the sequence (partial discharge monitoring signal). Specifically, for the same partial discharge monitoring point, the embodiments of this application can align the partial discharge monitoring signals corresponding to each acquisition cycle, that is, each acquisition cycle performs the steps of truncation and splicing to obtain the phase-aligned partial discharge monitoring signal corresponding to each acquisition cycle.

[0058] For example, for the same partial discharge monitoring point, taking a sampling frequency of 3600 Hz as an example, let the sampling point number corresponding to the first positive zero crossing be X. The original partial discharge monitoring signal sequence is: Pk=[pk,1,pk,2,…,pk,X−1,pk,X,pk,X+1,…,pk,3600]; Where Pk represents the sampling point sequence corresponding to the partial discharge monitoring signal, pk represents a partial discharge sampling point, "pk,1" represents the first sampling point, "pk,X" represents the Xth partial discharge sampling point, "pk,X+1" represents the (X+1)th partial discharge sampling point, and so on, which will not be elaborated here. k represents the kth acquisition period (k=1,2,…,N), and N is the total number of acquisition periods.

[0059] Figure 3 This is a schematic diagram of partial discharge monitoring signal phase alignment provided in an embodiment of this application.

[0060] like Figure 3 As shown in (a), the phase alignment point is the Xth partial discharge sampling point.

[0061] Furthermore, all partial discharge sampling points before the Xth partial discharge sampling point (i.e., the 1st to the (X-1th)th partial discharge sampling points) are extracted and moved to the end of the partial discharge monitoring signal sequence of this period to form a new sequence: Pk'=[pk,X, pk,X+1,…, pk,3600, pk,1, pk,2,…, pk,X-1]; Where Pk' represents the sequence of sampling points corresponding to the phase-aligned partial discharge monitoring signal after one acquisition cycle, pk represents a partial discharge sampling point, "pk,X" represents the Xth partial discharge sampling point, "pk,X+1" represents the (X+1)th partial discharge sampling point, and so on, which will not be elaborated here. k represents the kth acquisition cycle (k=1,2,…,N), and N is the total number of acquisition cycles.

[0062] Understandably, in this example, X+1 < 3600.

[0063] After the above transformation, the Xth partial discharge sampling point (i.e., the partial discharge sampling point corresponding to the voltage zero-crossing point) becomes the starting point of the new sequence. For example... Figure 3 As shown in (b), the aligned sequence starts at the phase (0°) corresponding to the zero crossing point.

[0064] Repeating the above steps to complete the time axis sliding alignment of the raw partial discharge monitoring signal data for 144 acquisition cycles can completely eliminate the deviation of "starting from different voltage phases at different acquisition times", ensuring that all partial discharge monitoring signal data are consistent in the voltage phase dimension, and providing a reliable foundation for subsequent multi-cycle signal superposition.

[0065] In some implementations, step S205 may be included before step S201.

[0066] S205: Filters the operating voltage signal within each acquisition cycle. The filtering process includes one or a combination of 50Hz notch filtering and low-pass filtering.

[0067] For the same partial discharge monitoring point, the embodiments of this application can filter the operating voltage signals collected in each acquisition cycle. Since the field environment contains various interference signals, including power frequency interference (50Hz and its harmonics) and high-frequency noise, these need to be filtered out.

[0068] Specifically, embodiments of this application can employ a 50Hz notch filter to remove power frequency interference. The notch filter has extremely high attenuation at 50Hz, effectively suppressing the influence of power frequency signals on the voltage waveform. Simultaneously, embodiments of this application can also employ a low-pass filter to remove high-frequency noise. The cutoff frequency of the low-pass filter can be set according to actual needs, for example, set to 500Hz, retaining the fundamental voltage signal while filtering out high-frequency interference. Through the above filtering processes, smooth voltage timing data can be obtained, providing a reliable basis for zero-crossing detection.

[0069] Further details can be found by referring to [link / reference]. Figure 1 Step S200 may be followed by step S300.

[0070] S300: For each partial discharge monitoring point, the partial discharge monitoring signals from multiple acquisition cycles after phase alignment are superimposed.

[0071] Specifically, for the phase-aligned partial discharge monitoring signals corresponding to each acquisition cycle of each partial discharge monitoring point, the embodiments of this application can superimpose the amplitudes of all partial discharge sampling points within the same phase interval. The phase interval includes multiple continuous phase angles, which are pre-divided based on the voltage phase range.

[0072] It should be noted that the voltage phase range is 0°-360°, corresponding to one complete power frequency cycle. In this embodiment, 0°-360° can be divided into several continuous phase intervals. For example, each 5° interval can be divided into a phase interval, resulting in [0°, 5°), [5°, 10°), ..., [355°, 360°]. Furthermore, partial discharge sampling points within the same phase interval refer to partial discharge sampling points falling within the same angular range in different sampling cycles. The voltage phases corresponding to these partial discharge sampling points are the same or similar, possessing a physical basis for superposition.

[0073] It is understood that when there are multiple partial discharge monitoring points, the embodiments of this application can superimpose the partial discharge monitoring signals corresponding to each partial discharge monitoring point.

[0074] It should also be noted that the granularity of the phase interval division can be flexibly adjusted according to the actual analysis accuracy requirements. For example, the phase interval can be divided at intervals of 1° or 2°, that is, the width of each phase interval is 1° or 2°, thereby obtaining higher phase resolution. The same phase interval may contain one or more partial discharge sampling points, or it may not contain any partial discharge sampling points.

[0075] For the same partial discharge monitoring point, in this embodiment, after phase alignment, the partial discharge monitoring signals corresponding to each acquisition cycle are reordered with the partial discharge sampling point corresponding to the zero-crossing point of the reference voltage as the starting point. In the aligned partial discharge monitoring signal, the sequence number of the partial discharge sampling point has changed, forming a new sequence number (hereinafter referred to as "aligned sequence number"). Since the preset sampling rate is fixed and the aligned sequence starts from the partial discharge sampling point corresponding to the zero-crossing point of the reference voltage, there is a one-to-one correspondence between the aligned sequence number and the phase interval.

[0076] For the same partial discharge monitoring point, if M partial discharge sampling points are collected in each power frequency cycle, then the phase interval corresponding to the i-th partial discharge sampling point is [(i-1)×360° / M, i×360° / M], where i=1,2,…,M. For example, when M=72, the phase interval corresponding to the first partial discharge sampling point of the phase-aligned partial discharge monitoring signal is [0°, 5°), the second partial discharge sampling point is [5°, 10°),…, and the 72nd sampling point is [355°, 360°].

[0077] Based on this, partial discharge sampling points with the same aligned sequence number in different acquisition cycles must be in the same phase interval and can be directly superimposed in terms of amplitude.

[0078] It should be further explained that the "superposition" operation involved in the embodiments of this application does not refer to the superposition of amplitudes between different partial discharge sampling points within the same acquisition cycle, but rather to the superposition of amplitudes of partial discharge sampling points within the same phase interval in multiple acquisition cycles, thereby obtaining the discharge characteristics of that phase interval under multiple measurement cycles.

[0079] To facilitate understanding, we can compare and explain from two perspectives: "single-signal plotting" and "multiple-signal superposition": Signal plotting for a single acquisition cycle: For a single acquisition cycle, the partial discharge monitoring signal can be plotted according to the phase angle [0°, 360°]. Each partial discharge sampling point falls within the corresponding phase interval based on its corresponding phase angle, reflecting the distribution of the partial discharge monitoring signal in the phase dimension within that acquisition cycle. At this time, the plotting result only reflects the raw information of a single cycle.

[0080] Superposition of signals from multiple acquisition cycles: In the embodiments of the present application, the phase-aligned partial discharge sampling points in each acquisition cycle can be classified according to the phase interval they belong to. For each phase interval, the amplitudes of the partial discharge sampling points falling within this phase interval in all acquisition cycles are superimposed to obtain the superposition result of this phase interval. This superposition process is equivalent to: aligning the partial discharge signals of multiple acquisition cycles on the phase dimension; within each phase interval, superimposing the amplitudes of the partial discharge monitoring signals in the "same phase interval" of different cycles.

[0081] Through the above manner, the embodiments of the present application can superimpose the partial discharge monitoring signals of each acquisition cycle corresponding to the same partial discharge monitoring point in the corresponding phase interval, so that the amplitude of the occasional partial discharge signal is significantly enhanced, while the background noise is relatively suppressed, the signal-to-noise ratio is improved, and the weak signal originally submerged in the noise can be revealed, thereby effectively improving the detection rate of the occasional partial discharge signal, and providing more stable and reliable map data for subsequent partial discharge type identification and insulation state assessment.

[0082] Further, the amplitude superposition includes any one of summation, maximum value superposition, or average value superposition.

[0083] ① Summation: In a specific embodiment, the superposition can be performed in a summation manner. For the same partial discharge monitoring point, the amplitudes of the partial discharge sampling points belonging to the same phase interval in N acquisition cycles are summed to obtain the superimposed amplitude corresponding to this phase interval. Suppose there are a total of M phase intervals (e.g., M=72), the amplitudes of the partial discharge sampling points (i.e., signal strength) of the i-th phase interval in N acquisition cycles are respectively , 0<i≤M, then the amplitude after summation is: ; wherein, is the amplitude of the partial discharge monitoring signal after summation, represents the amplitude of the partial discharge sampling point before summation, k represents the k-th acquisition cycle (k=1,2,…,N), and N is the total number of acquisition cycles.

[0084] It can be understood that the occasional partial discharge signal appears in a few cycles, and its amplitude is enhanced after summation (the amplitude increases linearly with the number of superimposed cycles). Moreover, the background noise is randomly distributed in each acquisition cycle, and the positive and negative parts cancel each other out or grow slowly after summation, which is relatively suppressed, so that the purpose of improving the signal-to-noise ratio can be achieved.

[0085] ② Maximum value superposition: In another specific embodiment, the superposition can be performed by taking the maximum value. For each phase interval, the maximum amplitude of this phase interval in N acquisition cycles of the same partial discharge monitoring point is taken as the superposition result: ; in, This represents the amplitude of the partial discharge monitoring signal after the maximum values ​​are superimposed. The value represents the amplitude of the partial discharge sampling point in the i-th phase interval within N acquisition cycles at the same partial discharge monitoring point, where k represents the k-th acquisition cycle (k=1,2,…,N) and N is the total number of acquisition cycles.

[0086] This method can effectively preserve strong, occasional signals and prevent them from being overwhelmed by average values.

[0087] ③ Take the average value and sum it up: In another specific embodiment, the summation can be performed by averaging. For each phase interval, the average amplitude of that phase interval is calculated over N acquisition cycles at the same partial discharge monitoring point: ; in, The amplitude of the partial discharge monitoring signal is the sum of the average values. This represents the sum of the amplitude values ​​of the partial discharge sampling points before accumulation, k represents the kth acquisition cycle (k=1,2,…,N), and N is the total number of acquisition cycles.

[0088] This method can suppress random noise and highlight the stable partial discharge characteristics, making it suitable for scenarios where the partial discharge signal is continuous but the amplitude is small.

[0089] For example, regarding early intermittent partial discharge (LPD) problems, because the acquisition window of a single acquisition cycle is short (e.g., 1 second), and the occurrence of intermittent LPD is random and intermittent, it is very likely that no LPD will occur within the acquisition window, meaning that no LPD pulse will be captured in a single acquisition cycle. If only the LPD monitoring signal acquired in a single cycle is used as the LPD judgment standard, when no LPD pulse is captured in that cycle, it is easy to miss the early intermittent LPD pulse, resulting in misjudgment and delaying the detection of the risk of equipment insulation degradation.

[0090] This application embodiment can superimpose data from multiple acquisition cycles. Taking 144 acquisition cycles as an example, even if only a portion (e.g., 50) of the acquisition cycles capture partial discharge pulses, and the remaining (e.g., 94) acquisition cycles do not detect any partial discharge pulses, after the superposition processing in step S300, the amplitudes of the partial discharge pulses acquired in some acquisition cycles will be superimposed within the corresponding phase intervals. Noise signals, due to their randomness, will be suppressed or canceled during the multi-cycle superposition process. Therefore, the partial discharge characteristics can still be significantly reflected in the final PRPD image, forming clear phase convergence features.

[0091] As can be seen, the embodiments of this application can effectively solve the problem of missed detection of intermittent partial discharge signals by superimposing multi-cycle partial discharge signals, and improve the sensitivity and reliability of partial discharge detection, which is especially suitable for the identification and early warning of early insulation defects.

[0092] Further details can be found by referring to [link / reference]. Figure 1 Step S300 may be followed by step S400.

[0093] S400: Based on the superimposed partial discharge monitoring signals, construct the partial discharge phase distribution map corresponding to each partial discharge monitoring point.

[0094] Specifically, for each partial discharge monitoring point, this embodiment of the application can construct a PRPD diagram with voltage phase as the abscissa and the amplitude of the superimposed partial discharge monitoring signal as the ordinate. The abscissa ranges from 0° to 360° and is divided into M equally spaced phase intervals (e.g., M=72). The ordinate is the amplitude of the superimposed partial discharge monitoring signal, in dBmV or mV.

[0095] It is understandable that multi-cycle signal superposition and PRPD image rendering essentially involves concatenating multiple one-dimensional arrays into a large array, and then rendering and outputting the entire array according to the rules for drawing PRPD images. In the output PRPD image, the amplitude corresponding to the intermittent partial discharge signal is significantly higher than the background noise, forming clear partial discharge feature points. Importantly, the embodiments of this application employ multi-cycle data superposition, making intermittent partial discharge signals that are originally weak or even undetectable in a single cycle clearly discernible in the superimposed PRPD image, increasing identification accuracy and avoiding early missed detection of intermittent partial discharge signals.

[0096] It is understood that when there are multiple partial discharge monitoring points, the embodiments of this application can construct a partial discharge phase distribution map for each partial discharge monitoring point separately.

[0097] Figure 4 A schematic diagram of the PRPD provided for an embodiment of this application.

[0098] like Figure 4As shown, this application embodiment provides a comprehensive PRPD spectrum after superimposing 144 cycles (24 hours). The horizontal axis represents the voltage phase, and the vertical axis represents the amplitude of the partial discharge monitoring signal superimposed using an average value superposition method. That is, the height of the bars represents the average amplitude. It can be understood that after the superposition step, the final PRPD spectrum does not show the original partial discharge sampling points, but rather the superimposed 72 data points (one value for each phase interval). Furthermore, the spectrum fully displays 72 phase intervals (each interval width 5°), with the discharge energy concentrated in phase intervals 18-23 (corresponding to 90°-115° phase) and phase intervals 54-59 (corresponding to 270°-295° phase), which conforms to the peak characteristics of surface discharge voltage.

[0099] Further details can be found by referring to [link / reference]. Figure 1 Step S400 may be followed by step S500.

[0100] S500: Input the partial discharge phase map into the pre-trained classification and recognition model to obtain the partial discharge recognition result.

[0101] The partial discharge identification results include whether the power equipment has partial discharge and the type of partial discharge. The partial discharge type includes one or more of the following: particulate discharge, surface discharge, suspension discharge, corona discharge, and air gap discharge.

[0102] Furthermore, the classification and recognition model in this embodiment can adopt a convolutional neural network (CNN) architecture, specifically an improved LeNet-5, ResNet, or VGG network structure. The classification and recognition model may include a feature extraction layer, a global average pooling layer, and a classification head.

[0103] Feature extraction layer: Composed of multiple convolutional and pooling layers, it is used to automatically extract partial discharge features from PRPD images, including the phase distribution pattern of discharge points, amplitude aggregation characteristics, and spectral morphology features.

[0104] Global average pooling layer: Converts feature maps into fixed-length feature vectors, reducing model parameters and preventing overfitting.

[0105] The classification head can include two parallel fully connected branches: an existence determination branch and a type identification branch. The existence determination branch outputs a binary classification result, characterizing whether partial discharge exists in the power equipment, in probability form, such as "probability of partial discharge: 87.3%". The type identification branch outputs a multi-class classification result, characterizing the specific type of partial discharge, with each type presented as a probability distribution, such as "surface discharge: 72.5%, particulate discharge: 15.2%, suspended discharge: 8.1%, corona discharge: 3.2%, air gap discharge: 1.0%".

[0106] In some implementations, the classification head may only include a type recognition branch. After the classification head outputs multi-classification results, this embodiment can determine the presence and type of localized discharge (LDD) based on probability magnitude. Specifically, when the highest probability value among the probabilities of each category output by the type recognition branch exceeds a preset threshold (e.g., 60%), the presence of LDD is determined, and the category corresponding to the highest probability is output as the LDD type; when the highest probability value is below the threshold, no LDD is determined. This implementation simplifies the network structure, reduces computation, and simultaneously addresses the dual needs of presence detection and type recognition through probability threshold judgment.

[0107] It is understood that when there are multiple partial discharge monitoring points, the embodiments of this application can input the PRPD maps corresponding to multiple partial discharge monitoring points into the classification and recognition model for joint diagnosis, or input the PRPD maps corresponding to each partial discharge monitoring point into the model independently for identification one by one; or input the PRPD maps of some partial discharge monitoring points into the model for preliminary identification, while using the PRPD maps of other partial discharge monitoring points as control samples or background noise references to cross-validate the consistency of the diagnostic results or help eliminate environmental interference, thereby improving the accuracy and reliability of partial discharge identification in multi-monitoring scenarios.

[0108] As can be seen, this application provides a partial discharge (PD) monitoring method based on multi-cycle data superposition. This method can perform multi-cycle data acquisition for at least one PD monitoring point to obtain the PD monitoring signal and operating voltage signal corresponding to each acquisition cycle. Then, this application can perform phase alignment on the PD monitoring signals of each acquisition cycle corresponding to the same PD monitoring point, and superimpose the phase-aligned PD monitoring signals of multiple acquisition cycles to obtain the "superimposed PD monitoring signal" corresponding to that PD monitoring point. Subsequently, this application can draw the PRPD spectrum corresponding to each PD monitoring point based on the "superimposed PD monitoring signal". In summary, the method provided by this application can take the same PD monitoring point as a unit, perform phase alignment and amplitude superposition on the PD monitoring signals acquired in multiple acquisition cycles, thereby forming the superimposed PD monitoring signal corresponding to that PD monitoring point. In other words, different acquisition cycles reflect the partial discharge state of the same monitoring location at different times. By superimposing the data, the weak or sporadic signals scattered in each cycle can be accumulated and enhanced in the phase dimension, and finally a comprehensive PRPD map that can truly characterize the partial discharge features of the monitoring point can be constructed.

[0109] Furthermore, the training steps for the classification and recognition model include the following steps S601-S603.

[0110] S601: Collect PRPD map samples of different types of partial discharges to construct a training dataset.

[0111] First, embodiments of this application can collect PRPD (partial discharge) spectrum samples of various types of partial discharges as basic data for model training. Partial discharge types include, but are not limited to, particulate discharge, surface discharge, suspended discharge, corona discharge, and air gap discharge. Sample sources can include: laboratory measured data, field measured data, and simulation data.

[0112] In some implementations, to ensure the diversity and generalization ability of the training dataset, a sufficient number of samples should be collected for each local release type (e.g., no less than 5,000 samples per class), and divided into training set, validation set and test set in a certain proportion, such as a 6:2:2 ratio.

[0113] S602: Perform data augmentation on the training dataset. Data augmentation includes at least one of rotation, scaling, and noise addition.

[0114] To expand the number of training samples and improve the robustness and generalization ability of the model, embodiments of this application can perform data augmentation processing on the training dataset constructed in S601. Data augmentation processing includes, but is not limited to, the following methods: Rotation enhancement: The PRPD map is rotated slightly (e.g., ±5°, ±10°) to simulate small phase shifts that may exist under different acquisition conditions, thereby enhancing the model's tolerance to phase alignment errors.

[0115] Scaling Enhancement: The amplitude dimension of the PRPD map is scaled (e.g., multiplied by a random coefficient between 0.8 and 1.2) to simulate amplitude variations caused by different acquisition gains and sensor sensitivity differences, thereby enhancing the model's adaptability to amplitude fluctuations.

[0116] Noise enhancement: Gaussian noise with different signal-to-noise ratios is added to the PRPD map to simulate electromagnetic interference, sensor noise, etc. in the field environment, thereby enhancing the model's recognition stability in noisy environments.

[0117] In some implementation methods, translation enhancement and elastic deformation processing can also be used.

[0118] Translation enhancement: The PRPD spectrum is slightly translated along the phase axis (e.g., ±1-3 phase intervals) to simulate the slight deviations that may exist in phase alignment.

[0119] Elastic deformation: Slight nonlinear deformation is applied to the PRPD spectrum to simulate subtle changes in partial discharge characteristics under different equipment and operating conditions.

[0120] The data augmentation operations described above can be performed individually or in combination (e.g., rotating, scaling, and adding noise) to generate diverse training samples. Data augmentation can increase the number of original training samples by several to tens of times, effectively preventing model overfitting and improving the model's generalization performance.

[0121] S603: Input the augmented training dataset into the convolutional neural network for training to obtain a classification and recognition model.

[0122] In this way, the trained classification and recognition model has the ability to automatically extract PRPD map features, accurately identify the type of partial discharge, and determine whether partial discharge exists, providing a reliable basis for the insulation status assessment of power equipment.

[0123] As can be seen from the above, this application provides a partial discharge monitoring method based on multi-cycle data superposition. The method includes: periodically acquiring partial discharge monitoring signals and operating voltage signals from power equipment; wherein the partial discharge sampling points of the partial discharge monitoring signals and the voltage sampling points of the operating voltage signals are synchronously acquired based on the same preset sampling rate; the power equipment includes at least one partial discharge monitoring point, and the partial discharge monitoring signal is obtained by monitoring the partial discharge monitoring point; for each partial discharge monitoring point, using the voltage zero-crossing point of the operating voltage signal as a reference, the phase of the partial discharge monitoring signals corresponding to each acquisition cycle is aligned; for each partial discharge monitoring point, the phase-aligned partial discharge monitoring signals from multiple acquisition cycles are superimposed; based on the superimposed partial discharge monitoring signals, a partial discharge phase distribution map corresponding to each partial discharge monitoring point is constructed; the partial discharge phase map is input into a pre-trained classification and recognition model to obtain the partial discharge recognition result. Through the above steps, this application can continuously acquire new data over time and repeatedly plot the relationship between the partial discharge signal and the voltage phase, thereby capturing discharge characteristics (such as early intermittent partial discharge characteristics). In other words, the embodiments of this application can extend the effective signal capture time from the second level of a single cycle to the hour level of multiple cycles by constructing a "virtual long window" for the same partial discharge monitoring point without changing the sensor acquisition mode (without shortening the acquisition interval or extending the acquisition window duration). Even if the early partial discharge signal only appears occasionally in a few acquisition cycles, the abnormal value can be clearly presented by superimposing it on the PRPD diagram, effectively avoiding missed detection and significantly improving the detection rate of early occasional partial discharge signals. At the same time, it avoids the problem of a surge in equipment power consumption caused by shortening the acquisition interval or extending the acquisition window, providing reliable technical support for the accurate assessment of the early insulation degradation state of power equipment.

[0124] Figure 5 This is a schematic diagram of the partial discharge monitoring device based on multi-period data superposition provided in an embodiment of this application.

[0125] like Figure 5 As shown in the figure, this application provides a partial discharge monitoring device based on multi-period data superposition. The device includes: The acquisition module 1001 is configured to periodically acquire partial discharge monitoring signals and operating voltage signals of power equipment; wherein, the partial discharge sampling points of the partial discharge monitoring signals and the voltage sampling points of the operating voltage signals are synchronously acquired based on the same preset sampling rate; the power equipment includes at least one partial discharge monitoring point, and the partial discharge monitoring signal is obtained by monitoring the partial discharge monitoring point; The phase alignment module 1002 is configured to: for each partial discharge monitoring point, use the voltage zero-crossing point of the operating voltage signal as a reference to perform phase alignment on the partial discharge monitoring signal corresponding to each acquisition cycle; The superposition module 1003 is configured to superimpose the partial discharge monitoring signals from multiple acquisition cycles after phase alignment for each partial discharge monitoring point. The drawing module 1004 is configured to: construct a partial discharge phase distribution map corresponding to each partial discharge monitoring point based on the superimposed partial discharge monitoring signal; The identification module 1005 is configured to input the partial discharge phase map into a pre-trained classification and identification model to obtain the partial discharge identification result.

[0126] In one possible implementation, the phase alignment module 1002 is specifically configured to: extract the reference voltage zero-crossing point of the operating voltage signal within each acquisition cycle, wherein when the voltage amplitude of one voltage sampling point is negative and the voltage amplitude of its adjacent next voltage sampling point is positive, the first voltage sampling point with a positive voltage amplitude is the reference voltage zero-crossing point; determine the sampling sequence number of each reference voltage zero-crossing point as the zero-crossing point sequence number corresponding to each acquisition cycle; determine the phase alignment point of the partial discharge monitoring signal corresponding to each acquisition cycle based on the zero-crossing point sequence number corresponding to each acquisition cycle; wherein the phase alignment point is the partial discharge sampling point in the partial discharge monitoring signal whose sampling point sequence number is the same as the zero-crossing point sequence number; for the partial discharge monitoring signal corresponding to each acquisition cycle, extract the partial discharge sampling points before the phase alignment point and concatenate them to the end to obtain the phase-aligned partial discharge monitoring signal corresponding to each acquisition cycle.

[0127] In one possible implementation, the phase alignment module 1002 is further configured to perform filtering on the operating voltage signal in each acquisition cycle, including one or a combination of 50Hz notch filtering and low-pass filtering.

[0128] In one possible implementation, the superposition module 1003 is specifically configured to: for the phase-aligned partial discharge monitoring signal corresponding to each acquisition cycle of each partial discharge monitoring point, superimpose the amplitudes of all partial discharge sampling points in the same phase interval; wherein, the phase interval is pre-divided based on the voltage phase range; the amplitude superposition includes any one of accumulation, maximum value superposition, or average value superposition.

[0129] In one possible implementation, the plotting module 1004 is specifically configured to: for each partial discharge monitoring point, construct a phase-resolved partial discharge (PRPD) map with voltage phase as the horizontal axis and the amplitude of the superimposed partial discharge monitoring signal as the vertical axis.

[0130] In one possible implementation, the partial discharge identification result includes whether the power equipment has partial discharge and the type of partial discharge. The partial discharge type includes one or more of the following: particulate discharge, surface discharge, suspension discharge, corona discharge, and air gap discharge.

[0131] In one possible implementation, the partial discharge monitoring signal and the operating voltage signal correspond to the same acquisition period and acquisition window; the length of the acquisition window is shorter than the length of the acquisition period.

[0132] In one possible implementation, the partial discharge monitoring device based on multi-cycle data superposition provided in this application embodiment further includes a training module, which is configured to: collect PRPD map samples of different types of partial discharges to construct a training dataset; perform data augmentation processing on the training dataset, including at least one of rotation, scaling, and noise addition; and input the data-augmented training dataset into a convolutional neural network for training to obtain a classification and recognition model.

[0133] In one possible implementation, the partial discharge monitoring signal is acquired by an ultrasonic sensor or an ultra-high frequency sensor, and the operating voltage signal is acquired by a voltage transformer; the power equipment is at least one of a transformer, a gas-insulated switchgear, a switch cabinet, and a cable.

[0134] In a specific implementation, the present invention also provides a computer storage medium, wherein the computer storage medium may store a program, which, when executed, may include some or all of the steps in various embodiments of the partial discharge monitoring method based on multi-cycle data superposition provided by the present invention. The storage medium may be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0135] It is readily understood that, based on the several embodiments provided in this application, those skilled in the art can combine, split, or reorganize the embodiments of this application to obtain other embodiments, none of which exceed the protection scope of this application.

[0136] The above detailed embodiments further illustrate the purpose, technical solution, and beneficial effects of the embodiments of this application. It should be understood that the above are merely specific embodiments of the embodiments of this application and are not intended to limit the protection scope of the embodiments of this application. Any modifications, equivalent substitutions, improvements, etc., made on the basis of the technical solutions of the embodiments of this application should be included within the protection scope of the embodiments of this application.

Claims

1. A partial discharge monitoring method based on multi-period data overlay, characterized in that, The method includes: The partial discharge monitoring signal and the operating voltage signal of the power equipment are periodically collected; wherein the partial discharge sampling point of the partial discharge monitoring signal and the voltage sampling point of the operating voltage signal are synchronously collected based on the same preset sampling rate; the power equipment includes at least one partial discharge monitoring point, and the partial discharge monitoring signal is obtained by monitoring the partial discharge monitoring point; For each partial discharge monitoring point, the phase alignment of the partial discharge monitoring signal corresponding to each acquisition cycle is performed based on the zero-crossing point of the operating voltage signal. For each partial discharge monitoring point, the partial discharge monitoring signals from multiple acquisition cycles after phase alignment are superimposed. Based on the superimposed partial discharge monitoring signals, a partial discharge phase distribution map corresponding to each partial discharge monitoring point is constructed. The partial discharge phase map is input into a pre-trained classification and recognition model to obtain the partial discharge recognition result.

2. The partial discharge monitoring method based on multi-period data superposition according to claim 1, characterized in that, The step of aligning the phase of the partial discharge monitoring signal corresponding to each acquisition cycle based on the zero-crossing point of the operating voltage signal includes: Extract the reference voltage zero-crossing point of the operating voltage signal in each acquisition cycle. In each acquisition cycle, when the voltage amplitude of one voltage sampling point is negative and the voltage amplitude of its next adjacent voltage sampling point is positive, the first voltage sampling point with a positive voltage amplitude is the reference voltage zero-crossing point. The sampling sequence number of each zero-crossing point of the reference voltage is determined as the zero-crossing point sequence number corresponding to each acquisition cycle; Based on the zero-crossing index corresponding to each acquisition cycle, the phase alignment point of the partial discharge monitoring signal corresponding to each acquisition cycle is determined; wherein, the phase alignment point is the partial discharge sampling point in the partial discharge monitoring signal whose sampling point index is the same as the zero-crossing index; For the partial discharge monitoring signal corresponding to each acquisition cycle, the partial discharge sampling points before the phase alignment point are extracted and spliced ​​to the end to obtain the phase-aligned partial discharge monitoring signal corresponding to each acquisition cycle.

3. The partial discharge monitoring method based on multi-period data superposition according to claim 2, characterized in that, Before the reference voltage of the operating voltage signal in each acquisition cycle is zeroed out, the process further includes: The operating voltage signal in each acquisition cycle is filtered, and the filtering process includes one or a combination of 50Hz notch filtering and low-pass filtering.

4. The partial discharge monitoring method based on multi-period data superposition according to claim 1, characterized in that, The step of superimposing the phase-aligned partial discharge monitoring signals from multiple acquisition cycles for each partial discharge monitoring point includes: For each acquisition cycle of the partial discharge monitoring point, the phase-aligned partial discharge monitoring signal is superimposed on the amplitude of all partial discharge sampling points in the same phase interval; wherein, the phase interval is pre-divided based on the voltage phase range; the amplitude superposition includes any one of accumulation, maximum value superposition or average value superposition.

5. The partial discharge monitoring method based on multi-period data superposition according to claim 1, characterized in that, The construction of a partial discharge phase distribution map corresponding to each partial discharge monitoring point based on the superimposed partial discharge monitoring signal includes: For each partial discharge monitoring point, a phase-resolved partial discharge (PRPD) diagram is constructed with voltage phase as the horizontal axis and the amplitude of the superimposed partial discharge monitoring signal as the vertical axis.

6. The partial discharge monitoring method based on multi-period data superposition according to claim 1, characterized in that, The partial discharge identification result includes whether the power equipment has partial discharge and the type of partial discharge. The partial discharge type includes one or more of the following: particulate discharge, surface discharge, suspension discharge, corona discharge, and air gap discharge.

7. The partial discharge monitoring method based on multi-period data superposition according to claim 1, characterized in that, The partial discharge monitoring signal and the operating voltage signal correspond to the same acquisition period and acquisition window; the length of the acquisition window is less than the length of the acquisition period.

8. The partial discharge monitoring method based on multi-period data superposition according to claim 1, characterized in that, The training steps of the classification and recognition model include: Collect PRPD map samples of different types of partial discharges to construct a training dataset; The training dataset is subjected to data augmentation processing, which includes at least one of rotation, scaling, and noise addition. The augmented training dataset is input into a convolutional neural network for training to obtain the classification and recognition model.

9. The partial discharge monitoring method based on multi-period data superposition according to claim 1, characterized in that, The partial discharge monitoring signal is acquired by an ultrasonic sensor or an ultra-high frequency sensor, and the operating voltage signal is acquired by a voltage transformer; the power equipment is at least one of a transformer, a gas-insulated switchgear, a switch cabinet, and a cable.

10. A partial discharge monitoring device based on multi-period data superposition, characterized in that, The device includes: The acquisition module is configured to periodically acquire partial discharge monitoring signals and operating voltage signals of power equipment; wherein the partial discharge sampling points of the partial discharge monitoring signals and the voltage sampling points of the operating voltage signals are acquired synchronously based on the same preset sampling rate; the power equipment includes at least one partial discharge monitoring point, and the partial discharge monitoring signal is obtained by monitoring the partial discharge monitoring point; The phase alignment module is configured to: for each partial discharge monitoring point, use the voltage zero-crossing point of the operating voltage signal as a reference to perform phase alignment on the partial discharge monitoring signal corresponding to each acquisition cycle; The superposition module is configured to superimpose the partial discharge monitoring signals from multiple acquisition cycles after phase alignment for each partial discharge monitoring point. The plotting module is configured to: construct a partial discharge phase distribution map corresponding to each partial discharge monitoring point based on the superimposed partial discharge monitoring signal; The identification module is configured to input the partial discharge phase map into a pre-trained classification and identification model to obtain the partial discharge identification result.

Images