A power line carrier metering data acquisition early warning method and system

Abstract

The application discloses a kind of power line carrier's metering data acquisition early warning method and system, it is related to information analysis technical field, the method includes: the power consumption information of each user node in power line network is preprocessed;Based on the power consumption information after preprocessing, transient event feature extraction is carried out using morphological operation to determine the target power consumption behavior characteristics of each user node;Based on target power consumption behavior characteristics, the metering data acquisition early warning model between user nodes is constructed;Based on metering data acquisition early warning model, abnormal event identification is carried out to determine early warning level;Based on abnormal event identification result, attribution positioning is carried out using waveform correlation degree;Based on abnormal event identification result, early warning level determination result and attribution positioning information, early warning message is generated, and early warning message is transmitted to operation and maintenance terminal.The application provides comprehensive enough early warning information, so that operation and maintenance personnel can directly carry out disposal action without secondary research and judgment, and significantly shorten abnormal response chain.

Description

Technical Field

[0001] This invention relates to the field of information analysis technology, and in particular to a method and system for collecting and warning metering data using power line carrier. Background Technology

[0002] In power line carrier communication-based metering data acquisition systems, accurately and promptly detecting and issuing early warnings of abnormal user electricity consumption behavior is crucial for improving the intelligence level of power distribution and utilization. Existing methods for detecting abnormal electricity consumption largely rely on steady-state electrical quantities, lacking sufficient ability to capture transient event characteristics, resulting in incomplete data support for abnormal event identification and early warning. Furthermore, when an anomaly is detected, current technologies often only conclude that an anomaly exists, but struggle to automatically trace the source of the anomaly or determine the event type through waveform characteristics. This leads to early warning information often containing only time, node, and simple threshold exceedance markers, lacking multi-dimensional information such as event category, severity, and fault location. This necessitates extensive manual analysis by maintenance personnel to determine the fault location and cause, resulting in low efficiency in power line network operation and maintenance. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides a method and system for collecting and issuing early warnings of metering data for power line carrier communication. It provides sufficiently comprehensive early warning information, enabling maintenance personnel to directly carry out handling actions without secondary analysis, and significantly shortening the abnormal response chain.

[0004] To address the aforementioned technical problems, this invention provides a method for collecting and issuing early warning data from power line carrier circuits, the method comprising: The power consumption information of each user node in the power line network is obtained based on the power line carrier acquisition module, and the power consumption information is preprocessed to obtain the preprocessed power consumption information. Based on the preprocessed electricity consumption information, transient event features are extracted using morphological operations to obtain transient event feature information, and the target electricity consumption behavior features of each user node are determined based on the transient event feature information. A metering data collection and early warning model between user nodes is constructed based on the target electricity consumption behavior characteristics. Based on the aforementioned measurement data acquisition and early warning model, abnormal events are identified to obtain abnormal event identification results. Based on the abnormal event identification results, an early warning level is determined to obtain an early warning level determination result. Based on the abnormal event identification results, attribution and localization are performed using waveform correlation to obtain attribution and localization information; Based on the abnormal event identification results, early warning level determination results, and attribution location information, an early warning message is generated and transmitted to the operation and maintenance terminal.

[0005] Optionally, the step of preprocessing the electricity consumption information to obtain preprocessed electricity consumption information includes: The power consumption information is decomposed into multiple channels to obtain wavelet decomposition coefficients, and the wavelet decomposition coefficients are then subjected to noise component suppression to obtain noise component-suppressed wavelet decomposition coefficients. Inverse wavelet transform is performed on the wavelet decomposition coefficients after noise component suppression to obtain preprocessed electricity consumption information.

[0006] Optionally, the step of extracting transient event features using morphological operations based on the preprocessed electricity consumption information to obtain transient event feature information, and determining the target electricity consumption behavior features of each user node based on the transient event feature information, includes: Extreme point detection is performed based on the preprocessed electricity consumption information to obtain an extreme point sequence, and zero-crossing point detection is performed based on the preprocessed electricity consumption information to obtain a zero-crossing point sequence. Based on the extreme point sequence and the zero-crossing point sequence, a multi-scale structural element is constructed, and the preprocessed electricity consumption information is subjected to morphological gradient enhancement based on the multi-scale structural element to obtain the morphological gradient enhancement result. Transient event feature extraction is performed based on the morphological gradient enhancement results to obtain transient event feature information; A steady-state interval is determined, and a background context feature vector is determined based on the steady-state interval and the preprocessed electricity consumption information. The target electricity consumption behavior characteristics of each user node are then determined based on the transient event feature information and the background context feature vector.

[0007] Optionally, the step of extracting transient event features based on the morphological gradient enhancement results to obtain transient event feature information includes: Based on the multi-scale structural elements, high-frequency transient features are extracted from the preprocessed electricity consumption information to obtain high-frequency transient feature information; Based on the multi-scale structural elements, low-frequency discontinuity features are extracted from the preprocessed electricity consumption information to obtain low-frequency discontinuity feature information. Based on the morphological gradient enhancement results, high-frequency transient feature information and low-frequency discontinuity feature information, a fused temporal transient feature map is generated, and the transient event region is determined based on the temporal transient feature map; Based on the transient event region, the time-domain waveform segment of the transient event is determined, and the transient event features are extracted based on the time-domain waveform segment of the transient event to obtain transient event feature information.

[0008] Optionally, the transient event feature extraction based on the time-domain waveform segment of the transient event to obtain transient event feature information includes: Based on the time-domain waveform segment, determine the duration, peak amplitude, energy decay slope, dominant oscillation frequency, and phase angle of the voltage zero-crossing sequence relative to the extreme point sequence and the zero-crossing sequence. Based on the duration, peak amplitude, energy decay slope, dominant oscillation frequency, and phase angle of the voltage zero-crossing sequence relative to the extreme point sequence and the zero-crossing sequence, determine the quantization parameter information of the transient event. Based on the quantization parameter information, a two-dimensional grid image is mapped to obtain a two-dimensional feature distribution map; Transient event information is determined based on the two-dimensional feature distribution map.

[0009] Optionally, the step of constructing a metering data collection and early warning model between user nodes based on the target electricity consumption behavior characteristics includes: Acquire power line network topology data and construct an electrical distance matrix between nodes based on the power line network topology data; Based on the preprocessed power consumption information, the dynamic correlation data of the carrier signals between nodes is determined, and a dynamic graph structure is constructed based on the electrical distance matrix between nodes and the dynamic correlation data of the carrier signals between nodes. Based on the target electricity consumption behavior characteristics, the time-domain pattern vector of each user node is determined, and based on the time-domain pattern vector and dynamic graph structure, a spatiotemporal graph attention network is used to construct a metering data collection and early warning model among user nodes.

[0010] Optionally, the step of identifying abnormal events based on the metering data acquisition and early warning model to obtain abnormal event identification results, and determining the early warning level based on the abnormal event identification results to obtain an early warning level determination result, includes: Based on the metering data acquisition and early warning model and the target electricity consumption behavior characteristics, a real-time reconstructed feature vector is determined. Based on the real-time reconstructed feature vector and the target electricity consumption behavior characteristics, a deviation vector analysis is performed to obtain a multi-dimensional deviation vector. Anomaly identification results are obtained by performing anomaly identification based on the multidimensional deviation vector. Based on the abnormal event identification results, the severity of single-node anomalies is assessed to obtain the severity assessment results of single-node anomalies. Based on the abnormal event identification results, an abnormal propagation trend analysis is performed to obtain the abnormal propagation trend analysis results. Based on the single-node abnormality severity assessment results and the abnormal propagation trend analysis results, the warning level is determined to obtain the warning level determination results.

[0011] Optionally, the step of identifying abnormal events based on the multidimensional deviation vector to obtain abnormal event identification results includes: Determine the boundary of the dynamic confidence ellipse, and perform anomaly initial state analysis of user nodes based on the boundary of the dynamic confidence ellipse and the multidimensional deviation vector to obtain anomaly initial state labels; Duration analysis is performed based on the abnormal initial state markers to obtain duration information; Anomaly event identification is performed based on the duration information to obtain anomaly event identification results.

[0012] Optionally, the step of using waveform correlation to perform attribution localization based on the abnormal event identification result to obtain attribution localization information includes: Based on the abnormal event identification results, abnormal waveform segment analysis is performed to obtain the target abnormal waveform segment; Based on the target abnormal waveform segment, anomaly attribution is performed using a fault waveform library to obtain anomaly attribution information. Based on the target abnormal waveform segment, waveform correlation analysis is performed between the abnormal node and the upstream and downstream nodes to obtain waveform correlation information. Based on the waveform correlation information, disturbance propagation path analysis is performed to obtain disturbance propagation path analysis results. Attribution location is performed based on the anomaly attribution information and disturbance propagation path analysis results to obtain attribution location information.

[0013] In addition, the present invention also provides a power line carrier metering data acquisition and early warning system, the system comprising: Information preprocessing module: used to acquire electricity consumption information of each user node in the power line network based on the power line carrier acquisition module, preprocess the electricity consumption information, and obtain preprocessed electricity consumption information; Feature extraction module: used to extract transient event features based on preprocessed electricity consumption information using morphological operations, obtain transient event feature information, and determine the target electricity consumption behavior features of each user node based on the transient event feature information; Model building module: used to build a metering data collection and early warning model between user nodes based on the target electricity consumption behavior characteristics; Early warning determination module: used to identify abnormal events based on the metering data acquisition early warning model, obtain abnormal event identification results, and determine the early warning level based on the abnormal event identification results, and obtain the early warning level determination result; Attribution localization module: used to perform attribution localization based on waveform correlation according to the abnormal event identification results, and obtain attribution localization information; Early warning message transmission module: used to generate early warning messages based on the abnormal event identification results, early warning level determination results and attribution location information, and transmit the early warning messages to the operation and maintenance terminal.

[0014] In this embodiment of the invention, the electricity consumption information of each user node in the power line network is preprocessed; based on the preprocessed electricity consumption information, transient event features are extracted using morphological operations; based on the transient event feature information, the target electricity consumption behavior characteristics of each user node are determined, improving the reliability of transient event feature extraction and providing more comprehensive data support for subsequent abnormal event identification and early warning. A metering data acquisition and early warning model between user nodes is constructed based on the target electricity consumption behavior characteristics; abnormal event identification is performed based on the metering data acquisition and early warning model, improving the accuracy of abnormal event identification. Early warning level is determined based on the abnormal event identification results; attribution and location are performed based on waveform correlation based on the abnormal event identification results; an early warning message is generated based on the abnormal event identification results, early warning level determination results, and attribution and location information, and transmitted to the operation and maintenance terminal. This completes the information flow from data acquisition and intelligent analysis to the operation and maintenance closed loop, providing sufficiently comprehensive early warning information, enabling operation and maintenance personnel to directly carry out disposal actions without secondary analysis, significantly shortening the abnormal response chain and improving the level of electricity management. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating the metering data acquisition and early warning method for power line carrier in an embodiment of the present invention. Figure 2 This is a flowchart illustrating a power line carrier metering data acquisition and early warning method according to another embodiment of the present invention. Figure 3 This is a schematic diagram of the structural composition of the power line carrier metering data acquisition and early warning system in an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1 Please see Figure 1 , Figure 1This is a flowchart illustrating the metering data acquisition and early warning method for power line carrier communication according to an embodiment of the present invention. The method includes: S11: Obtain the power consumption information of each user node in the power line network based on the power line carrier acquisition module, preprocess the power consumption information, and obtain the preprocessed power consumption information; In the specific implementation of this invention, the power consumption information of each user node in the power line network is acquired based on the power line carrier acquisition module. The power consumption information is then subjected to channel multi-band decomposition to obtain wavelet decomposition coefficients. Noise component suppression is then applied to the wavelet decomposition coefficients to obtain noise component-suppressed wavelet decomposition coefficients. Based on the noise component-suppressed wavelet decomposition coefficients, inverse wavelet transform is performed to obtain preprocessed power consumption information. This provides a high-quality data foundation for subsequent feature extraction and avoids subsequent feature distortion and misjudgment due to data quality issues.

[0019] S12: Based on the preprocessed electricity consumption information, morphological operations are used to extract transient event features to obtain transient event feature information, and the target electricity consumption behavior features of each user node are determined based on the transient event feature information. In the specific implementation of this invention, extreme point detection is performed based on preprocessed electricity consumption information to obtain an extreme point sequence, and zero-crossing point detection is performed based on the preprocessed electricity consumption information to obtain a zero-crossing point sequence. Multi-scale structural elements are constructed based on the extreme point sequence and the zero-crossing point sequence, and morphological gradient enhancement is performed on the preprocessed electricity consumption information based on the multi-scale structural elements to obtain morphological gradient enhancement results. Transient event features are extracted based on the morphological gradient enhancement results to obtain transient event feature information. A steady-state interval is determined, and a background context feature vector is determined based on the steady-state interval and the preprocessed electricity consumption information. Finally, the target electricity consumption behavior features of each user node are determined based on the transient event feature information and the background context feature vector. This improves the reliability of transient event feature extraction and provides more comprehensive data support for subsequent abnormal event identification and early warning.

[0020] S13: Construct a metering data collection and early warning model between user nodes based on the target electricity consumption behavior characteristics; In the specific implementation of this invention, power line network topology data is acquired, and an electrical distance matrix between nodes is constructed based on the power line network topology data; dynamic correlation data of carrier signals between nodes is determined based on preprocessed electricity consumption information, and a dynamic graph structure is constructed based on the electrical distance matrix between nodes and the dynamic correlation data of carrier signals between nodes; the time-domain pattern vector of each user node is determined based on the target electricity consumption behavior characteristics, and a spatiotemporal graph attention network is used to construct a metering data collection and early warning model between user nodes based on the time-domain pattern vector and the dynamic graph structure, which can improve the accuracy of abnormal event identification.

[0021] S14: Based on the metering data acquisition and early warning model, perform abnormal event identification to obtain abnormal event identification results, and based on the abnormal event identification results, determine the early warning level to obtain the early warning level determination result; In the specific implementation of this invention, a real-time reconstructed feature vector is determined based on the metering data acquisition and early warning model and the target electricity consumption behavior characteristics. Deviation vector analysis is then performed based on the real-time reconstructed feature vector and the target electricity consumption behavior characteristics to obtain a multi-dimensional deviation vector. Anomaly identification is performed based on the multi-dimensional deviation vector to obtain anomaly identification results. Single-node anomaly severity assessment is performed based on the anomaly identification results to obtain single-node anomaly severity assessment results. Anomaly propagation trend analysis is performed based on the anomaly identification results to obtain anomaly propagation trend analysis results. Finally, an early warning level determination is performed based on the single-node anomaly severity assessment results and the anomaly propagation trend analysis results to obtain an early warning level determination result. This achieves intelligent classification and risk grading of anomaly events, enabling maintenance resources to prioritize high-risk events and improve the accuracy and timeliness of maintenance.

[0022] S15: Based on the abnormal event identification results, use waveform correlation to perform attribution localization and obtain attribution localization information; In the specific implementation of this invention, abnormal waveform segments are analyzed based on the abnormal event identification results to obtain target abnormal waveform segments; abnormal attribution is performed using a fault waveform library based on the target abnormal waveform segments to obtain abnormal attribution information; waveform correlation analysis is performed between the abnormal node and its upstream and downstream nodes based on the target abnormal waveform segments to obtain waveform correlation information, and disturbance propagation path analysis is performed based on the waveform correlation information to obtain disturbance propagation path analysis results; attribution location is performed based on the abnormal attribution information and disturbance propagation path analysis results to obtain attribution location information, providing clear attribution location information, which can significantly shorten the time for fault investigation and electricity theft verification, and improve operation and maintenance efficiency.

[0023] S16: Generate an early warning message based on the abnormal event identification result, early warning level determination result, and attribution location information, and transmit the early warning message to the operation and maintenance terminal.

[0024] In the specific implementation of this invention, an early warning message is generated based on the abnormal event identification result, the early warning level determination result, and the attribution location information, and the early warning message is transmitted to the operation and maintenance terminal. The content of the early warning information is comprehensive and is directly pushed to the operation and maintenance terminal through the communication channel, which allows operation and maintenance personnel to grasp the early warning information at the first time, thereby improving the reliability of power supply and the level of power management.

[0025] In this embodiment of the invention, the electricity consumption information of each user node in the power line network is preprocessed; based on the preprocessed electricity consumption information, transient event features are extracted using morphological operations; based on the transient event feature information, the target electricity consumption behavior characteristics of each user node are determined, improving the reliability of transient event feature extraction and providing more comprehensive data support for subsequent abnormal event identification and early warning. A metering data acquisition and early warning model between user nodes is constructed based on the target electricity consumption behavior characteristics; abnormal event identification is performed based on the metering data acquisition and early warning model, improving the accuracy of abnormal event identification. Early warning level is determined based on the abnormal event identification results; attribution and location are performed based on waveform correlation based on the abnormal event identification results; an early warning message is generated based on the abnormal event identification results, early warning level determination results, and attribution and location information, and transmitted to the operation and maintenance terminal. This completes the information flow from data acquisition and intelligent analysis to the operation and maintenance closed loop, providing sufficiently comprehensive early warning information, enabling operation and maintenance personnel to directly carry out disposal actions without secondary analysis, significantly shortening the abnormal response chain and improving the level of electricity management.

[0026] Example 2 Please see Figure 2 , Figure 2 This is a flowchart illustrating a power line carrier metering data acquisition and early warning method according to another embodiment of the present invention, the method comprising: S201: Obtain the power consumption information of each user node in the power line network based on the power line carrier acquisition module, preprocess the power consumption information, and obtain the preprocessed power consumption information; In the specific implementation of this invention, the preprocessing of the electricity consumption information to obtain preprocessed electricity consumption information includes: performing channel multi-band decomposition on the electricity consumption information to obtain wavelet decomposition coefficients, and performing noise component suppression on the wavelet decomposition coefficients to obtain noise component suppressed wavelet decomposition coefficients; and performing inverse wavelet transform based on the noise component suppressed wavelet decomposition coefficients to obtain preprocessed electricity consumption information.

[0027] Specifically, the power line carrier acquisition module acquires the power consumption information of each user node in the power line network. The power line carrier acquisition module is an edge intelligent unit that deeply integrates high-precision synchronous acquisition and power line carrier communication functions. Utilizing the master clock source deployed in the distribution area aggregation terminal, it periodically broadcasts synchronization beacon frames containing high-precision timestamps through the power line carrier channel. The power line carrier acquisition module of each user node corrects the local clock phase offset of its temperature-compensated crystal oscillator according to the timestamp of the received synchronization beacon frame, realizing microsecond-level synchronization of the sampling clock of all nodes in the network. Based on the unified synchronization clock trigger, the analog-to-digital converter built into each node samples the voltage and current analog signals on the power line at a fixed sampling rate much higher than the power frequency at the same time, generating a high-time-resolution original voltage and current waveform sampling sequence that is strictly aligned in time. According to the unified time tag, the continuous sampling sequence is divided into data frames of a fixed length of several power frequency cycles. Each frame of data is accompanied by a unified timestamp of its starting sampling point across the entire network, thus obtaining the power consumption information of each user node.

[0028] The power consumption information is subjected to multi-band channel decomposition. The automatic gain control parameters and received signal strength indication values ​​synchronously collected with the data frame inside the carrier communication module are read in real time. Combined with the load fluctuation information caused by the switching of electrical equipment on the current line, the current background noise energy level and probability distribution of burst impulse noise of the power line channel are estimated to form a real-time noise basis estimate. Based on the selected lifting wavelet basis, the power consumption information is subjected to multi-level lifting wavelet decomposition and mapped to sub-bands of different frequencies. The high-frequency sub-band captures noise and transient spikes, while the low-frequency sub-band retains the power frequency fundamental wave and the slowly varying envelope, thus obtaining the wavelet decomposition coefficients.

[0029] The wavelet decomposition coefficients are subjected to noise component suppression. For each sub-band, a soft threshold is set using the dynamic noise basis estimate. For coefficients in high-frequency sub-bands that exceed the threshold, peak clipping is performed to suppress impulse noise. For sub-band coefficients containing narrowband interference, they are directly set to zero, thereby accurately removing various noise components and obtaining the wavelet decomposition coefficients after noise component suppression.

[0030] Based on the wavelet decomposition coefficients after noise component suppression, an inverse wavelet transform is performed. The wavelet decomposition coefficients after noise component suppression are then reconstructed by a corresponding lifting wavelet inverse transform. The reconstructed time-domain waveform significantly suppresses noise while retaining the real transient information of electricity consumption, and is output as the preprocessed pure electricity consumption information.

[0031] S202: Based on the preprocessed electricity consumption information, perform extreme point detection to obtain an extreme point sequence; based on the preprocessed electricity consumption information, perform zero-crossing point detection to obtain a zero-crossing point sequence. In the specific implementation of this invention, extreme point detection is performed based on the preprocessed power consumption information, and zero-crossing point detection is performed based on the preprocessed power consumption information. That is, the preprocessed power consumption information is used as the analysis object, and the first-order differential zero-crossing method is used to detect all local maxima and local minima in the waveform of the preprocessed power consumption information, thus obtaining the extreme point sequence. The precise time and position of the voltage and current waveforms crossing zero points are analyzed, thus obtaining the zero-crossing point sequence.

[0032] S203: Construct multi-scale structural elements based on the extreme point sequence and zero-crossing point sequence, and perform morphological gradient enhancement on the preprocessed electricity consumption information based on the multi-scale structural elements to obtain morphological gradient enhancement results; In the specific implementation of this invention, multi-scale structural elements are constructed based on the extreme point sequence and the zero-crossing sequence. According to the statistical value of the time interval between extremes in the extreme point sequence, combined with the current signal period length reflected by the zero-crossing sequence, a cluster of structural elements containing different scales and shapes is generated, which is the multi-scale structural element. Morphological gradient enhancement is performed on the preprocessed power consumption information based on the multi-scale structural elements. Morphological gradient operations are performed on each piece of preprocessed power consumption information according to the multi-scale structural elements. This operation uses differential operations of dilation and erosion to highlight the instantaneous power transition edges caused by load switching, short-circuit faults, etc., transforming the originally gradual power changes into a clear pulse sequence with large amplitude, thus obtaining the morphological gradient enhancement result.

[0033] S204: Based on the morphological gradient enhancement results, extract transient event features to obtain transient event feature information; In a specific implementation of this invention, the step of extracting transient event features based on the morphological gradient enhancement results to obtain transient event feature information includes: extracting high-frequency transient features from the preprocessed electricity consumption information based on the multi-scale structuring elements to obtain high-frequency transient feature information; extracting low-frequency discontinuity features from the preprocessed electricity consumption information based on the multi-scale structuring elements to obtain low-frequency discontinuity feature information; generating a fused time-domain transient feature map based on the morphological gradient enhancement results, high-frequency transient feature information, and low-frequency discontinuity feature information, and determining the transient event region based on the time-domain transient feature map; determining the time-domain waveform segment of the transient event based on the transient event region, and extracting transient event features based on the time-domain waveform segment of the transient event to obtain transient event feature information.

[0034] Specifically, high-frequency transient features are extracted from the preprocessed power consumption information based on the multi-scale structuring elements. A white-hat transformation is then performed on the preprocessed power consumption information based on the multi-scale structuring elements, i.e., the result of performing a morphological opening operation is subtracted from the preprocessed power consumption information. This operation can completely remove the high-frequency oscillations superimposed on the power frequency fundamental wave, the high-frequency pulse trains caused by arc discharge, and the high-order harmonic peaks generated by the switching of power electronic equipment from the fundamental wave background, thereby obtaining high-frequency transient feature information.

[0035] Based on the multi-scale structuring elements, low-frequency discontinuity features are extracted from the preprocessed power consumption information. A black hat transformation is then performed on the preprocessed power consumption information according to the multi-scale structuring elements, i.e., the preprocessed power consumption information is subtracted from the morphological closing operation result. This operation can highlight negative abnormal features such as deep voltage dips, short-term power outages, and periodic low current ranges caused by certain electricity theft behaviors, thereby obtaining low-frequency discontinuity feature information.

[0036] Based on the morphological gradient enhancement results, high-frequency transient feature information, and low-frequency discontinuity feature information, a fused temporal transient feature map is generated. This involves aligning the morphological gradient enhancement results, high-frequency transient feature information, and low-frequency discontinuity feature information along a unified time axis, and then superimposing and fusing them on the same time axis to construct a multi-layered temporal transient feature map. Regions of different intensities in the map accurately indicate the occurrence time and waveform position of various transient disturbances. Based on this temporal transient feature map, transient event regions are determined. An adaptive energy threshold segmentation algorithm is applied to the temporal transient feature map to divide regions in the map where the energy is continuously higher than the threshold into independent transient event regions.

[0037] Based on the transient event region, the time-domain waveform segment of the transient event is determined. According to the start and end time boundaries of each transient event region, the corresponding waveform segment is extracted from the preprocessed power consumption information, thus obtaining the time-domain waveform segment of the transient event. Transient event feature extraction is performed based on the time-domain waveform segment of the transient event. The duration, peak amplitude, energy decay slope, dominant oscillation frequency, and phase angle of the voltage zero-crossing sequence in the extreme point sequence and zero-crossing sequence are determined according to the time-domain waveform segment. Transient event feature extraction is performed in this way to obtain transient event feature information.

[0038] Furthermore, the transient event feature extraction based on the time-domain waveform segment of the transient event to obtain transient event feature information includes: determining the duration, peak amplitude, energy decay slope, dominant oscillation frequency, and phase angle of the event start point relative to the extreme point sequence and the voltage zero-crossing sequence in the zero-crossing sequence based on the time-domain waveform segment; determining the quantization parameter information of the transient event based on the duration, peak amplitude, energy decay slope, dominant oscillation frequency, and phase angle of the event start point relative to the extreme point sequence and the voltage zero-crossing sequence in the zero-crossing sequence; performing two-dimensional grid image mapping based on the quantization parameter information to obtain a two-dimensional feature distribution map; and determining the transient event information based on the two-dimensional feature distribution map.

[0039] Specifically, based on the time-domain waveform segment, the duration, peak amplitude, energy decay slope, dominant oscillation frequency, and phase angle of the event start point relative to the voltage zero-crossing sequence in the extreme point sequence and zero-crossing sequence are determined. Using the time-domain waveform segment as a basis, a scan is performed backward from the start point of the segment to find the first sampling point whose amplitude exceeds a preset energy threshold, which is taken as the actual physical start time of the event. Similarly, a scan is performed forward from the end of the segment to find the last sampling point that falls back below the energy threshold, which is taken as the actual end time of the event. The time difference between the start and end times is the duration of the transient event. Within the duration interval, the waveform segment is fully scanned, and the maximum positive peak value and minimum negative valley value of the voltage and current waveforms are recorded. The sum of the absolute values ​​of the positive peak value and negative valley value of the voltage and current are calculated as a representation of the amplitude intensity. Simultaneously, the maximum instantaneous power within the interval is recorded as the power peak value. If the transient event exhibits oscillation decay characteristics, the local maximum sequence within the time period is used to extract each positive peak point in the oscillation waveform. The natural logarithm of the amplitude of these peak points is taken, and a linear fit is applied to the time. The slope of the fitted straight line is the energy decay slope. For transient events with obvious oscillatory components, the time interval between consecutive zero-crossings is statistically analyzed using the zero-crossing sequence within the event duration. The dominant oscillation frequency is obtained by calculating the average period of multiple complete oscillation cycles and taking its reciprocal. Using the voltage zero-crossing sequence, the previous positive voltage zero-crossing point that is closest in time to the start time of the transient event is found. The time difference between the zero-crossing point and the start time of the event is calculated, and this time difference is divided by the power frequency period and multiplied by 360 degrees to obtain the phase angle of the event start point relative to the voltage fundamental zero-crossing point.

[0040] The quantization parameters of the transient event are determined based on the time length, peak amplitude, energy decay slope, dominant oscillation frequency, and phase angle of the voltage zero-crossing sequence in the extreme point sequence and zero-crossing sequence relative to the event start point. In other words, the quantization parameters of the transient event are composed of the time length, peak amplitude, energy decay slope, dominant oscillation frequency, and phase angle of the voltage zero-crossing sequence in the extreme point sequence and zero-crossing sequence relative to the event start point.

[0041] Based on the quantization parameter information, a two-dimensional grid image mapping is performed, mapping the quantization parameter information onto a two-dimensional grid image with time and frequency as axes, according to the event occurrence time and dominant frequency. The grayscale or color intensity of each pixel in the image represents the transient disturbance energy accumulation in that time-frequency interval, forming an intuitive two-dimensional feature distribution map. Based on the two-dimensional feature distribution map, transient event information is determined, that is, the time interval and frequency range corresponding to the significantly high-energy regions in the two-dimensional feature distribution map, the center time and center frequency of the high-energy regions, the duration of the high-energy regions on the time axis and the bandwidth on the frequency axis, and the energy contrast between the high-energy regions and the surrounding areas, etc. These descriptive parameters, together with the two-dimensional image itself, constitute the transient event feature information.

[0042] S205: Determine the steady-state interval, determine the background context feature vector based on the steady-state interval and the preprocessed electricity consumption information, and determine the target electricity consumption behavior characteristics of each user node based on the transient event feature information and the background context feature vector; In the specific implementation of this invention, a steady-state interval is determined. Based on the steady-state interval and the preprocessed electricity consumption information, a background context feature vector is determined. For each user node, outside the transient event region, a steady-state interval of several power frequency cycles before and after the event is selected. From the preprocessed electricity consumption information, the effective values ​​of current, effective values ​​of voltage, active power, reactive power, total harmonic distortion rate, and the content of each major subharmonic are extracted from the steady-state interval to form the background context feature vector. Based on the transient event feature information and the background context feature vector, the target electricity consumption behavior characteristics of each user node are determined. That is, the transient event feature information and the background context feature vector are cascaded and fused at the feature level to form a unified high-dimensional feature vector.

[0043] S206: Construct a metering data collection and early warning model between user nodes based on the target electricity consumption behavior characteristics; In a specific implementation of this invention, the step of constructing a metering data collection and early warning model between user nodes based on the target electricity consumption behavior characteristics includes: acquiring power line network topology data and constructing an electrical distance matrix between nodes based on the power line network topology data; determining dynamic correlation data of carrier signals between nodes based on preprocessed electricity consumption information and constructing a dynamic graph structure based on the electrical distance matrix between nodes and the dynamic correlation data of carrier signals between nodes; determining the time-domain pattern vector of each user node based on the target electricity consumption behavior characteristics, and constructing a metering data collection and early warning model between user nodes using a spatiotemporal graph attention network based on the time-domain pattern vector and the dynamic graph structure.

[0044] Specifically, the process involves acquiring power line network topology data, importing complete topology data of the power line network from the distribution geographic information system, including the connection relationships between transformer outgoing lines, branch lines, and meter box nodes, constructing an electrical distance matrix between nodes based on the power line network topology data, and calculating the equivalent impedance parameters along the power line between any two user nodes using the known model, wire diameter, length, and material of each line segment based on the power line network topology data. The impedance values ​​of each line segment, expressed as complex impedances, are then calculated based on the equivalent impedance parameters. Finally, the reciprocal of the impedance values ​​of each line segment is taken to convert them into the corresponding branch admittance parameters. Based on this, taking the transformer outlet of the distribution area as the reference node, the user meter access point and branch node are uniformly modeled as network nodes. According to the series and parallel connection relationship of the network topology, the admittance parameters of each branch are accumulated into the self-admittance and mutual admittance of the corresponding node through the branch addition method, and the node admittance matrix of the low-voltage distribution network is assembled. The node admittance matrix is ​​inverted to obtain the node impedance matrix. The magnitude of the mutual impedance between any two nodes in the node impedance matrix is ​​taken as the equivalent electrical distance between the pair of nodes. The equivalent electrical distances of all node pairs are arranged in the order of node number to form the electrical distance matrix between nodes.

[0045] Based on preprocessed power consumption information, dynamic correlation data of inter-node carrier signals is determined. Then, based on this preprocessed power consumption information, synchronous demodulation analysis is performed on the power line carrier communication signal components carried within it. The received signal attenuation and waveform cross-correlation coefficients of each inter-node carrier signal are continuously monitored, and these two are fused to form dynamic correlation data of inter-node carrier signals reflecting channel state fluctuations. A dynamic graph structure is constructed based on the inter-node electrical distance matrix and the dynamic correlation data of inter-node carrier signals. Each user node is used as a vertex in the graph. The normalized magnitude of the equivalent impedance of each pair of nodes in the inter-node electrical distance matrix is ​​used as the edge weight benchmark. This benchmark is then corrected in real time based on the dynamic correlation data of inter-node carrier signals. When carrier attenuation increases or waveform similarity decreases, the edge weights are increased accordingly to strengthen the potential coupling of abnormal propagation, forming a graph structure with weights that dynamically change over time.

[0046] Based on the target electricity consumption behavior characteristics, the temporal pattern vector of each user node is determined. Based on the temporal pattern vector and the dynamic graph structure, a spatiotemporal graph attention network is used to construct a metering data collection and early warning model among user nodes. For each user node in the dynamic graph structure, the target electricity consumption behavior characteristics generated in each consecutive time frame are arranged in chronological order to form a node-level feature time series. A spatiotemporal graph attention network integrating temporal convolution and spatial graph attention is constructed: the front end performs sliding convolution on each node-level feature time series through multi-scale one-dimensional convolution kernels to capture short-term sudden trends, gradual drifts, and periodic patterns of electricity consumption behavior, and outputs a temporal pattern vector. The middle layer utilizes the edge weights provided by the dynamic graph structure as priors and automatically calculates the attention influence coefficient of each neighbor node on the target node through a learnable attention function. The back-end aggregates the neighbor temporal pattern vectors according to the attention coefficients, concatenates them with the target node's own temporal pattern vector, and then feeds them into a multilayer perceptron for deep spatiotemporal feature fusion. It outputs the reconstructed prediction value and reconstruction probability of the target's electricity consumption behavior characteristics at the next moment. The network is trained using a large number of node-level feature time series under historical normal working conditions. All parameters are optimized with the goal of minimizing the reconstruction error. The final solidified network is the metering data collection and early warning model between user nodes.

[0047] S207: Based on the metering data acquisition and early warning model, perform abnormal event identification to obtain abnormal event identification results, and based on the abnormal event identification results, determine the early warning level to obtain the early warning level determination result; In the specific implementation of this invention, the step of identifying abnormal events based on the metering data acquisition and early warning model, obtaining abnormal event identification results, and determining the early warning level based on the abnormal event identification results, to obtain an early warning level determination result, includes: determining a real-time reconstructed feature vector based on the metering data acquisition and early warning model and target electricity consumption behavior characteristics; performing deviation vector analysis based on the real-time reconstructed feature vector and target electricity consumption behavior characteristics to obtain a multi-dimensional deviation vector; identifying abnormal events based on the multi-dimensional deviation vector to obtain abnormal event identification results; assessing the severity of single-node anomalies based on the abnormal event identification results to obtain a single-node anomaly severity assessment result; analyzing the anomaly propagation trend based on the abnormal event identification results to obtain anomaly propagation trend analysis results; and determining the early warning level based on the single-node anomaly severity assessment result and the anomaly propagation trend analysis result to obtain an early warning level determination result.

[0048] Specifically, based on the metering data acquisition and early warning model and the target electricity consumption behavior characteristics, a real-time reconstructed feature vector is determined. The target electricity consumption behavior characteristics of each user node are input into the metering data acquisition and early warning model. Temporal patterns are extracted through a temporal convolutional layer, neighbor information is aggregated through a graph attention layer, and forward inference is performed through a multilayer perceptron. The model outputs a predictive reconstructed feature vector for the current characteristics of each node, which is the real-time reconstructed feature vector. Based on the real-time reconstructed feature vector and the target electricity consumption behavior characteristics, deviation vector analysis is performed. For each user node, the target electricity consumption behavior characteristics and the real-time reconstructed feature vector of each user node are compared and calculated dimension by dimension. The feature vector is decomposed into three subspaces, and the amplitude deviation reflecting the difference in electricity consumption amplitude, the morphological deviation characterizing the similarity of waveform contour shape, and the harmonic structure deviation describing the difference in the distribution of harmonic components are calculated respectively. The three deviation values ​​are combined to form a multi-dimensional deviation vector.

[0049] Based on the multidimensional deviation vector, abnormal event identification can be performed to determine the boundary of the dynamic confidence ellipse. Based on the boundary of the dynamic confidence ellipse and the multidimensional deviation vector, abnormal initial state analysis of user nodes can be performed. Based on the abnormal initial state marker, duration analysis can be performed. Based on the duration, abnormal event identification can be performed to obtain the abnormal event identification result.

[0050] Based on the abnormal event identification results, a single-node anomaly severity assessment is performed. The assessment comprehensively evaluates the maximum deviation peak, duration, and potential harm to power grid safety and electricity consumption order of the event, taking into account the matched anomaly type. According to a preset hierarchical standard, the severity of this single-node anomaly event is divided into three internal levels: minor, moderate, or severe, thus obtaining the single-node anomaly severity assessment result.

[0051] Based on the anomaly identification results, anomaly propagation trend analysis is performed. This involves retrieving neighboring nodes with larger edge weights from the current anomalous node within a graph structure whose weights dynamically change over time during the same period. The spatial clustering characteristics of multiple anomalous nodes and the temporal sequence of their onset times are analyzed. If multiple neighboring nodes are confirmed to have anomalies within a short time window, an anomaly propagation trend is identified, thus obtaining the anomaly propagation trend analysis result. Based on the single-node anomaly severity assessment results and the anomaly propagation trend analysis results, a warning level is determined. A comprehensive decision is made based on these results: if the severity is minor and there is no propagation trend, a blue warning is issued; if the severity is minor but there is a propagation trend, it is upgraded to a yellow warning; if the severity is moderate, it corresponds directly to an orange warning; if the severity is severe or a clear multi-node propagation trend has formed, it is determined to be the highest level, a red warning, thus obtaining the warning level determination result.

[0052] Furthermore, the step of identifying abnormal events based on the multidimensional deviation vector to obtain abnormal event identification results includes: determining the boundary of the dynamic confidence ellipse, and performing abnormal initial state analysis of the user node based on the dynamic confidence ellipse boundary and the multidimensional deviation vector to obtain an abnormal initial state marker; performing duration analysis based on the abnormal initial state marker to obtain duration information; and performing abnormal event identification based on the duration information to obtain abnormal event identification results.

[0053] Specifically, the dynamic confidence ellipse boundary is determined, and the historical distribution characteristics of the multidimensional deviation vector of each node during long-term normal operation are statistically analyzed. An adaptive dynamic confidence ellipse boundary based on Mahalanobis distance is established. Based on the dynamic confidence ellipse boundary and the multidimensional deviation vector, abnormal initial state analysis of user nodes is performed. When the position of the multidimensional deviation vector output by a node at the current moment exceeds the dynamic confidence ellipse boundary in the multidimensional space, it is determined that the node has an abnormal initial state at the current moment, that is, an abnormal initial state label is obtained.

[0054] Duration analysis is performed based on the abnormal initial state markers. According to the unified time reference established by the original power waveform frames with time tags, each node marked as an abnormal initial state is continuously tracked. The power frequency cycle is used as the smallest time unit. The cumulative number of power frequency cycles in which the abnormal initial state marker appears is counted. The start time of the first appearance of the marker and the end time of the last appearance are recorded simultaneously. The cumulative number of power frequency cycles, the start time, and the end time are encapsulated together to form structured data describing the duration of the abnormal initial state, which is the duration information.

[0055] Anomaly identification is performed based on the duration information by comparing the cumulative number of power frequency cycles in the duration information with a preset minimum duration cycle threshold. If the cumulative number of cycles reaches the threshold, and the multidimensional deviation vector of each node during the time period is always outside the boundary of the dynamic confidence ellipse, then the continuous abnormal initial state sequence is confirmed as a complete anomaly event. After confirmation, the unique identifier of the node, the start and end times in the duration information, the peak value and evolution trend of the multidimensional deviation vector of each node during the time period are extracted. The anomaly type is matched according to the dominant component of the deviation vector. The above information is integrated to obtain the anomaly event identification result.

[0056] S208: Based on the abnormal event identification results, attribution and localization are performed using waveform correlation to obtain attribution and localization information; In a specific implementation of this invention, the step of using waveform correlation to perform attribution and location based on the abnormal event identification result to obtain attribution and location information includes: analyzing abnormal waveform segments based on the abnormal event identification result to obtain target abnormal waveform segments; performing abnormal attribution based on the target abnormal waveform segments using a fault waveform library to obtain abnormal attribution information; performing waveform correlation analysis between the abnormal node and upstream and downstream nodes based on the target abnormal waveform segments to obtain waveform correlation information, and performing disturbance propagation path analysis based on the waveform correlation information to obtain disturbance propagation path analysis results; and performing attribution and location based on the abnormal attribution information and the disturbance propagation path analysis results to obtain attribution and location information.

[0057] Specifically, based on the abnormal event identification results, abnormal waveform segment analysis is performed. According to the abnormal start and end times recorded in the abnormal event identification results, several power frequency cycles are extended from the preprocessed power consumption information to both ends with the abnormal start and end times as the center, and multi-cycle continuous waveform segments containing the steady state before the abnormality, the entire abnormal process, and the recovery after the abnormality are accurately extracted, thus obtaining the target abnormal waveform segment.

[0058] Based on the target abnormal waveform segment, anomaly attribution is performed using a fault waveform library. The target abnormal waveform segment is represented by dual-domain features. On the one hand, its original voltage and current time-domain sampling sequences are retained as time-domain features. On the other hand, a Hilbert transform is performed on the segment to extract the instantaneous amplitude envelope trajectory and instantaneous frequency change trajectory, forming a transform domain feature that reflects the instantaneous dynamics of signal energy and frequency. The transform domain feature is compared one by one with each standard sample in a preset typical fault waveform library. The fault waveform library contains standard feature templates for common anomaly types such as internal short circuits in electrical appliances, line-to-ground arcs, abnormal equipment startup, and carrier communication conflicts. The dynamic time warping distance between the abnormal waveform and each template is calculated, and the fault type with the smallest distance and the highest correlation is selected to obtain the anomaly attribution information.

[0059] Based on the target abnormal waveform segment, waveform correlation analysis is performed between the abnormal node and its upstream and downstream nodes. The time-domain and transform-domain dual feature representations of the target abnormal waveform segment are extracted. According to the weight values ​​of each edge in the graph structure where the weights change dynamically over time, the upstream and downstream nodes with the closest electrical coupling to the abnormal node are determined. Waveform segments of these neighboring nodes in the same time period are extracted from the preprocessed power consumption information, and the dual feature representations are also extracted. The dynamic time warping distances between the abnormal node waveform and the upstream node waveform, and between the abnormal node waveform and the downstream node waveform are calculated respectively. The two sets of distance values, along with the corresponding node identifiers and upstream-downstream relationships, are encapsulated to obtain waveform correlation information. Based on the waveform correlation information, disturbance propagation path analysis is performed. The waveform correlation information is analyzed to extract the dynamic time warping distance values ​​between the abnormal node and its upstream and downstream nodes. By comparing the order in which the abnormal characteristic waveform appears among the abnormal, upstream, and downstream nodes, the node with the earliest appearance of the abnormal characteristic waveform and the smallest dynamic time warping distance is identified as the disturbance source node. Simultaneously, the attenuation ratio of the abnormal characteristic amplitude and the waveform distortion evolution trend are calculated node by node along the upstream and downstream directions. Combined with the direction of power energy flow reflected by the electrical distance matrix between nodes, the physical propagation direction of the disturbance signal in the power grid topology is determined whether it is attenuating and spreading outward from the source node or being conducted from the outside to the abnormal node. The disturbance source node identification, propagation direction, attenuation ratio, and distortion trend on each path segment are integrated to obtain the disturbance propagation path analysis results.

[0060] Based on the anomaly attribution information and disturbance propagation path analysis results, attribution and location are performed. Combining the anomaly attribution information and disturbance propagation path analysis results, when it can be clearly determined that the abnormal disturbance waveform first appears from a certain node or a certain branch line and propagates to other nodes with attenuation, the specific physical location of that node or that branch line is determined as the true source of the abnormal disturbance. The location result includes a precise geographical location description and the type of suspected equipment parsed from the topology data in the electrical distance matrix between nodes, thus obtaining the attribution and location information.

[0061] S209: Generate an early warning message based on the abnormal event identification result, early warning level determination result, and attribution location information, and transmit the early warning message to the operation and maintenance terminal.

[0062] In the specific implementation of this invention, an early warning message is generated based on the abnormal event identification results, early warning level determination results, and attribution location information. This message is then transmitted to the operation and maintenance terminal. The abnormal event identification results, early warning level determination results, and attribution location information are uniformly collected. Simultaneously, user profiles and geographic coordinate data of abnormal nodes are retrieved from the power distribution management system to generate a complete, all-element early warning message, including early warning time, node identifier, abnormality type, early warning level, waveform screenshot, deviation radar chart, transient feature map, abnormal propagation path topology highlighting map, and targeted on-site operation and maintenance suggestions. The early warning message is marked with the highest priority in the power line carrier communication protocol. The optimal multi-hop relay path is selected using a graph structure with dynamically changing weights over time. The message is forwarded hop-by-hop to the distribution area aggregation terminal via a node self-organizing network. The aggregation terminal then transmits the message to the operation and maintenance terminal via a 5G or 4G encrypted tunnel. Upon receiving the message, the operation and maintenance terminal automatically triggers a mandatory audio-visual alert and highlights the precise location of the abnormal source on an electronic map, completing a closed-loop early warning process from cloud analysis to on-site reception.

[0063] In this embodiment of the invention, the electricity consumption information of each user node in the power line network is preprocessed; based on the preprocessed electricity consumption information, transient event features are extracted using morphological operations; based on the transient event feature information, the target electricity consumption behavior characteristics of each user node are determined, improving the reliability of transient event feature extraction and providing more comprehensive data support for subsequent abnormal event identification and early warning. A metering data acquisition and early warning model between user nodes is constructed based on the target electricity consumption behavior characteristics; abnormal event identification is performed based on the metering data acquisition and early warning model, improving the accuracy of abnormal event identification. Early warning level is determined based on the abnormal event identification results; attribution and location are performed based on waveform correlation based on the abnormal event identification results; an early warning message is generated based on the abnormal event identification results, early warning level determination results, and attribution and location information, and transmitted to the operation and maintenance terminal. This completes the information flow from data acquisition and intelligent analysis to the operation and maintenance closed loop, providing sufficiently comprehensive early warning information, enabling operation and maintenance personnel to directly carry out disposal actions without secondary analysis, significantly shortening the abnormal response chain and improving the level of electricity management.

[0064] Example 3 Please see Figure 3 , Figure 3 This is a schematic diagram of the structural composition of the power line carrier metering data acquisition and early warning system in an embodiment of the present invention. The system includes: Information preprocessing module 31: used to acquire the power consumption information of each user node in the power line network based on the power line carrier acquisition module, preprocess the power consumption information, and obtain preprocessed power consumption information; Feature extraction module 32: used to extract transient event features based on preprocessed electricity consumption information using morphological operations, obtain transient event feature information, and determine the target electricity consumption behavior features of each user node based on the transient event feature information; Model building module 33: used to build a metering data collection and early warning model between user nodes based on the target electricity consumption behavior characteristics; Early warning determination module 34: used to identify abnormal events based on the metering data acquisition early warning model, obtain abnormal event identification results, and determine the early warning level based on the abnormal event identification results, and obtain early warning level determination results; Attribution localization module 35: used to perform attribution localization based on waveform correlation according to the abnormal event identification result, and obtain attribution localization information; Early warning message transmission module 36: used to generate early warning messages based on the abnormal event identification results, early warning level determination results and attribution location information, and transmit the early warning messages to the operation and maintenance terminal.

[0065] In the specific implementation of this invention, the specific implementation methods of the system items can be referred to the implementation methods of the above-mentioned method items, and will not be repeated here.

[0066] In this embodiment of the invention, the electricity consumption information of each user node in the power line network is preprocessed; based on the preprocessed electricity consumption information, transient event features are extracted using morphological operations; based on the transient event feature information, the target electricity consumption behavior characteristics of each user node are determined, improving the reliability of transient event feature extraction and providing more comprehensive data support for subsequent abnormal event identification and early warning. A metering data acquisition and early warning model between user nodes is constructed based on the target electricity consumption behavior characteristics; abnormal event identification is performed based on the metering data acquisition and early warning model, improving the accuracy of abnormal event identification. Early warning level is determined based on the abnormal event identification results; attribution and location are performed based on waveform correlation based on the abnormal event identification results; an early warning message is generated based on the abnormal event identification results, early warning level determination results, and attribution and location information, and transmitted to the operation and maintenance terminal. This completes the information flow from data acquisition and intelligent analysis to the operation and maintenance closed loop, providing sufficiently comprehensive early warning information, enabling operation and maintenance personnel to directly carry out disposal actions without secondary analysis, significantly shortening the abnormal response chain and improving the level of electricity management.

[0067] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, which may include: read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, etc.

[0068] Furthermore, the above provides a detailed description of the metering data acquisition and early warning method and system for power line carrier provided by the embodiments of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for collecting and issuing early warning data for power line carrier communication, characterized in that, The method includes: The power consumption information of each user node in the power line network is obtained based on the power line carrier acquisition module, and the power consumption information is preprocessed to obtain the preprocessed power consumption information. Based on the preprocessed electricity consumption information, transient event features are extracted using morphological operations to obtain transient event feature information, and the target electricity consumption behavior features of each user node are determined based on the transient event feature information. A metering data collection and early warning model between user nodes is constructed based on the target electricity consumption behavior characteristics. Based on the aforementioned measurement data acquisition and early warning model, abnormal events are identified to obtain abnormal event identification results. Based on the abnormal event identification results, an early warning level is determined to obtain an early warning level determination result. Based on the abnormal event identification results, attribution and localization are performed using waveform correlation to obtain attribution and localization information; Based on the abnormal event identification results, early warning level determination results, and attribution location information, an early warning message is generated and transmitted to the operation and maintenance terminal.

2. The method for metering data acquisition and early warning based on power line carrier according to claim 1, characterized in that, The step of preprocessing the electricity consumption information to obtain preprocessed electricity consumption information includes: The power consumption information is decomposed into multiple channels to obtain wavelet decomposition coefficients, and the wavelet decomposition coefficients are then subjected to noise component suppression to obtain noise component-suppressed wavelet decomposition coefficients. Inverse wavelet transform is performed on the wavelet decomposition coefficients after noise component suppression to obtain preprocessed electricity consumption information.

3. The method for collecting and issuing early warning data for power line carrier communication according to claim 1, characterized in that, The process of extracting transient event features using morphological operations based on preprocessed electricity consumption information to obtain transient event feature information, and determining the target electricity consumption behavior features of each user node based on the transient event feature information, includes: Extreme point detection is performed based on the preprocessed electricity consumption information to obtain an extreme point sequence, and zero-crossing point detection is performed based on the preprocessed electricity consumption information to obtain a zero-crossing point sequence. Based on the extreme point sequence and the zero-crossing point sequence, a multi-scale structural element is constructed, and the preprocessed electricity consumption information is subjected to morphological gradient enhancement based on the multi-scale structural element to obtain the morphological gradient enhancement result. Based on the morphological gradient enhancement results, transient event features are extracted to obtain transient event feature information; A steady-state interval is determined, and a background context feature vector is determined based on the steady-state interval and the preprocessed electricity consumption information. The target electricity consumption behavior characteristics of each user node are then determined based on the transient event feature information and the background context feature vector.

4. The method for collecting and issuing early warning data for power line carrier communication according to claim 3, characterized in that, The step of extracting transient event features based on the morphological gradient enhancement results to obtain transient event feature information includes: Based on the multi-scale structural elements, high-frequency transient features are extracted from the preprocessed electricity consumption information to obtain high-frequency transient feature information; Based on the multi-scale structural elements, low-frequency discontinuity features are extracted from the preprocessed electricity consumption information to obtain low-frequency discontinuity feature information. Based on the morphological gradient enhancement results, high-frequency transient feature information and low-frequency discontinuity feature information, a fused temporal transient feature map is generated, and the transient event region is determined based on the temporal transient feature map; Based on the transient event region, the time-domain waveform segment of the transient event is determined, and the transient event features are extracted based on the time-domain waveform segment of the transient event to obtain transient event feature information.

5. The method for collecting and issuing early warning data for power line carrier communication according to claim 4, characterized in that, The transient event feature extraction is performed on the time-domain waveform segment based on the transient event to obtain transient event feature information, including: Based on the time-domain waveform segment, determine the duration, peak amplitude, energy decay slope, dominant oscillation frequency, and phase angle of the voltage zero-crossing sequence relative to the extreme point sequence and the zero-crossing sequence. Based on the duration, peak amplitude, energy decay slope, dominant oscillation frequency, and phase angle of the voltage zero-crossing sequence relative to the extreme point sequence and the zero-crossing sequence, determine the quantization parameter information of the transient event. Based on the quantization parameter information, a two-dimensional grid image is mapped to obtain a two-dimensional feature distribution map; Transient event information is determined based on the two-dimensional feature distribution map.

6. The method for metering data acquisition and early warning based on power line carrier according to claim 1, characterized in that, The method for constructing a metering data collection and early warning model between user nodes based on the target electricity consumption behavior characteristics includes: Acquire power line network topology data and construct an electrical distance matrix between nodes based on the power line network topology data; Based on the preprocessed power consumption information, the dynamic correlation data of the carrier signals between nodes is determined, and a dynamic graph structure is constructed based on the electrical distance matrix between nodes and the dynamic correlation data of the carrier signals between nodes. Based on the target electricity consumption behavior characteristics, the time-domain pattern vector of each user node is determined, and based on the time-domain pattern vector and dynamic graph structure, a spatiotemporal graph attention network is used to construct a metering data collection and early warning model among user nodes.

7. The method for metering data acquisition and early warning based on power line carrier according to claim 1, characterized in that, The process of identifying abnormal events based on the metering data acquisition and early warning model, obtaining abnormal event identification results, and determining the early warning level based on the abnormal event identification results, includes: Based on the metering data acquisition and early warning model and the target electricity consumption behavior characteristics, a real-time reconstructed feature vector is determined. Based on the real-time reconstructed feature vector and the target electricity consumption behavior characteristics, a deviation vector analysis is performed to obtain a multi-dimensional deviation vector. Anomaly identification results are obtained by performing anomaly identification based on the multidimensional deviation vector. Based on the abnormal event identification results, the severity of single-node anomalies is assessed to obtain the severity assessment results of single-node anomalies. Based on the abnormal event identification results, an abnormal propagation trend analysis is performed to obtain the abnormal propagation trend analysis results. Based on the single-node abnormality severity assessment results and the abnormal propagation trend analysis results, the warning level is determined to obtain the warning level determination results.

8. The method for metering data acquisition and early warning of power line carrier according to claim 7, characterized in that, The process of identifying abnormal events based on the multidimensional deviation vector to obtain abnormal event identification results includes: Determine the boundary of the dynamic confidence ellipse, and perform anomaly initial state analysis of user nodes based on the boundary of the dynamic confidence ellipse and the multidimensional deviation vector to obtain anomaly initial state labels; Duration analysis is performed based on the abnormal initial state markers to obtain duration information; Anomaly event identification is performed based on the duration information to obtain anomaly event identification results.

9. The method for metering data acquisition and early warning based on power line carrier according to claim 1, characterized in that, The attribution and localization based on waveform correlation, obtained by utilizing the abnormal event identification results, includes: Based on the abnormal event identification results, abnormal waveform segment analysis is performed to obtain the target abnormal waveform segment; Based on the target abnormal waveform segment, anomaly attribution is performed using a fault waveform library to obtain anomaly attribution information. Based on the target abnormal waveform segment, waveform correlation analysis is performed between the abnormal node and the upstream and downstream nodes to obtain waveform correlation information. Based on the waveform correlation information, disturbance propagation path analysis is performed to obtain disturbance propagation path analysis results. Attribution location is performed based on the anomaly attribution information and disturbance propagation path analysis results to obtain attribution location information.

10. A metering data acquisition and early warning system based on power line carrier wave, characterized in that, The system includes: Information preprocessing module: used to acquire electricity consumption information of each user node in the power line network based on the power line carrier acquisition module, preprocess the electricity consumption information, and obtain preprocessed electricity consumption information; Feature extraction module: used to extract transient event features based on preprocessed electricity consumption information using morphological operations, obtain transient event feature information, and determine the target electricity consumption behavior features of each user node based on the transient event feature information; Model building module: used to build a metering data collection and early warning model between user nodes based on the target electricity consumption behavior characteristics; Early warning determination module: used to identify abnormal events based on the metering data acquisition early warning model, obtain abnormal event identification results, and determine the early warning level based on the abnormal event identification results, and obtain the early warning level determination result; Attribution localization module: used to perform attribution localization based on waveform correlation according to the abnormal event identification results, and obtain attribution localization information; Early warning message transmission module: used to generate early warning messages based on the abnormal event identification results, early warning level determination results and attribution location information, and transmit the early warning messages to the operation and maintenance terminal.

Images