A power distribution network multi-path parallel fault prediction and accurate judgment system based on big data analysis

Abstract

This invention provides a multi-parallel fault prediction and accurate assessment system for power distribution networks based on big data analysis, including a multi-parallel acquisition module, a big data-based basic prediction module, a meteorological-linked dynamic bypass module, and a topology-based source tracing and congestion prevention scheduling module. The system acquires real-time electrical and environmental micro-meteorological parameters of nodes, performs standard denoising filtering on the electrical parameters, and outputs basic early warnings. When a surge in windblown sand concentration accompanied by a sudden drop in temperature is detected, standard denoising filtering is paused, and high-frequency micro-distortion segments without denoising are extracted to output a high-risk warning for insulation flashover. To address the situation where multiple nodes concurrently upload large amounts of fault data due to the extraction of these high-frequency micro-distortion segments, leading to near-saturation of communication computing power, a full-feature upload certificate is issued to the core hub node based on the topology depth and the slope of the zero-sequence current mutation. This invention balances the objective capture of early flashover characteristics under specific micro-meteorological conditions with the rational scheduling of network resources under high-concurrency data storms.

Description

Technical Field

[0001] This invention relates to the field of power system distribution network operation monitoring and data analysis technology, and in particular to a distribution network multi-parallel fault prediction and accurate judgment system based on big data analysis. Background Technology

[0002] As a crucial component of the power system, the operating status of the distribution network directly affects the reliability and stability of power supply. In the daily operation and maintenance of the distribution network, data acquisition terminals are typically deployed at various branch lines and key nodes to obtain basic electrical parameters such as voltage and current in real time. To ensure the accuracy of subsequent status analysis, the conventional data processing procedure generally performs standard noise reduction filtering on the acquired electrical signals. This filtering is mainly used to remove high-frequency electromagnetic noise and transient interference generated by switching operations, thereby extracting a smooth and stable power frequency fundamental signal to meet the status over-limit early warning requirements under most normal operating conditions.

[0003] Distribution network lines are often located in areas with variable terrain and climate. When lines are situated in windy or dusty environments, sudden changes in local microclimates can affect equipment operation. Under conditions of rapid increases in air dust concentration accompanied by a sharp drop in ambient temperature, contaminants adhering to the surface of line insulators are prone to physical changes and deliquescence, leading to early insulation degradation. Insulators in the early stages of degradation typically exhibit weak partial discharges, which appear as high-frequency, small, distorted pulses interspersed within broadband signals. Conventional fixed noise reduction filtering mechanisms, when performing standard filtering operations, tend to attenuate these high-frequency discharge characteristics, which contain precursors to insulation flashover, as ordinary background white noise. This objectively increases the difficulty for the system to capture effective early warning signals in the early stages of fault development.

[0004] When a distribution network is subjected to widespread severe weather or regional transient disturbances, multiple upstream and downstream nodes within the affected area often detect abnormal fluctuations within similar time slices and simultaneously trigger local alarm upload mechanisms. Because each monitoring terminal concurrently sends transient waveform files containing massive amounts of high-frequency sampling point characteristics to the superior master station system within a short period, this sudden large-scale data flow can quickly exhaust the physical bandwidth of the distribution network's underlying communication. The queuing of numerous data packets in the transmission network can cause the communication link's computing power to saturate or even become congested. Without targeted and coordinated scheduling, the indiscriminate competition for upload resources among nodes makes it easy for the full waveform files of nodes at the core fault location to experience transmission delays, affecting the time for the dispatch center to obtain effective judgment criteria and hindering the centralized analysis of sudden faults by the backend. Summary of the Invention

[0005] The purpose of this invention is to provide a multi-path parallel fault prediction and accurate judgment system for distribution networks based on big data analysis, in order to solve the technical problems that existing conventional noise reduction and filtering mechanisms are prone to filtering out the early high-frequency discharge characteristics of insulation flashover under specific meteorological conditions, and that the indiscriminate uploading of massive waveform data when multiple nodes are simultaneously alarmed can easily cause congestion in the underlying communication network and delay the transmission of critical data in core nodes.

[0006] This invention provides a system for multi-path parallel fault prediction and accurate judgment in power distribution networks based on big data analysis, comprising:

[0007] The multi-channel parallel acquisition module is used to simultaneously acquire real-time electrical parameters and real-time environmental micro-meteorological parameters of multiple related nodes in the target area's power distribution network;

[0008] The big data basic prediction module is used to perform standard denoising filtering on the real-time electrical parameters and input the denoised electrical parameters into a preset analysis architecture to output basic early warning results.

[0009] The distribution network multi-parallel fault prediction and accurate judgment system based on big data analysis also includes:

[0010] The meteorological linkage dynamic bypass module is used to generate a filtering blocking command and send it to the associated node corresponding to the target area when the real-time environmental micro-meteorological parameters meet the conditions of a sudden increase in wind and sand concentration accompanied by a sudden drop in temperature. The standard denoising filtering operation is paused, and the high-frequency micro-distortion fragments that have not been denoised are extracted. Based on the high-frequency micro-distortion fragments, a high-risk warning for insulation flashover is output.

[0011] The topology tracing and congestion prevention scheduling module is used to comprehensively evaluate the topology depth of each node within the distribution network structure and the slope of zero-sequence current mutation when multiple associated nodes concurrently upload large amounts of fault data containing high-frequency micro-distortion segments and transient waveform files due to the execution of the filtering and blocking instructions, which leads to the saturation of communication computing power. The module marks the node with the highest evaluation score as the core hub node and issues it a full feature upload certificate. Other nodes that do not hold the full feature upload certificate are forced to enter a degraded transmission state as edge branch nodes and only upload lightweight feature summaries. This ensures that subsequent processing computing power is focused on the high-risk early warning and judgment of insulation pollution flashover of the core hub node.

[0012] Optionally, the multi-channel parallel acquisition module includes:

[0013] The high-frequency synchronous sampling unit is used to perform synchronous data acquisition on multiple parallel branch lines in the target area distribution network using the positioning satellite clock signal, and to obtain three-phase voltage waveform data, three-phase current waveform data and zero-sequence current change as the real-time electrical parameters.

[0014] The micro-meteorological sensing and collection unit is used to acquire the real-time wind speed index, air sand concentration index, and real-time temperature change rate of the multiple associated nodes as the real-time environmental micro-meteorological parameters.

[0015] The electricity information collection unit is used to acquire user payment status, electricity load, meter operating status, meter box voltage and current, and transformer area operating parameters within the target area distribution network, and to form a unified time-stamped data source.

[0016] Optionally, the big data basic prediction module includes:

[0017] The feature baseline construction unit is used to extract electrical waveform data within historical normal operating cycles to construct a multidimensional reference envelope.

[0018] The parallel differential comparison unit is used to perform a parallel comparison between the real-time electrical parameters and the multidimensional reference envelope. When it is determined that the waveform corresponding to the real-time electrical parameters deviates from the multidimensional reference envelope and the deviation exceeds a preset safety threshold, the basic warning result is generated.

[0019] Optionally, the big data basic prediction module further includes:

[0020] The status cache traceability unit is used to store the real-time electrical parameters and real-time environmental micro-meteorological parameters of the corresponding node into a distributed cache array after generating the basic early warning result, and to add a timestamp tag for subsequent traceability and retrieval.

[0021] Optionally, the system further includes:

[0022] The panoramic topology mapping module is used to construct a three-dimensional distribution network topology map based on the actual geographical orientation of the target distribution network, mark vulnerable nodes located at the wind gap of the canyon in the three-dimensional distribution network topology map, and display the basic early warning results in the three-dimensional distribution network topology map by coloring and highlighting.

[0023] Optionally, the meteorological linkage dynamic bypass module includes:

[0024] The micro-meteorological change monitoring unit is used to monitor in real time the slope of the increase in sand content and the slope of the decrease in temperature within the real-time environmental micro-meteorological parameters.

[0025] The bypass instruction execution unit is used to determine that the conditions of a sudden increase in sand concentration accompanied by a sudden drop in temperature are met when the slope of the increase in sand concentration exceeds the preset sandstorm warning line and the slope of the decrease in temperature exceeds the preset temperature drop threshold line. At this time, the unit sends the filtering blocking instruction to the associated node corresponding to the target area to suspend the execution of the standard noise reduction filtering operation.

[0026] Optionally, the meteorological linkage dynamic bypass module further includes:

[0027] A high-frequency micro-discharge interception unit is used to acquire the undenoised original signal as the remaining unfiltered signal after the bypass instruction execution unit suspends the standard denoising and filtering operation, and to intercept the undenoised high-frequency micro-distortion segment including high-frequency spike pulses from the remaining unfiltered signal.

[0028] The pollution flashover trend analysis unit is used to take the associated node that generates the high-frequency micro-distortion segment as the target node, and obtain the normal micro-frequency signal of the adjacent associated node that meets the preset normal operation conditions; perform a lateral differential comparison between the undenoised high-frequency micro-distortion segment of the target node and the normal micro-frequency signal; if the differential amplitude exceeds the preset discharge threshold, the high-risk warning of insulation pollution flashover is sent to the dispatch control center.

[0029] Optionally, after receiving the high-risk warning of insulation flashover, the dispatch control center initiates multi-path parallel analysis logic in conjunction with the unified time-stamped data source; the multi-path parallel analysis logic includes simultaneously executing line outage information analysis, work order analysis, payment status analysis, and recall and curve analysis for users, meter boxes, and transformer areas;

[0030] The dispatch control center is equipped with two modes: accurate judgment and precise judgment. The accurate judgment prioritizes response speed and is configured to output the first round of emergency repair work orders and power outage information judgment results within 5 seconds, and output voltage and current measurement results within 30 seconds. The precise judgment is configured to accurately determine the cause and location of the fault based on multi-dimensional data cross-validation, and finally output a comprehensive judgment result including the power outage type, fault location and fault cause.

[0031] Optionally, the topology tracing and congestion prevention scheduling module includes:

[0032] The congestion situation awareness unit is used to continuously count the concurrent occupancy rate of each of the associated nodes uploading the large-capacity fault data. When the concurrent occupancy rate exceeds the communication bandwidth load limit, the congestion prevention scheduling logic is activated.

[0033] The topology and electrical evaluation unit is used to calculate the topology level depth and the zero-sequence current mutation slope of the associated nodes that concurrently upload the large-capacity fault data after the congestion prevention scheduling logic is activated.

[0034] Optionally, the topology tracing and congestion prevention scheduling module further includes:

[0035] The credential dynamic issuance unit is used to calculate a comprehensive evaluation score based on the topology level depth and the zero-sequence current mutation slope, mark the associated node with the highest comprehensive evaluation score as the core hub node, and issue the full feature upload credential to the core hub node, so that the core hub node can obtain the highest level network priority to transmit the complete transient waveform file.

[0036] Optionally, the topology tracing and congestion prevention scheduling module further includes:

[0037] An edge node degradation unit is used to issue a forced degradation instruction to the edge branch node that does not hold the full feature upload certificate; the edge branch node enters the degradation transmission state after receiving the forced degradation instruction;

[0038] The feature summary extraction unit is used to issue feature extraction instructions to the edge branch nodes in the degraded transmission state, instructing the node to extract only the effective voltage and current values ​​within its own transient waveform file, compare the effective current value with a set limit value to generate an over-limit flag, and combine the effective voltage value, the effective current value, and the over-limit flag to form the lightweight feature summary for uploading, thereby reducing network bandwidth usage and preventing massive concurrent data from blocking the network and affecting the overall judgment process.

[0039] The present invention has achieved the following beneficial effects:

[0040] The meteorological linkage dynamic bypass module configured in this invention can monitor the micro-meteorological evolution of the area where the node is located in real time. When faced with a sudden change in the environment, such as a surge in sand concentration accompanied by a sudden drop in temperature, it actively suspends the standard denoising and filtering operation, thereby objectively retaining and extracting the high-frequency small distortion segments that have not been denoised for outputting insulation flashover warning. This mechanism prevents the risk of conventional solidified filtering misfiltering the high-frequency characteristics of early insulation discharge and ensures the effective acquisition of flashover precursor signals under special meteorological conditions. However, extracting and retaining the original high-frequency data that has not been denoised will generate a huge amount of data. When a sudden change in the micro-meteorology causes multiple nodes in the area to trigger the bypass extraction simultaneously, it is very easy to cause a data congestion storm in the underlying communication network. To address this, the present invention deeply integrates a topology tracing and congestion prevention scheduling module to comprehensively manage concurrent upload tasks. Specifically targeting such high-concurrency situations, it comprehensively evaluates each concurrent node based on the topology depth within the distribution network and the slope of zero-sequence current mutation. It issues full feature upload credentials to the core hub node and forces edge branch nodes into a degraded transmission state that only sends lightweight features. This cause-and-effect linkage communication scheduling strategy takes into account both the complete return of critical waveform data and the reasonable reduction of non-critical data, alleviating the network channel load pressure caused by the synchronous alarm of massive data induced by micro-meteorological mutations. It ensures the orderly transmission of core fault feature data and the stable operation of the overall analysis process under high-concurrency conditions.

[0041] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0042] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0043] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0044] Figure 1 A block diagram of the overall structure of a multi-path parallel fault prediction and accurate judgment system for power distribution networks based on big data analysis, provided in an embodiment of the present invention;

[0045] Figure 2 This is a block diagram of the internal structure of the multi-channel parallel acquisition module and the big data basic prediction module provided in the embodiments of the present invention;

[0046] Figure 3 This is a block diagram of the internal structure of the meteorological linkage dynamic bypass module provided in an embodiment of the present invention;

[0047] Figure 4 This is a flowchart illustrating the data capture and early warning workflow of the meteorological linkage dynamic bypass module provided in this embodiment of the invention.

[0048] Figure 5 This is a block diagram of the internal structure of the topology tracing and congestion prevention scheduling module provided in an embodiment of the present invention;

[0049] Figure 6 The flowchart illustrates the congestion control process of the topology tracing and congestion prevention scheduling module provided in this embodiment of the invention. Detailed Implementation

[0050] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0051] This application provides a system for predicting and accurately assessing multi-path parallel faults in a power distribution network based on big data analysis. For example... Figure 1 As shown, the system includes:

[0052] The multi-channel parallel acquisition module is used to simultaneously acquire real-time electrical parameters and real-time environmental micro-meteorological parameters of multiple related nodes in the target area's power distribution network;

[0053] The big data basic prediction module is used to perform standard denoising filtering on the real-time electrical parameters and input the denoised electrical parameters into a preset analysis architecture to output basic early warning results.

[0054] The meteorological linkage dynamic bypass module is used to generate a filtering blocking command and send it to the associated node corresponding to the target area when the real-time environmental micro-meteorological parameters meet the conditions of a sudden increase in wind and sand concentration accompanied by a sudden drop in temperature. The standard denoising filtering operation is paused, and the high-frequency micro-distortion fragments that have not been denoised are extracted. Based on the high-frequency micro-distortion fragments, a high-risk warning for insulation flashover is output.

[0055] The topology tracing and congestion prevention scheduling module is used to comprehensively evaluate the topology depth of each node within the distribution network structure and the slope of zero-sequence current mutation when multiple associated nodes concurrently upload large amounts of fault data containing high-frequency micro-distortion segments and transient waveform files due to the execution of the filtering and blocking instructions, which leads to the saturation of communication computing power. The module marks the node with the highest evaluation score as the core hub node and issues it a full feature upload certificate. Other nodes that do not hold the full feature upload certificate are forced to enter a degraded transmission state as edge branch nodes and only upload lightweight feature summaries. This ensures that subsequent processing computing power is focused on the high-risk early warning and judgment of insulation pollution flashover of the core hub node.

[0056] Furthermore, such as Figure 2As shown, the multi-channel parallel acquisition module includes a high-frequency synchronous sampling unit and a micro-meteorological sensing and collection unit. The multi-channel parallel acquisition module is configured with a data bus to collect the physical parameters of the associated nodes and convert these physical parameters into digital messages. In the underlying circuit board routing layout, physical isolation trenches are set between the high-voltage signal traces and the meteorological signal traces, and an insulating layer is set between the copper layers.

[0057] Specifically, the high-frequency synchronous sampling unit is used to perform synchronous data acquisition on multiple parallel branch lines within the target area's power distribution network using a positioning satellite clock signal, acquiring three-phase voltage waveform data, three-phase current waveform data, and zero-sequence current abrupt change as the real-time electrical parameters. The high-frequency synchronous sampling unit is internally equipped with a satellite time synchronization receiver, used to receive time synchronization commands from the positioning satellite clock signal and generate a trigger pulse sequence. The trigger pulse sequence is transmitted to each analog-to-digital converter (ADC) channel via a communication bus, driving the associated ADCs to perform sampling operations within the same time slice. The high-frequency synchronous sampling unit receives analog level signals, filters and amplifies them, and then inputs them to the ADCs, outputting the three-phase voltage waveform data and three-phase current waveform data with the same clock reference.

[0058] After acquiring the waveform data, the high-frequency synchronous sampling unit performs vector superposition and summation of the instantaneous current values ​​of phase A, phase B, and phase C to generate a transient zero-sequence current sequence. A dual-buffered queue is established in the storage area to store the transient zero-sequence current sequence of the current power frequency cycle and the transient zero-sequence current sequence of the previous power frequency cycle, respectively. To eliminate cycle duration drift errors caused by minute fluctuations (frequency offset) in the actual operating frequency of the distribution network, before performing the point-by-point subtraction operation, the high-frequency synchronous sampling unit first extracts the positive zero-crossing time coordinates of the fundamental voltage corresponding to the previous and current power frequency cycles. Using this positive zero-crossing point as the phase alignment anchor point, the transient zero-sequence current sequence of the current power frequency cycle is shifted and overlapped on the time axis at the microsecond level. Then, the point-by-point subtraction operation is performed strictly according to the relative time offset, thereby effectively avoiding false differential current caused by frequency fluctuations. Data points with the same phase coordinates within the double-buffered queue are extracted and subtracted point by point to output a difference matrix. The unsigned peak values ​​within the difference matrix are further extracted and defined as the zero-sequence current abrupt change. The three-phase voltage waveform data, three-phase current waveform data, and zero-sequence current abrupt change are combined and encapsulated into the real-time electrical parameters.

[0059] Furthermore, the micro-meteorological sensing and collection unit is used to acquire real-time wind speed indicators, air dust concentration indicators, and real-time temperature change rates at the locations of the multiple associated nodes as real-time environmental micro-meteorological parameters. Specifically, the micro-meteorological sensing and collection unit acquires the real-time wind speed indicators based on a wind speed sensor; acquires real-time wind direction vector data based on a wind direction sensor to establish the dominant wind direction in the region; acquires the air dust concentration indicators based on a dust sensor; and acquires temperature values ​​using a temperature probe.

[0060] For calculating the real-time temperature change rate, the micro-meteorological sensing and collection unit records the time span between two consecutive temperature data acquisitions; calculates the difference between the currently measured temperature value and the previously measured temperature value; divides this difference by the time span to obtain the real-time temperature change rate. The real-time wind speed index, air dust concentration index, and real-time temperature change rate are combined to form the real-time environmental micro-meteorological parameters. After generating the real-time environmental micro-meteorological parameters, the micro-meteorological sensing and collection unit extracts the timestamp corresponding to the current data and appends the real-time environmental micro-meteorological parameters to the real-time electrical parameter message with the same timestamp.

[0061] Furthermore, to enrich the data dimensions for subsequent comprehensive analysis, the multi-channel parallel acquisition module also includes an electricity consumption information acquisition unit. This unit synchronously acquires user payment status, electricity load, meter operating status, meter box voltage and current, and transformer operating parameters within the target area's distribution network. It then integrates these data with real-time electrical parameters and real-time environmental micro-meteorological parameters to form a unified time-stamped data source, which is submitted to the backend system. Addressing the sampling rate differences between high and low frequency data, the fusion specifically employs a timestamp mapping and zero-order hold strategy: using the high-frequency time axis of the real-time electrical parameters as a reference, minute-level or static electricity consumption information parameters are mapped to this high-frequency reference axis according to the nearest time principle. Before the next status update, zero-order hold logic is used to broaden and pad the data values, thus encapsulating them into a multi-dimensional feature matrix with strictly aligned long and short periods.

[0062] Furthermore, the big data basic prediction module performs standard denoising filtering on the real-time electrical parameters and inputs the denoised electrical parameters into a preset analysis architecture to output basic early warning results. Specifically, the big data basic prediction module uses a low-pass digital filter to filter the real-time electrical parameters, filtering out signals outside the preset frequency band. Combining the electrical engineering laws of steady-state operation and harmonic distribution in the distribution network, the preset frequency band is specifically set to 0Hz to 2.5kHz, and the corresponding low-pass digital filter cutoff frequency is 2.5kHz. Specifically, a 4th-order Butterworth IIR digital filter is used to ensure smooth amplitude-frequency characteristics within the passband. This accurately filters out high-frequency noise caused by high-frequency electromagnetic interference in space or transient arcs at switch breaks, retaining only the fundamental frequency and effective harmonic characteristics within the 50th order for subsequent construction of the steady-state envelope. At the same time, power frequency harmonics are eliminated through morphological processing. The morphological processing specifically uses a one-dimensional flat structuring element, whose length span is set to include 3 to 5 consecutive discrete sampling points. The waveform data retained after the above processing forms the denoised electrical parameters, which are then input into the preset analysis architecture for calculation.

[0063] Specifically, the preset analysis architecture includes a feature baseline construction unit. This unit extracts electrical waveform data from historical normal operating cycles to construct a multi-dimensional baseline envelope. The feature baseline construction unit extracts electrical waveform data from historical normal operating cycles and divides it into multiple time-series windows using a sliding time slicing method. Adhering to the evolution law of steady-state electrical parameters in the distribution network, the time span of each time-series window is specifically defined as containing 10 to 20 complete power frequency cycles (i.e., 200 to 400 milliseconds in a 50Hz system). This span is sufficient to filter out extremely short-duration asynchronous closing transient impacts while accurately capturing the true envelope of the steady-state waveform. For the voltage amplitude, current RMS value, and harmonic distortion rate within each time-series window, their feature center coordinates and variances are calculated to establish a time-series matrix. The structure of the time series matrix is ​​defined as follows: the row vectors of the matrix are the chronological order of each time series window, and the column vectors are the six feature values ​​(center coordinates and variances corresponding to the voltage amplitude, current RMS value, and harmonic distortion rate calculated and extracted within each time series window), forming a two-dimensional feature space. Principal component analysis (PCA) is used to perform dimensionality reduction on the time series matrix, retaining feature dimensions with a cumulative variance contribution rate of not less than 95%, to obtain principal component feature points, which are then mapped to the spatial coordinate system. The Welzl algorithm (or Support Vector Data Description algorithm) is used to calculate the boundary hypersphere parameters enclosing the principal component feature points. These boundary hypersphere parameters include a multidimensional sphere center coordinate vector and a multidimensional sphere radius scalar. The multidimensional sphere radius scalar is multiplied by an expansion tolerance coefficient to obtain the outer envelope radius, and the boundary surface generated by the outer envelope radius is used as the multidimensional baseline envelope. The external tolerance coefficient is specifically set by extracting the maximum Euclidean distance of the characteristic coordinate drift induced by the alternation of daytime load peaks and valleys and the normal switching of capacitor banks during the historical fault-free period of the target node; the ratio of this maximum drift Euclidean distance to the multidimensional spherical radius scalar, plus a physical safety margin of 0.05 to 0.10, is derived and established as the external tolerance coefficient specific to that node (typically ranging from 1.10 to 1.25). This derivation mechanism ensures that the reference envelope can adaptively and well accommodate reasonable steady-state operating condition fluctuations of the power grid.

[0064] Furthermore, the preset analysis architecture includes a parallel differential comparison unit. This unit performs a parallel comparison between the real-time electrical parameters and the multidimensional reference envelope. When it is determined that the waveform corresponding to the real-time electrical parameters deviates from the multidimensional reference envelope and the deviation exceeds a preset safety threshold, a basic warning result is generated. The parallel differential comparison unit extracts the real-time feature coordinates corresponding to the denoised electrical parameters and calculates the Euclidean distance between these real-time feature coordinates and the multidimensional sphere center coordinate vector. When this Euclidean distance is greater than the outer envelope radius, the ratio of the distance difference to the outer envelope radius is calculated to obtain the deviation magnitude. If the deviation magnitude exceeds the preset safety threshold for a consecutive preset number of sampling points, a basic warning result containing an over-limit label, an associated node identification code, and an alarm timestamp is generated. In this judgment logic, the preset safety threshold is explicitly defined as 10% to 15% of the multidimensional spherical radius scalar under the corresponding time window; the preset quantity is set as the total number of sampling points corresponding to 3 to 5 consecutive power frequency cycles (i.e., at 50Hz power frequency, the duration is 60 to 100 milliseconds) under the current system sampling rate. By imposing dual rigid constraints on the deviation amplitude and duration span, the extremely short-term data exceeding the limit misjudgment caused by a single lightning-induced overvoltage or asymmetric excitation inrush current is effectively eliminated at the physical level.

[0065] Furthermore, the big data basic prediction module also includes a state cache tracing unit. This state cache tracing unit, after generating the basic early warning result, stores the real-time electrical parameters and real-time environmental micro-meteorological parameters of the corresponding node into a distributed cache array, and adds a timestamp tag for subsequent tracing and retrieval. The state cache tracing unit delineates the distributed cache array in the storage area, using the alarm timestamp and node identification code as the primary index key for data storage. When the basic early warning result is received, the state cache tracing unit timestamps the cached data blocks within a set period before and after the early warning time, and modifies the storage attribute of the cached data blocks to read-only mode, so that the corresponding transient waveform data is retained in the memory. The set period range is specifically defined as: a time window that traces back 50 complete power frequency cycles from the alarm timestamp as the origin, and extends forward to extract 100 complete power frequency cycles.

[0066] Furthermore, the system also includes a panoramic topology mapping module. This module constructs a three-dimensional distribution network topology map based on the actual geographical orientation of the target distribution network, marks vulnerable nodes located at wind gaps in the canyon topology map, and highlights the basic early warning results in the three-dimensional distribution network topology map. The panoramic topology mapping module reads elevation model and layer file data to generate the three-dimensional distribution network topology map. The elevation model uses digital elevation model (DEM) raster data with a spatial resolution of at least 30 meters. By extracting the slope parameters of the terrain grid, physical nodes located in the terrain contraction zone and oriented parallel to the prevailing wind direction are marked as vulnerable nodes. Specifically, the terrain contraction zone is defined as a canyon-like area where the terrain grids on both sides perpendicular to the node's orientation converge inwards, and the calculated lateral slope parameters on both sides are greater than a set terrain contraction threshold (e.g., 15 degrees). Upon receiving the basic early warning results, the panoramic topology mapping module parses the physical topological location of the warning node and performs highlighting and coloring rendering on the model area corresponding to that node in the three-dimensional distribution network topology map.

[0067] Furthermore, such as Figure 3 As shown, the meteorological linkage dynamic bypass module includes a micro-meteorological change monitoring unit and a bypass command execution unit. The micro-meteorological change monitoring unit is used to monitor the rising slope of sand concentration and the falling slope of temperature within the real-time environmental micro-meteorological parameters. The micro-meteorological change monitoring unit establishes a dual-path sliding buffer window in the storage area. Given that the evolution of meteorological environmental parameters (wind speed, sand concentration, and temperature) has physical thermal inertia and time lag effects, the time span of the dual-path sliding buffer window is clearly defined to accommodate a continuous data point set of 10 to 15 minutes, and the data update step size is set to 1 minute. This large-scale slowly varying time window effectively filters out sensor sampling glitches caused by local instantaneous gusts, truly reflecting the trend of insulation degradation caused by micro-meteorology. The first sliding buffer window stores the time series of the real-time wind speed index and the air sand concentration index; the second sliding buffer window stores the time series of the real-time temperature change rate. Newly received data is sequentially stored in the corresponding window, and the earliest data at the end is removed synchronously. For the characteristic variable of airborne dust concentration within the window, a linear regression equation is established with discrete time series as independent variables and corresponding dust concentration values ​​as dependent variables. The coefficients corresponding to the independent variables in this linear equation are extracted and used as the slope of the dust concentration increase. Using the same linear regression operation logic, the sequence within the second sliding buffer window is analyzed, and the coefficients corresponding to the independent variables are extracted as the slope of the temperature decrease.

[0068] Specifically, in combination Figure 4The illustrated early warning workflow involves a bypass command execution unit that determines the conditions for a sudden increase in sand concentration accompanied by a sharp drop in temperature when the sand concentration rises above a preset sandstorm warning line and the temperature drop rises above a preset temperature drop threshold line. In this case, a filtering blocking command is sent to the associated node corresponding to the target area to pause the standard noise reduction filtering operation. The bypass command execution unit uses comparison logic to determine whether the sand concentration rises above the preset sandstorm warning line and whether the temperature drop rises above the preset temperature drop threshold line. To ensure the system's actions closely align with the climate-induced disaster characteristics of the target implementation area, the derivation and determination steps for the preset sandstorm warning line and the preset temperature drop threshold line are as follows: Based on the local power distribution network's historical fault log, real physical fault sample points are selected where large-scale deliquescence and flashover of insulator surfaces caused by microclimate deterioration are triggered. For all selected sample points, the maximum abrupt change slope of sand content and the maximum drop slope of temperature within the one-hour time window prior to the fault occurrence are statistically analyzed, forming two sets of historical disaster-inducing meteorological characteristic sample sets. The arithmetic mean of each sample set is calculated and subtracted by one standard deviation to establish the lower limit extreme value of meteorological abrupt change. Based on this, the empirical range of the preset sandstorm warning line is 1.5 mg / (m³·min) to 2.5 mg / (m³·min), and the absolute range of the preset temperature drop threshold line is 0.5℃ / min to 1.2℃ / min. The synchronous exceeding of the limits of the dual physical meteorological characteristics constitutes a reliable benchmark for external causes of insulation degradation, avoiding false triggering of bypass mechanisms. When both conditions are met simultaneously, the system generates the filtering blocking instruction containing the target node's address range and sends it to the associated node. Upon receiving the filtering blocking instruction, the associated node modifies its internal data flow address, changing the data flow pointer from pointing to the digital filter to pointing to the high-frequency analysis buffer, thereby pausing the standard denoising filtering operation.

[0069] Furthermore, the meteorological linkage dynamic bypass module also includes a high-frequency micro-discharge interception unit. This unit is used to extract the un-denoised high-frequency micro-distortion segment containing high-frequency spike pulses from the remaining unfiltered signal after the bypass command execution unit pauses the standard denoising filtering operation. The high-frequency micro-discharge interception unit uses a high-pass filter to filter the remaining unfiltered signal to attenuate the fundamental frequency energy. Specifically, the high-pass filter is an equiripple FIR high-pass digital filter with a cutoff frequency of 10kHz. Within a set time window, the root mean square energy value of the high-pass signal is calculated. When the root mean square energy value is greater than a preset background energy threshold, the current time coordinate is recorded as the distortion start boundary; when the root mean square energy value falls back to within the preset background energy threshold range and remains within a set duration, the current time coordinate is recorded as the distortion termination boundary. To effectively extract high-frequency micro-discharge signals while shielding background white noise, the adaptive determination step of the preset background energy threshold is as follows: During system initialization, a baseline healthy state is selected where the target node is in a clear, windless, and stable temperature condition. The root mean square energy sequence of the residual signal after high-pass filtering is recorded for 10 consecutive power frequency cycles. Based on the classical engineering statistics 3σ criterion, the average root mean square energy of this sequence is calculated and superimposed with three times the root mean square energy standard deviation, which is used as the preset background energy threshold representing the limit of background interference in the power grid. Furthermore, considering the continuous characteristics of early micro-discharge pulse clusters, the set duration is specifically defined as 1 to 2 milliseconds. Data sequences within the corresponding time range are extracted based on the distortion start boundary and distortion termination boundary to generate the undenoised high-frequency micro-distortion segment.

[0070] Furthermore, the meteorological linkage dynamic bypass module also includes a flashover trend analysis unit. This unit performs a lateral differential comparison between the undenoised high-frequency micro-distortion segment of the target node and the normal micro-frequency signal of the adjacent healthy node. If the differential amplitude exceeds a preset discharge threshold, a high-risk flashover warning is sent to the dispatch control center. The flashover trend analysis unit obtains the normal micro-frequency signal of the nearest adjacent healthy node based on the distribution network's graph structure data. To ensure the relative purity of the lateral differential comparison benchmark and avoid extracting signals already contaminated by homogeneous discharge, the adjacent healthy node must be verified by the system and simultaneously pass the following three preset judgment conditions:

[0071] No basic alerts of any level were triggered within the current time window;

[0072] The standard deviation of the fluctuation of the three-phase current RMS value and the voltage RMS value over the past 10 minutes has remained below 3% of the rated nominal value;

[0073] It is located in the same distribution bus section as the target node, but does not belong to the same direct feeder line in terms of topology.

[0074] The system extracts the micro-frequency signal from the node with the shortest straight-line distance in Euclidean space only from the set of nodes that meet the above necessary conditions. Fast Fourier Transform (FFT) is performed on the undenoised high-frequency micro-distortion segment of the target node and the normal micro-frequency signal of the adjacent healthy node to transfer them into the frequency domain. The cross-power spectral density sequence of the two is calculated, and an inverse FFT is performed to generate a cross-correlation function curve. The group delay difference is calculated based on the peak coordinate offset of the cross-correlation function curve. Specifically, during the conversion, the group delay difference is equal to the offset of the discrete sampling point index corresponding to the peak of the cross-correlation function curve, multiplied by the single-step sampling period of the underlying hardware single-channel analog-to-digital converter (e.g., 100 nanoseconds). After obtaining this product result, the microprocessor directly assigns it to the internal timer / counter register as a time shift compensation, thereby achieving strict phase alignment of the two signals at the nanosecond level of the underlying instruction set. The normal micro-frequency signal is time-domain shifted based on the group delay difference to achieve phase alignment of the two signals. After alignment, a point-by-point subtraction operation is performed on the two signals to generate a differential amplitude array. The peak value of the differential amplitude array is extracted. When the peak value is determined to be greater than the preset discharge threshold, the high-risk warning for insulation flashover is encapsulated and sent to the dispatch control center. Considering the crosstalk effect of common-mode electromagnetic interference in the distribution network, the preset discharge threshold is derived as follows: the maximum peak value of the normal micro-frequency signal of the adjacent healthy nodes at the same time segment is extracted, and multiplied by a discharge discrimination sensitivity gain coefficient of 2.5 to 3.0 times to obtain the preset discharge threshold. If the peak value of the differential amplitude exceeds this dynamic benchmark, it proves in an electrophysical sense that the energy jump has significantly exceeded the system's common-mode environmental noise limit, belonging to a substantial specific discharge precursor caused by the deliquescence of conductive pollution.

[0075] Furthermore, upon receiving the high-risk warning for insulation flashover, the dispatch control center, in conjunction with the unified time-stamped data source obtained by the electricity consumption information collection unit, simultaneously initiates six parallel analysis logics. Specifically, the six parallel analysis logics include: line outage information analysis, work order analysis, payment status analysis, user recruitment and curve analysis, meter box recruitment and curve analysis, and transformer area recruitment and curve analysis.

[0076] In terms of specific judgment logic: the first round of work order analysis is used to determine whether the user, line, branch line, transformer area, and meter box are experiencing power outages, so as to quickly locate the scope of impact; the payment status analysis is used to determine whether the user is experiencing a power outage due to unpaid bills; the recall and curve analysis (covering users, meter boxes, and transformer areas) are used to comprehensively determine whether the specific cause of the warning is a tripped circuit breaker of the user or an electrical fault in the meter, meter box, or transformer area itself.

[0077] To balance the timeliness and accuracy of emergency response to sudden faults, the system established two modes—accurate assessment and precise assessment—through multiple rounds of experiments. Accurate assessment prioritizes response speed and sets strict rules: it expects to output the first round of emergency repair work orders and power outage information assessment results within 5 seconds, and the voltage and current recall results within 30 seconds. To ensure the 5-second time limit, the accurate assessment mode is configured to directly call static topology cache data via hardware interrupts and forcibly bypass time-domain to frequency-domain Fourier transforms and floating-point envelope comparisons, extracting only the Boolean flags indicating switch over-limits for logical algebraic matching. Precise assessment prioritizes location accuracy, achieving precise determination of the fault cause and location based on cross-validation of the six parallel data streams. The system ultimately outputs a comprehensive assessment result to maintenance personnel, including the power outage type, fault location, fault cause, and repair priority.

[0078] Furthermore, in order to achieve automated qualitative analysis of fault causes using the precise judgment mode, the system has a preset logical mapping rule matrix based on the six parallel data streams. The specific extraction steps and judgment logic are as follows:

[0079] User internal fault determination: When the power outage information of the line is determined to be no power outage, the payment status is normal, and the voltage of the meter box / transformer area is normal, but the voltage of the user terminal's voltage of the user terminal disappears, the logical mapping result is that the user side circuit breaker trips or there is an internal line fault.

[0080] Determination of power outage due to overdue payment: When the payment status assessment result is overdue payment and the user's survey feedback indicates that the voltage has disappeared, the logical mapping result is remote power outage due to overdue payment.

[0081] Equipment failure determination: When the power outage information is determined to be no power outage, but the specific meter box or meter fails to be detected three times in a row, and the voltage feedback from its upstream distribution area is normal, the logic mapping result is that the meter box / meter body is damaged or the wiring is loose.

[0082] Common fault determination on the distribution network side: When the line outage information is determined to be a power outage, the voltage feedback from the transformer area is lost, and the work order is determined to have a repair request along the same path, the logical mapping result is a planned power outage or a sudden trip on the distribution network side.

[0083] The system uses the above Boolean logic combination to perform real-time matching of the six data streams, thereby establishing the fault cause field in the comprehensive judgment result.

[0084] Furthermore, such as Figure 5 As shown, the topology tracing and congestion prevention scheduling module includes a congestion situation awareness unit. This unit continuously monitors the concurrent occupancy rate of the large-capacity fault data uploaded by each associated node. When the concurrent occupancy rate exceeds the communication bandwidth load limit, the congestion prevention scheduling logic is activated. Combined with... Figure 6 The congestion control process illustrated involves the congestion situation awareness unit periodically reading the number of cached bytes and connection parameters of the transport layer input queue, and comprehensively calculating the current physical link concurrency occupancy rate. The concurrency occupancy rate is calculated as follows: extract the total number of cached bytes in the transport layer within the current statistical period, multiply it by 8 to convert it into the total number of bits to be sent; divide this total number of bits by the duration of the statistical period to obtain the real-time concurrent throughput demand rate; then divide this demand rate by the nominal maximum bandwidth throughput rate allocated to the network segment by the physical communication link to obtain a dimensionless percentage value as the concurrency occupancy rate. The concurrency occupancy rate is then compared with a preset communication bandwidth load limit; wherein the communication bandwidth load limit is explicitly defined as 80% to 85% of the nominal maximum bandwidth throughput rate of the target distribution network physical communication link. The physical purpose of this setting is to forcibly reserve 15% to 20% of redundant channel bandwidth resources, dedicated to carrying control commands with stringent real-time requirements, such as relay protection tripping signaling and positioning satellite timing heartbeat packets, thereby fundamentally preventing the risk of grid infrastructure control network failure due to congestion caused by massive transient waveform recordings. When the concurrent occupancy rate is determined to exceed the communication bandwidth load limit, the flag in the system scheduling and control register is set to activate the anti-congestion scheduling logic.

[0085] Furthermore, the topology tracing and congestion prevention scheduling module includes a topology and electrical evaluation unit. This unit, after activating the congestion prevention scheduling logic, calculates the topology level depth and the zero-sequence current mutation slope of the associated nodes concurrently uploading the large-capacity fault data. The calculation process for the topology level depth includes: extracting the directed acyclic graph data structure generated by transforming the 3D distribution network topology graph; the specific mapping transformation rule is as follows: devices with electrical entity breaking or power aggregation attributes, such as substation outgoing circuit breakers, branch load switches, and distribution transformers in the geographic space, are uniformly abstracted and extracted as graph nodes; and physical overhead line segments are abstracted as associated edges with specific power flow directionality (strictly from the power source side to the load side). This mapping rule effectively maps the geographic model containing complex 3D elevations to a standard electrical network graph structure suitable for breadth-first search. Using the substation outgoing node as the root node, a breadth-first search algorithm is used to traverse along the node's associated edges to the branch child nodes. For each cascaded level traversed downwards, the depth attribute value increases by a set step size scalar. The step size scalar set here is fixed at an integer value of 1, which actually maps the number of physical series stages from the upstream substation bus to the downstream edge branch switch in the distribution network. When traversing and addressing to the associated node, the accumulated depth value is output as the topology level depth. The calculation process for the zero-sequence current mutation slope includes: extracting the zero-sequence current waveform sequence of the associated node; calculating the ratio of the difference between adjacent discrete sampling points to the time step difference within a specific time window; the span of the specific time window is defined as one-quarter of the power frequency cycle after the detection of the mutation start time (i.e., 5 milliseconds at 50Hz), so as to accurately lock the initial sharp rising edge characteristics of the zero-sequence current before the transient peak value is reached at the moment of single-phase ground fault. The arithmetic mean of the extracted ratios is calculated and the unsigned absolute value is taken as the zero-sequence current mutation slope.

[0086] Specifically, the topology tracing and congestion prevention scheduling module also includes a dynamic credential issuance unit. This unit calculates a comprehensive evaluation score based on the topology layer depth and the zero-sequence current mutation slope, marks the associated node with the highest comprehensive evaluation score as the core hub node, and issues the full feature upload credential to the core hub node, enabling it to obtain the highest network priority to transmit the complete transient waveform file. The dynamic credential issuance unit establishes a weighted scoring function, assigning weights to the reciprocal of the topology layer depth and the zero-sequence current mutation slope corresponding to each associated node, to calculate the comprehensive evaluation score.

[0087] Specifically, the detailed analytical mathematical formula of the weighted scoring function is as follows: .

[0088] In the formula, Representing the The overall evaluation score of each associated node; Represents the topological level depth corresponding to the node (its value is an integer variable or feature parameter greater than or equal to 1). This represents the slope of the zero-sequence current abrupt change extracted from the previous time window at that node; This is the maximum absolute value of the slope of the zero-sequence current mutation at each alarm node in the current concurrent queue, and is used to strictly perform dimensionless normalization. If... If the value is 0, then a minimal non-zero constant is superimposed during the actual denominator calculation. (like To prevent arithmetic overflow.

[0089] parameter and The network structure priority weights and electrical transient intensity weights are respectively represented, and the two satisfy the following conditions: Based on the underlying objective laws governing distribution network emergency repair analysis: the severity of transient zero-sequence current distortion can directly reflect the physical source of insulation breakdown, and its scheduling importance should take precedence over a single topological depth that only represents spatial distance. Therefore, the system directly locks the weight distribution boundaries based on engineering guidance: optimal setting... and This analytical expression relies on conventional normalization and explicit non-equal weighting, and can directly output a highly recognizable descending queue without iterative optimization. Its computational logic closely aligns with the protection action criteria of the power system.

[0090] The sorting algorithm is invoked to sort the alarm node queue in the concurrent queuing sequence in descending order based on the comprehensive evaluation score. The node at the top of the sequence is marked as the core hub node, and a full feature upload credential with an additional priority identifier is generated and sent to that node. The core hub node performs network scheduling transmission based on the full feature upload credential and uploads the complete transient waveform file.

[0091] Furthermore, the topology tracing and congestion prevention scheduling module includes an edge node degradation unit. This unit issues a forced degradation instruction to edge branch nodes that do not possess the full feature upload credential; upon receiving the forced degradation instruction, the edge branch node enters the degraded transmission state. Specifically, the edge node degradation unit generates the forced degradation instruction containing a data simplification opcode and delivers it to the edge branch node that does not possess the upload credential via network transmission. After parsing the forced degradation instruction, the edge branch node closes the high-speed transmission channel used for sending the full original waveform data and enters the restricted degraded transmission state.

[0092] Furthermore, the topology tracing and congestion prevention scheduling module includes a feature summary extraction unit. The feature summary extraction unit sends a feature extraction instruction to the edge branch node in the degraded transmission state, instructing the node to extract only the effective voltage and current values ​​from its own transient waveform file. The effective current value is compared with a set limit value to generate an over-limit flag. The effective voltage value, the effective current value, and the over-limit flag are combined into a lightweight feature summary for uploading, thereby reducing network bandwidth usage and preventing massive concurrent data from blocking the network and affecting the overall analysis process. In the degraded transmission state, the processor inside the edge branch node extracts sampling points of discrete voltage and current waveform sequences, calculates the root mean square parameter, and outputs the effective voltage and current values. The acquired effective current value is compared with a set limit value. If the effective current value is greater than the limit value, a corresponding status bit is set to generate the over-limit flag. In this step, the physical definition of the set limit value is equivalent to the instantaneous trip or overcurrent protection setting current threshold configured for the edge branch line. It is obtained by directly reading 1.2 to 1.5 times the historical maximum continuous operating load peak current in the register of the corresponding edge branch node protection device. This mapping ensures that although nodes in degraded transmission mode cannot upload massive broadband waveforms, the lightweight over-limit flags they report still possess the core practical value of reflecting the over-limit tripping status of the distribution network. The processor packages and encapsulates the voltage RMS value, current RMS value, and over-limit flags generated by the above calculations into a lightweight feature summary with reduced data volume, and then performs the communication upload operation.

[0093] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A system for multi-path parallel fault prediction and accurate assessment in distribution networks based on big data analysis, comprising: The multi-channel parallel acquisition module is used to simultaneously acquire real-time electrical parameters and real-time environmental micro-meteorological parameters of multiple related nodes in the target area's power distribution network; The big data basic prediction module is used to perform standard denoising filtering on the real-time electrical parameters and input the denoised electrical parameters into a preset analysis architecture to output basic early warning results. The feature is that the distribution network multi-path parallel fault prediction and accurate judgment system based on big data analysis further includes: The meteorological linkage dynamic bypass module is used to generate a filtering blocking command and send it to the associated node corresponding to the target area when the real-time environmental micro-meteorological parameters meet the conditions of a sudden increase in wind and sand concentration accompanied by a sudden drop in temperature. The standard denoising filtering operation is paused, and the high-frequency micro-distortion fragments that have not been denoised are extracted. Based on the high-frequency micro-distortion fragments, a high-risk warning for insulation flashover is output. The topology tracing and congestion prevention scheduling module is used to comprehensively evaluate the topology depth of each node within the distribution network structure and the slope of zero-sequence current mutation when multiple associated nodes concurrently upload large amounts of fault data containing high-frequency micro-distortion segments and transient waveform files due to the execution of the filtering and blocking instructions, which leads to the saturation of communication computing power. The module marks the node with the highest evaluation score as the core hub node and issues it a full feature upload certificate. Other nodes that do not hold the full feature upload certificate are forced to enter a degraded transmission state as edge branch nodes and only upload lightweight feature summaries. This ensures that subsequent processing computing power is focused on the high-risk early warning and judgment of insulation pollution flashover of the core hub node.

2. The system for multi-path parallel fault prediction and accurate judgment of distribution networks based on big data analysis as described in claim 1, characterized in that, The multi-channel parallel acquisition module includes: The high-frequency synchronous sampling unit is used to perform synchronous data acquisition on multiple parallel branch lines in the target area distribution network using the positioning satellite clock signal, and to obtain three-phase voltage waveform data, three-phase current waveform data and zero-sequence current change as the real-time electrical parameters. The micro-meteorological sensing and collection unit is used to acquire the real-time wind speed index, air sand concentration index, and real-time temperature change rate of the multiple associated nodes as the real-time environmental micro-meteorological parameters. The electricity information collection unit is used to acquire user payment status, electricity load, meter operating status, meter box voltage and current, and transformer area operating parameters within the target area distribution network, and to form a unified time-stamped data source.

3. The system for multi-path parallel fault prediction and accurate judgment of distribution networks based on big data analysis as described in claim 2, characterized in that, The big data-based prediction module includes: The feature baseline construction unit is used to extract electrical waveform data within historical normal operating cycles to construct a multidimensional reference envelope. The parallel differential comparison unit is used to perform a parallel comparison between the real-time electrical parameters and the multidimensional reference envelope. When it is determined that the waveform corresponding to the real-time electrical parameters deviates from the multidimensional reference envelope and the deviation exceeds a preset safety threshold, the basic warning result is generated.

4. The system for multi-path parallel fault prediction and accurate judgment of distribution networks based on big data analysis as described in claim 3, characterized in that, The big data basic prediction module also includes: The status cache traceability unit is used to store the real-time electrical parameters and real-time environmental micro-meteorological parameters of the corresponding node into a distributed cache array after generating the basic early warning result, and to add a timestamp tag for subsequent traceability and retrieval.

5. The system for multi-path parallel fault prediction and accurate judgment of distribution networks based on big data analysis according to claim 4, characterized in that, Also includes: The panoramic topology mapping module is used to construct a three-dimensional distribution network topology map based on the actual geographical orientation of the target distribution network, mark vulnerable nodes located at the wind gap of the canyon in the three-dimensional distribution network topology map, and display the basic early warning results in the three-dimensional distribution network topology map by coloring and highlighting.

6. The system for multi-path parallel fault prediction and accurate judgment of distribution networks based on big data analysis as described in claim 1, characterized in that, The meteorological linkage dynamic bypass module includes: The micro-meteorological change monitoring unit is used to monitor in real time the slope of the increase in sand content and the slope of the decrease in temperature within the real-time environmental micro-meteorological parameters. The bypass instruction execution unit is used to determine that the conditions of a sudden increase in sand concentration accompanied by a sudden drop in temperature are met when the slope of the increase in sand concentration exceeds the preset sandstorm warning line and the slope of the decrease in temperature exceeds the preset temperature drop threshold line. At this time, the unit sends the filtering blocking instruction to the associated node corresponding to the target area to suspend the execution of the standard noise reduction filtering operation.

7. The system for multi-path parallel fault prediction and accurate judgment of distribution networks based on big data analysis as described in claim 6, characterized in that, The meteorological linkage dynamic bypass module also includes: A high-frequency micro-discharge interception unit is used to acquire the undenoised original signal as the remaining unfiltered signal after the bypass instruction execution unit suspends the standard denoising and filtering operation, and to intercept the undenoised high-frequency micro-distortion segment including high-frequency spike pulses from the remaining unfiltered signal. The pollution flashover trend analysis unit is used to take the associated node that generates the high-frequency micro-distortion segment as the target node, and obtain the normal micro-frequency signal of the adjacent associated node that meets the preset normal operation conditions; perform a lateral differential comparison between the undenoised high-frequency micro-distortion segment of the target node and the normal micro-frequency signal; if the differential amplitude exceeds the preset discharge threshold, the high-risk warning of insulation pollution flashover is sent to the dispatch control center.

8. The system for multi-path parallel fault prediction and accurate judgment of distribution networks based on big data analysis according to claim 1, characterized in that, The topology tracing and congestion prevention scheduling module includes: The congestion situation awareness unit is used to continuously count the concurrent occupancy rate of each of the associated nodes uploading the large-capacity fault data. When the concurrent occupancy rate exceeds the communication bandwidth load limit, the congestion prevention scheduling logic is activated. The topology and electrical evaluation unit is used to calculate the topology level depth and the zero-sequence current mutation slope of the associated nodes that concurrently upload the large-capacity fault data after the congestion prevention scheduling logic is activated.

9. The system for multi-path parallel fault prediction and accurate judgment of distribution networks based on big data analysis as described in claim 8, characterized in that, The topology tracing and congestion prevention scheduling module also includes: The credential dynamic issuance unit is used to calculate a comprehensive evaluation score based on the topology level depth and the zero-sequence current mutation slope, mark the associated node with the highest comprehensive evaluation score as the core hub node, and issue the full feature upload credential to the core hub node, so that the core hub node can obtain the highest level network priority to transmit the complete transient waveform file.

10. The system for multi-path parallel fault prediction and accurate judgment of distribution networks based on big data analysis according to claim 9, characterized in that, The topology tracing and congestion prevention scheduling module also includes: An edge node degradation unit is used to issue a forced degradation instruction to the edge branch node that does not hold the full feature upload certificate; the edge branch node enters the degradation transmission state after receiving the forced degradation instruction; The feature summary extraction unit is used to issue feature extraction instructions to the edge branch nodes in the degraded transmission state, instructing the node to extract only the effective voltage and current values ​​within its own transient waveform file, compare the effective current value with a set limit value to generate an over-limit flag, and combine the effective voltage value, the effective current value, and the over-limit flag to form the lightweight feature summary for uploading, thereby reducing network bandwidth usage and preventing massive concurrent data from blocking the network and affecting the overall judgment process.

Images