A method, device and medium for dynamic sensing of the entire power distribution network

Abstract

This invention discloses a method, device, and medium for dynamic sensing across the entire distribution network, belonging to the field of power system operation and control technology. The method includes constructing an active sensing layer to collect transient waveform data; fusing the transient waveform data with distribution network operation data to generate a global data correlation cloud map; performing reverse backtracking based on the global data correlation cloud map when a tripping event occurs to form an event chain; constructing an operational stress accumulation model based on the event chain for latent defect early warning; sensing the defect source based on the latent defect early warning results; and generating a dynamic sensing map of the distribution network by integrating the event chain, early warning results, and defect source. This invention achieves end-to-end connectivity and closed-loop optimization from bottom-level data sensing to top-level decision support.

Description

Technical Field

[0001] This invention relates to the field of power system operation and control technology, specifically to a method, device, and medium for dynamic sensing across the entire distribution network. Background Technology

[0002] Currently, this field mainly relies on Supervisory Control and Data Acquisition (SCADA) systems, fault indicators, and the gradually expanding synchronous phasor measurement technology. SCADA systems enable macroscopic monitoring of steady-state quantities such as voltage and current, but their long sampling periods make it difficult to capture transient processes. Fault indicators can provide segment location after a fault occurs, but their trigger thresholds are high and they lack precise synchronization and waveform recording capabilities. While synchronous phasor measurement technology offers excellent accuracy, its high cost and stringent communication synchronization requirements make large-scale, dense deployment difficult in complex distribution networks with numerous nodes. Overall, existing technologies focus more on wide-area measurement of transmission networks or post-fault isolation in distribution networks. For capturing early, weak disturbances and transient characteristics in distribution networks, and for deep fusion of multi-source data, a dynamic sensing solution that balances economy, efficiency, and comprehensiveness is still lacking.

[0003] However, current technical architectures have significant limitations in achieving comprehensive and in-depth perception of distribution networks, which is precisely the key problem this invention aims to solve. For example, the settings of existing devices are usually designed to isolate existing faults and are insensitive to early, minor electrical changes below the protection settings, creating a perception blind spot. Simultaneously, low sampling rates lead to the loss of key transient characteristic information such as harmonics and oscillations, rendering subsequent analysis unfounded. Furthermore, various types of data are often processed in isolation, lacking a unified spatiotemporal fusion framework; traditional correlation analysis is mostly based on static physical topology, failing to incorporate dynamic factors such as real-time power flow direction and switch status, resulting in judgments of event propagation paths and impact ranges that do not match the actual electrical coupling strength. In addition, existing operation and maintenance models rely heavily on periodic inspections and post-incident repairs, lacking predictive models for hidden equipment defects based on the quantitative accumulation of historical stress events, making it impossible to issue early warnings before abnormal external parameters of equipment. Even when anomalies are detected, it is difficult to quickly and accurately locate the electrical source of the defect from wide-area transient data, making preventative maintenance measures lack specificity. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, this invention aims to overcome the limitations of existing technologies, such as insensitivity to early weak disturbances, loss of transient features, data isolation, and static analysis.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a method for dynamic sensing of the entire distribution network, comprising, An active sensing layer is constructed, and transient waveform data of the distribution network is collected based on the active sensing layer; the transient waveform data is fused with the distribution network operation data to generate a global data correlation cloud map; when a tripping event occurs, the global data correlation cloud map is used to perform reverse backtracking to generate an event chain; based on the event chain, an operational stress accumulation model is constructed to perform latent defect early warning; based on the latent defect early warning results, the defect source is determined; and based on the event chain, early warning results, and defect source, a dynamic sensing map of the distribution network is generated.

[0007] As a preferred embodiment of the dynamic sensing method for the entire distribution network described in this invention, the construction of the active sensing layer includes: Based on the topology of the distribution network, monitoring terminals are deployed at key nodes of the distribution network, and micro-disturbance thresholds and lists of electrically related neighbor nodes are set. When any monitoring terminal detects that the electrical quantity exceeds the micro-disturbance threshold, it sends a synchronization trigger command packet. After receiving the synchronization trigger command packet, it performs command validity verification and parsing to obtain the transient waveform packet. The transient waveform packet is time-aligned and encapsulated to generate a wide-area correlated transient data group.

[0008] As a preferred embodiment of the power distribution network dynamic sensing method described in this invention, the fusion includes: Receive the wide-area correlated transient data group to obtain distribution network operation data; The dynamic topological time-varying correlation degree between any two nodes is calculated by fusing the basic electrical coupling factor, power flow influence factor, and applied switch state attenuation coefficient. The calculated dynamic topological time-varying correlation degree is used for visualization mapping to construct a global data correlation cloud map.

[0009] As a preferred embodiment of the dynamic sensing method for the entire distribution network described in this invention, the reverse backtracking includes: When a tripping event occurs, determine the backtracking root node corresponding to the tripping point, and set the backtracking time window with the tripping time as the endpoint. From the global data association cloud map, extract all historical data events within the backtracking time window that have an association strength exceeding a preset threshold with the backtracking root node, and form association events; Starting from the backtracking root node, the process of event impact propagation is simulated to construct a potential event chain; For each event included in the potential event chain, calculate the contribution weight of the current event to the tripping event and perform normalization processing to obtain the event chain.

[0010] As a preferred embodiment of the dynamic sensing method for the entire distribution network described in this invention, the latent defect early warning includes: Based on the event chain, the equipment identifiers of the involved equipment are marked, and an operational stress accumulation file is established for all marked equipment. The stress increment is set and merged into a comprehensive cumulative stress value that represents the overall cumulative damage level of the equipment, and the normal aging process of all equipment is depicted as an aging curve; A warning threshold is defined from the aging curve. The comprehensive cumulative stress value is compared with the warning threshold. When the comprehensive cumulative stress value is greater than or equal to the warning threshold, a latent defect warning is generated for the current equipment.

[0011] As a preferred embodiment of the power distribution network dynamic sensing method described in this invention, the step of determining the defect source includes: When a latent defect warning is received, the wide-area correlated transient data group prior to the warning time is acquired; The current and voltage waveforms of the corresponding nodes in the wide-area correlated transient data group are preprocessed; The preprocessed waveform is subjected to a parallel three-layer analysis process, and harmonic features, pulse features and oscillation mode features are extracted respectively; By comparing the energy and amplitude of different characteristics, the dominant characteristic type of this disturbance is determined, and the source of each dominant characteristic type is located.

[0012] As a preferred embodiment of the power distribution network dynamic sensing method described in this invention, the source location includes: For the dominant feature type, a distributed parameter network is constructed, and the nodes in the network are successively assumed to be disturbance sources to simulate the propagation process of feature components; Calculate the error between theoretical propagation results and actual observed characteristics under different disturbance sources; The node with the smallest error is identified as the source of the early warning.

[0013] As a preferred embodiment of the power distribution network dynamic sensing method of the present invention, the power distribution network dynamic sensing map includes: For any device node, the event chain, the latent defect warning result, and the warning source are received; Dynamically perceived risk values ​​are calculated by analyzing the risk components of historical event chains, current early warning risk components, and risk components of disturbance sources. The calculated dynamic sensing risk values ​​are mapped onto a color spectrum to obtain the dynamic sensing map of the power distribution network.

[0014] The present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the aforementioned dynamic sensing method for the entire distribution network.

[0015] The present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that the computer program, when executed by a processor, implements the steps of the aforementioned dynamic sensing method for the entire distribution network.

[0016] The beneficial effects of the present invention are as follows: By deploying terminals with high-precision synchronization capabilities and setting a micro-disturbance threshold lower than the protection setpoint, the present invention achieves sensitive capture and wide-area collaborative acquisition of weak electrical signals of early defects, providing a high-quality, spatiotemporally synchronized raw transient data foundation for the entire system.

[0017] It also proposes a dynamic topology time-varying correlation degree, which integrates real-time electrical coupling, power flow direction and switch status, and uses this to drive the fusion of multi-source data to generate a global correlation cloud map, realizing the visualization of the power grid situation from static topology to dynamic impact propagation.

[0018] Furthermore, when a trip occurs, the event chain leading to the trip is constructed and quantitatively evaluated by tracing back the correlation of historical moments, revealing the spatiotemporal causal sequence of the fault.

[0019] Furthermore, a stress accumulation model for nonlinear operation of the equipment is established based on the event chain. This model takes into account the damage acceleration effect and realizes early warning of latent defects. For the early warning equipment, multi-feature hierarchical analysis and network inversion calculation are performed on the wide-area transient waveform to accurately locate the physical source of the defect.

[0020] Furthermore, by integrating historical event chains, current early warnings, and source risks, a dynamic perception map is generated that intuitively displays the global risk level and priority. Overall, this invention elevates traditional post-event fault analysis of distribution networks to a new paradigm of proactive defense and intelligent operation and maintenance, encompassing pre-event latent defect early warning, in-event event chain tracing and source location, and global risk visualization. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0022] Figure 1 The above is a flowchart of a dynamic sensing method for the entire distribution network provided in one embodiment of the present invention.

[0023] Figure 2 This is a flowchart illustrating the data acquisition process of the active sensing layer in a dynamic sensing method for the entire power distribution network, as provided in an embodiment of the present invention.

[0024] Figure 3 This is a flowchart illustrating a latent defect early warning method for a power distribution network dynamic sensing method according to an embodiment of the present invention. Detailed Implementation

[0025] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0026] Example 1, referring to Figure 1 This is one embodiment of the present invention, which provides a method for dynamic sensing of the entire distribution network, including: S100. Construct an active sensing layer to collect transient waveform data; S200: Integrate transient waveform data with distribution network operation data to generate a global data association cloud map; S300: When a tripping event occurs, the event chain is obtained by combining the global data association cloud map for reverse tracing. S400, based on the event chain, constructs a running stress accumulation model to provide early warning of hidden defects; S500: Based on the early warning results of hidden defects, perceive the source of defects; S600, integrated event chain, early warning results and defect sources generate a dynamic perception map of the power distribution network; It should be noted that existing technologies still have limitations, such as: the high settings of existing protection or monitoring devices make it difficult to sensitively capture early, weak electrical changes below the protection settings; the lack of deep fusion of multi-source data such as SCADA and fault recording based on a unified spatiotemporal framework; difficulty in automatically and quantitatively tracing back and constructing a series of precursory events leading to the fault after a tripping event, making it impossible to clearly reveal the fault evolution process; the lack of a damage model based on the quantitative accumulation of multi-source stress throughout the equipment's entire life cycle, making it impossible to predict and warn of the development of hidden defects in the equipment before external parameters become significantly abnormal; and even if an anomaly is detected or a warning is issued, it is difficult to locate the physical source of the abnormal electrical characteristics from the wide-area, multi-point transient waveforms.

[0027] Therefore, to address the aforementioned issues, the following steps (S100-S600) are employed: starting with wide-area synchronous high-density transient data acquisition; forming a panoramic situation cloud map through multi-source data fusion driven by dynamic correlation; performing reverse causal chain analysis of tripping events based on this cloud map; using the event chain to drive equipment stress accumulation modeling and latent defect early warning; accurately locating the defect source triggered by multi-feature inversion in early warning; and finally aggregating all chain information to generate a dynamic comprehensive risk map serving operation and maintenance decisions. This achieves seamless integration and closed-loop optimization from bottom-level data perception to top-level decision support.

[0028] Example 2, refer to Figures 1-3 This is one embodiment of the present invention, which provides a method for dynamic sensing of the entire distribution network, including: In this embodiment of the invention, an active sensing layer is constructed in S100 to collect transient waveform data, including the following steps S101-S105: S101. Deploy monitoring terminals with high-speed sampling and high-precision clock synchronization functions. Specific locations include, but are not limited to, key electrical locations such as 10kV outgoing switch cabinets in substations, main line nodes, important branch nodes, key power user access points, and distributed power grid connection points in the distribution network. Each terminal undergoes parameter pre-configuration before being put into operation: S1011. Set the current micro-perturbation threshold and voltage micro-perturbation threshold for each terminal. Among them, the current micro-disturbance threshold is set to five to ten percent of the rated current of the line where the current terminal is located; the voltage micro-disturbance threshold is set to one to three percent of the rated voltage of the system where the current terminal is located. It should be noted that these thresholds are lower than the instantaneous overcurrent protection settings and overvoltage protection settings of the line relay protection device, in order to detect subtle electrical changes caused by early defects.

[0029] S1012. Configure an electrical associated neighbor node list for each terminal according to the topology of the distribution network. The list of electrically associated neighbor nodes is generated by searching along the power supply side and the load side from the node where the current terminal is located, and including all other smart monitoring terminal nodes whose electrical distance is less than a preset impedance threshold in the list of electrically associated neighbor nodes. The list of electrically associated neighbor nodes defines the scope of which neighboring terminals need to be notified for collaborative monitoring when a local disturbance occurs. It should be noted that the electrical distance is calculated by the sum of the line impedance and the line resistance, and the preset impedance threshold is set to 0.5 to 2.0 ohms based on the line voltage level and typical line parameters.

[0030] S1013. The time length of the preset high-density acquisition window is set to 5 to 10 power frequency cycles, i.e. 100 milliseconds to 200 milliseconds; The sampling rate is preset to at least 1 million sampling points per second within the current time window.

[0031] S102. Each intelligent monitoring terminal continuously monitors the three-phase voltage and three-phase current of the connected line in real time, and calculates the instantaneous and effective values ​​of each phase current and voltage. Specifically, the terminal continuously compares the absolute value of the instantaneous value of any phase current with the current micro-perturbation threshold. Simultaneously, the absolute value of the difference between the instantaneous value of any phase voltage and the rated voltage is continuously compared with the voltage micro-perturbation threshold. When the absolute value of the instantaneous current of any phase is greater than or equal to the current micro-disturbance threshold, or the absolute value of the voltage deviation of any phase is greater than or equal to the voltage micro-disturbance threshold, a local micro-disturbance event is determined to have occurred.

[0032] In response to the determination that a local micro-disturbance event has occurred, the timestamp of the current disturbance trigger is immediately recorded. At the same time, a preliminary analysis is performed on the electrical quantity that caused the trigger to determine whether it is a current-dominated or voltage-dominated disturbance, and a preliminary identifier is generated.

[0033] S103. After completing the event determination and time recording, create a synchronization trigger instruction package, which includes: the unique identifier ID of the triggering terminal, the disturbance trigger timestamp, and the initial identifier that caused the trigger. The generated synchronization trigger command packet is sent simultaneously in two directions via the data communication interface: The first direction is to send a message to all terminals in the list of electrically associated neighbor nodes, via multicast or fast unicast, on a dedicated low-latency command channel or a shared data network, to notify neighboring terminals to prepare for collaborative data collection.

[0034] The second direction is to unicast and upload to the regional data aggregation server to report the occurrence of the triggered event to the next higher level system.

[0035] S104. All online monitoring terminals, corresponding to their internal communication modules, continuously listen for synchronization trigger command packets in the network. When a terminal receives a synchronization trigger command packet, it verifies and parses the command validity, parses the command packet, and performs high-density data collection. The start time of data acquisition is calculated and calibrated based on the received disturbance trigger timestamp and the current high-precision clock time. Due to network latency in command transmission, the start time of data acquisition is set to the disturbance trigger timestamp. All terminals open the acquisition window after a unified logical start time. Within the current window, the instantaneous waveforms of three-phase voltage and three-phase current are recorded synchronously at a sampling rate of at least 1 million sampling points per second.

[0036] After the acquisition window ends, each terminal participating in this collaborative acquisition will package the acquired raw instantaneous waveform, terminal identifier, and trigger timestamp corresponding to this acquisition into a transient waveform package.

[0037] S105. After receiving the synchronization trigger instruction packet, start the data collection waiting period. During the waiting period, continuously receive transient waveform packets uploaded from various terminals in the distribution network. It should be noted that the waiting period needs to be greater than the sum of the acquisition window duration and the maximum network transmission delay. The maximum network transmission delay is the difference between the disturbance trigger timestamp and the current high-precision clock time. Using the unique identifier in each received synchronization trigger instruction packet as the index key, all transient waveform packets are aggregated together.

[0038] All the collected transient waveform packets are time-aligned and encapsulated, and the waveform data from different spatial nodes are aligned on the time axis with the same start time as the time reference to generate a wide-area associated transient data group.

[0039] It should be noted that, directly addressing the issues of perception blind spots and transient feature loss, by setting a micro-perturbation threshold below the protection setpoint and synchronous sampling up to 1MHz, sensitive capture of early defect signals and complete recording of transient waveforms were achieved, laying a high-quality data foundation for all subsequent advanced analyses.

[0040] In this embodiment of the invention, step S200 involves fusing the transient waveform data with the power distribution network operation data to generate a global data association cloud map, including the following steps S201-S204: S201. Configure a unified data access bus to continuously receive data streams, specifically including: Receive wide-area correlated transient data groups, which correspond to transient event data; The system receives and encapsulates the operation data of the distribution network, including: for analog values ​​such as line current and node voltage, sampling is performed at a set period; when any analog value changes by more than 0.5% of its rated value relative to the previous sampling point, a steady-state telemetry event data is generated; when new defect records, hidden danger ledgers, or maintenance work orders are entered, operation and maintenance history event data is generated. All data is arranged in chronological order of occurrence and placed in a cache pool to await data fusion.

[0041] S202. In order to calculate the dynamic correlation degree, it is necessary to grasp the structure and operation status of the distribution network in real time; Using the latest electrical connections and real-time switch status of the distribution network as input, the depth-first search algorithm in graph theory is used to calculate and update the connectivity of the current distribution network and generate the topological adjacency matrix at the current moment.

[0042] The voltage phasors of all nodes in the entire network, including magnitude and phase angle, and the active and reactive power flow data of all branches are obtained. Using the above data, the power flow distribution of the entire network is deduced through linearized state estimation or power flow calculation based on Kirchhoff's laws, and the power flow direction of each branch from the head end to the tail end is obtained.

[0043] S203. Based on the data in the buffer pool and the structure and operating status of the distribution network, calculate the dynamic correlation value for any two nodes i and j in the distribution network at any time t. In this invention, the distance is not based on a fixed physical distance, but rather on a combination of real-time electrical coupling strength and operating status. For node pair i and j to be calculated, read the current topology adjacency matrix of the distribution network, the current branch power flow direction data, the fixed resistance R parameter and reactance X parameter of the distribution network; S2031. Calculate the basic electrical coupling factor, specifically: Based on the current topology, find all valid electrical paths from node i to node j, wherein a valid electrical path is one in which all switches on the current path are in the closed state; Calculate the path coupling impedance for each valid path, which is the sum of the absolute values ​​of the impedances of all series branches on the current path. Furthermore, the reciprocal of the minimum path coupling impedance among all valid paths is taken as the basic electrical coupling factor; that is, the basic electrical coupling factor is equal to 1 divided by the minimum path coupling impedance; if there is no valid path, that is, all paths between i and j have open switches, then the basic electrical coupling factor is 0. S2032, Calculate the tidal current impact factor, specifically: Iterate through all branches on the path with the minimum impedance from node i to node j; For each branch on the path, determine whether its power flow direction is consistent with the path direction from i to j. If the path directions are consistent, assign a positive weight w to the current branch from i to j; if the path directions are opposite, assign a negative weight w; if the absolute value of the power flow is less than 5% of the rated value, assign a neutral weight w=1. Multiply the weights w of all branches on the path together to obtain the power flow direction product factor P of the entire path. The tidal current influence factor is defined as the product factor of the tidal current direction plus 1, divided by 2, mapping P to between 0 and 1. If the product factor of the tidal current direction is positive and greater than 1, the tidal current influence factor is greater than 0.5, indicating a strong correlation with the tidal current direction. If the product factor of the tidal current direction is negative and less than 1, the tidal current influence factor is less than 0.5, indicating a weak correlation with the counter-tidal current direction.

[0044] S2033, Application switch state attenuation coefficient, specifically: The switch state attenuation coefficient is defined as follows: if there is at least one fully closed valid path between nodes i and j, then the switch state attenuation coefficient = 1; if there is no fully closed path, but there is a path containing N open switches (assuming that they may be connected through tie switches), then the switch state attenuation coefficient is calculated by multiplying the attenuation coefficient by the number of open paths, where the attenuation coefficient is a value between 0 and 1. For example, if the attenuation coefficient is 0.3, this means that for every additional disconnected switch on the path, the correlation decreases exponentially by 0.3. If there is no connection at all, the basic electrical coupling factor is already 0, and the switch state attenuation coefficient no longer has any effect.

[0045] S2034, Synthetic Dynamic Topological Time-Varying Correlation Degree : In this invention, it is necessary to overcome the shortcomings of simple weighted averaging or linear combination in accurately describing the nonlinear interaction between electrical coupling, power flow direction, and switching states. Therefore, the above three factors are organically combined: , in, The basic electrical coupling factor; This is the maximum value of the basic electrical coupling factor calculated for all node pairs under the same topology. In order to Perform normalization to ensure its range falls between 0 and 1, and Recalculation is required after each topology change; The trend influence factor typically ranges from 0.45 to 1.1. β is the attenuation coefficient of the switching state, ranging from 0 to 1; β and γ are the adjustable exponents of different factors. Through extensive simulation and historical data analysis, β=0.7 and γ=0.5 are set.

[0046] S204. Using the calculated time-varying correlation degree All data will be used to construct a global data association cloud map.

[0047] S2041. Whenever new event data is retrieved from the cache pool for fusion, the original occurrence node is set to m, and the original occurrence time is... It assigns an initial influence of 1.

[0048] S2042. Calculate the time-varying correlation degree between the current occurrence node and all other nodes in the entire network based on the distribution network status at the current occurrence time.

[0049] S2043. The influence of the current event will spread from node m to the entire network. For any remaining node, the influence received from the current event is the product of the initial influence and the current time-varying correlation. It should be noted that the nodes with tighter electrical coupling, smoother power flow direction, and more direct switch connection are more affected by the event.

[0050] S2044. Define the decay function. For an event that occurred at a historical moment, its residual influence at node m at the current evaluation moment is: calculated by multiplying the decay coefficient by the time difference of the event occurrence using the natural exponential function. The time difference of the event occurrence is the current evaluation moment minus the initial occurrence node. At the current moment, the residual influence of the event spreading from node m to the remaining nodes is the product of the decay coefficient calculated by the natural exponential function, the time difference of the event occurrence, and the corresponding time-varying correlation degree.

[0051] S2045. For the current moment, iterate through all historical event objects and calculate the remaining influence of each event on every node in the entire network according to the calculation method of the remaining influence of node m in S2044. On the graphical interface, the remaining influence value of each node is mapped to color depth and graphic size; the higher the remaining influence of a node, the more eye-catching and larger the red marker will be displayed on the global data association cloud map.

[0052] S2046. When a new event occurs, immediately repeat S2041 to S2045. Its initial influence is added to the matrix, and the global data association cloud map is updated in real time. As time goes by, the residual influence of old events gradually diminishes, their contribution to the global data association cloud weakens, and their color fades. At the same time, the time-varying association degree is recalculated. When the time-varying association degree changes, the influence of the diffusion of all historical events based on the new association degree is recalculated, and the global data association cloud is updated.

[0053] It should be noted that the problem of data silos and static analysis is solved by realizing the spatiotemporal integration of multi-source data through a dynamic correlation model; a dynamic topology time-varying correlation is proposed, which integrates real-time electrical coupling, power flow direction and switch status to generate a panoramic visualization cloud map that can accurately reflect the dynamic propagation and impact range of events, so that situational awareness can leap from static topology to dynamic impact assessment.

[0054] In an embodiment of the present invention, when a tripping event occurs in step S300, the event chain is formed by reverse tracing using the global data association cloud map, including the following steps S301-S305: S301. In this invention, the analysis is performed under the background of distribution network tripping. When the distribution network protection device operates and causes the switch to trip, a protection tripping signal is generated. The protection tripping signal is used as a trigger input. The tripping signal is parsed to determine the physical switchgear that tripped. The first node of the line protected by the current physical switchgear is marked as the backtracking root node of this analysis.

[0055] S302. Based on the voltage level and line type of the tripped line, as well as historical tripping analysis and statistics, dynamically configure the key parameters for this retrospective. It should be noted that the key parameters include the retrospective time window and the minimum correlation strength threshold: S3021. Set a retrospective time window, which is a time interval extending in the historical direction with the tripping time as the endpoint. It should be noted that the length of the retrospective time window is not fixed, but is dynamically adjusted according to the importance of the line, the nature of the fault, and the weather conditions; the basic window is 24 hours before the tripping. In an optional embodiment, the backtracking time window may be extended to 72 hours if it involves a cable line or important users; In another alternative embodiment, the backtracking time window can also be shortened to 2 hours before the trip if the fault is transient due to a lightning strike.

[0056] S3022. Set a minimum association strength threshold, wherein the minimum association strength threshold is set to a value between 0.1 and 0.3, and the initial default value is 0.15.

[0057] S3023. After the parameters are determined, a query request is initiated to the global data association cloud map. The request content is: to obtain the complete record of all data event points whose real-time association strength with the backtracking root node is greater than or equal to the minimum association strength threshold within the time interval [backtracking time window, tripping time]. These data event points are the various events on each node of the global data association cloud map in S204 that have undergone diffusion attenuation calculation, and are put into the medium association event pool. S303. Starting from the backtracking root node, simulate the process of event impact propagation; Divide the backtracking time window into two or more consecutive time slices, and scan backwards in reverse order, starting from the time slice closest to the tripping time. Within each time slice, examine the events that occurred in the associated event pool within the current time slice. For each event, denoted as event E, the node where the event occurred is... The time of occurrence is ; Query the cloud map related to the entire data domain to find the time of the event. The node it is located at The topological time-varying correlation degree between the current event and the leading edge node; If the topological time-varying correlation degree is greater than or equal to the path filtering threshold at this time, it is considered that when event E occurs, its location is closely electrically connected with the leading edge node of the current event, and the influence of event E may propagate along the electrical connection to the leading edge node of the influencing wave. Therefore, event E will be included in the potential event chain currently being constructed, and the occurrence node of event E will be... Set as the new frontier node; among them, through extensive simulation and historical data analysis, the path selection threshold can be set to 0.25.

[0058] Starting with new frontier nodes, we continue to search for other events with high historical relevance, forming an event sequence that is linked backward in time and space.

[0059] S304. For each potential event chain and each event E contained therein, initiate a quantitative assessment to calculate the contribution weight of the current event to the final trip. The calculation model comprehensively considers four dimensions, each of which is a sub-factor. Specifically: The first factor is the basic weight of the event type: a built-in event type weight mapping table is set based on power system operation experience; The second factor is the spatiotemporal correlation strength factor: directly using the topological time-varying correlation degree at historical moments; The third factor is the time decay factor: set the decay time constant, calculate the time difference between the tripping time and the event occurrence time, and then calculate the time decay factor with the natural constant as the base and the negative value of the ratio of the time difference to the decay time constant as the exponent. The fourth factor is the event's own intensity factor: normalized based on the magnitude of the event's own physical quantity; The contribution weight is calculated as the product of the first factor, the second factor, the third factor, and the fourth factor. The contribution weights of all events in an event chain are standardized and normalized so that the sum of the weights of all events in the current chain is 1.

[0060] S305. Analyze all event chains. If different chains contain the same historical event, add the weights of the current event in different chains and merge them into one record. Set a total chain weight threshold of 0.7, calculate the sum of the normalized contribution weights of all events on each merged event chain, and select the event chain with a total chain weight greater than 0.7 as the dominant trigger chain for this trip.

[0061] It should be noted that by simulating the propagation of events in reverse based on the correlation of historical moments, the potential event chain leading to the tripping can be quantitatively assessed, clearly revealing the evolution sequence of the fault and the contribution of each precursor event, thus realizing the transformation from discussing the event itself to tracing its root cause.

[0062] In this embodiment of the invention, S400 constructs a running stress accumulation model based on the event chain to perform latent defect early warning, including the following steps S401-S404: S401. Continuously receive the event chain, automatically extract all mentioned device identifiers, and label the device identifiers; Each labeled device is scored, and the initial score value is the contribution weight of the current device to the event in the event chain. If the same device is mentioned twice or more in event chains generated at different times, its score is accumulated, and each time it is accumulated, the accumulation coefficient of the new event chain is higher than that of the old event chain. All labeled devices are dynamically sorted based on their scores. The higher the score, the more historical abnormal stress events the device has experienced, and the higher the risk of developing latent defects.

[0063] Check the internal database to determine if each labeled device already has an operational stress accumulation file; If it does not exist, a new file will be created for the current equipment immediately. The operational stress accumulation file includes the equipment's unique identifier, equipment type, rated parameters, and commissioning date. At the same time, the factory test data of the current equipment will be synchronized as a health baseline reference for stress accumulation calculation. S402. Calculate the real-time stress input for all devices with established files, including but not limited to labeled devices, specifically: overload stress increment, operating stress increment, neighborhood fault impact stress increment, and equivalent stress increment. S4021. The overload stress increment is obtained by receiving the current measurement value of the circuit where the current equipment is located in real time. When the effective value of the current continuously exceeds 80% of the rated current of the equipment, it is considered to have entered the stress accumulation range. The overload stress increment calculation is based on the magnitude and duration of the current excess. Specifically, using 1 minute as the basic time unit, the percentage excess of the average current relative to the rated current within this unit is calculated, which is the ratio of the current average current to the rated current minus 0.8. If the result is negative, it is taken as zero. Multiplying the percentage excess by the basic time unit yields the basic load stress micro-element; The load stress increment for the day is obtained by summing up the load stress micro-values ​​of all time units throughout the day and then multiplying them by the load stress coefficient related to the equipment type and cooling method. It should be noted that the load stress coefficient is determined by the equipment thermal model or manufacturer data, and its range is usually between 0.1 and 1.0.

[0064] S4022, The operating stress increment is generated with each operation. The operating stress increment consists of two parts: one is a fixed base value, which represents mechanical wear; the other is a variable value, which is related to the magnitude of the current cut off during operation.

[0065] When the operation occurs, the transient waveform packet is invoked to extract the current amplitude at the instant of operation. The incremental operating stress is obtained by multiplying the ratio of the operating current to the rated breaking current by the arc wear coefficient, and then adding the mechanical stress constant to the obtained value. The mechanical stress constant and the arc wear coefficient are both derived from the mechanical life and electrical life curve parameters provided by the equipment model and manufacturer.

[0066] S4023, Neighborhood Fault Impact Stress Increment: When short-circuit faults, lightning strikes, or other events are detected in the power grid through data correlation cloud maps, the impact of the current event on all equipment is calculated. The specific steps are as follows: For the equipment in the archive, when the fault occurs, the topological time-varying correlation degree between the node where the current equipment is located and the fault point is calculated according to the steps in S2034. The increase in the impact stress of the neighboring fault is proportional to the severity of the fault and inversely proportional to the electrical distance, and decays over time.

[0067] S4024, Equivalent stress increment: Identify unresolved defects in the current device from the event chain. Each defect type is mapped to an equivalent stress rate. The equivalent stress increment will continue to accumulate until the defect is eliminated, at which point the equivalent stress increment input stops.

[0068] S403. Combine all the above stress increments into a comprehensive cumulative stress value that represents the overall cumulative damage level of the equipment; The daily dynamic stress increment is obtained by directly adding the overload stress increment, the operating stress increment, and the neighboring fault impact stress increment. The equivalent force increment is also accumulated daily.

[0069] To avoid the limitations of simple linear accumulation, the overall cumulative stress value of the equipment is updated daily: , in, This represents the overall cumulative stress value after the equipment upgrade. This represents the comprehensive cumulative stress value before equipment upgrades, at the initial commissioning stage. =0; The stress accumulation attenuation factor, 0 < <1, determined by fitting data from accelerated aging tests of the equipment, with typical values ​​between 0.95 and 0.99; This represents the daily increase in dynamic stress before the update. The daily increase in equivalent stress of defects before the update; α is the stress accumulation sensitivity coefficient, α>0, which is related to the existing cumulative damage level of the equipment. A positive correlation was obtained by regression analysis of the relationship between historical stress sequences and final lifespan of a large number of similar equipment failure cases; The time-based stress increment before the update, i.e. ; It should be noted that, It is based on the unresolved defects identified in S4024, and is obtained based on existing power equipment condition-based maintenance rules, equipment failure rate statistical reports, and the experience accumulated by the power grid company. middle This represents the background aging stress value accumulated by the equipment from the time of commissioning to the present due to the mere passage of time (without considering any additional stress). To represent the background aging stress value of the equipment from the time of commissioning to the same time yesterday, both are obtained by fitting the historical life statistics of similar equipment using the material aging theory; in one embodiment, the material aging theory can use the simplified model of the Arrhenius equation.

[0070] S404. The normal aging process of all equipment is depicted as an aging curve, wherein the aging curve has the comprehensive cumulative stress value on the horizontal axis and the remaining health percentage on the vertical axis. Define the attention threshold from the aging curve. and warning threshold ; It should be noted that the attention threshold corresponds to the starting point where the curve begins to rise steeply, while the warning threshold corresponds to the critical point before the failure risk enters an unacceptably high probability range. The thresholds are determined comprehensively through historical operating data of the current equipment group, CS inversion analysis of failure cases, and reliability theory.

[0071] Furthermore, after updating the overall cumulative stress value, the judgment process is executed immediately: like < The device is marked as healthy; like ≤ < The device is marked as requiring attention, its current status is recorded, and a notification is displayed in the device file, but it usually does not trigger an active alarm. like ≥ Immediately issue a latent defect warning, indicating that internal damage has developed to a critical level, but externally observable parameters have not yet shown any abnormalities.

[0072] It should be noted that efforts are being made to address the problem of delayed early warnings and to achieve early prediction of equipment health. By establishing a nonlinear stress accumulation model, the multi-source stresses on the equipment are quantified, and combined with the equipment aging curve, early warnings are issued before the damage reaches the critical point, thus upgrading the operation and maintenance model from reactive maintenance to predictive maintenance.

[0073] In this embodiment of the invention, step S500, based on the latent defect warning result, senses the defect source, including the following steps S501-S504: S501. When a latent defect warning is received, it is timestamped. Based on this, backtrack by a preset analysis time window. It should be noted that the setting of the analysis time window is based on engineering experience. In an optional embodiment, it is set to... In the preceding 24 hours, ensure coverage of critical events that could lead to stress buildup; Furthermore, a query is initiated to the wide-area associated transient data group, with the query condition being within the time interval [ - , Within this scope, all wide-area correlated transient data groups containing early warning devices; S502. Select the wide-area correlated transient data groups that meet the query conditions, prioritizing those with the closest timestamps based on time proximity. Data group; In an optional embodiment, the order of priority can also be based on event intensity: if there are multiple data groups with similar times, the data group with the largest combined waveform amplitude across all nodes is selected. It should be noted that the calculation of the combined maximum value is to take the maximum absolute value of the current change at each node within the data group, and then calculate the arithmetic mean of this maximum value across all nodes, selecting the data group with the largest average value.

[0074] In another alternative embodiment, the data group with the most nodes and the most complete waveform records can be selected according to the order of data integrity priority.

[0075] The selected optimal wide-area correlated transient data set will be locked as the input data for this round of source identification. The original current and voltage waveforms of all nodes in the current data set will be preprocessed, including removing the power frequency component baseline, performing anti-aliasing filtering, and data alignment correction to ensure the accuracy of subsequent analysis.

[0076] S503. Perform a parallel hierarchical analysis process on the preprocessed current waveform: The first layer of analysis uses windowed Fourier transform to analyze the waveforms within 10 power frequency cycles before and after the disturbance, and calculates the amplitude and phase changes of the fundamental wave and its integer harmonics during the disturbance process, usually up to the 13th order. The second layer of analysis uses a time-domain analysis window to slide and scan the preprocessed waveform. The rate of change of the waveform within the window is calculated by summing the absolute values ​​of the differences. When the rate of change exceeds the pulse threshold obtained from the statistical analysis of historical calm period data, it is determined that a pulse may exist. It should be noted that the pulse threshold is usually set to 3 to 5 times the average rate of change of historical calm period. For the detected potential pulse segments, standard exponentially decaying pulses with different rise time constants and decay time constants are used for matching and fitting. The root mean square error is calculated by adjusting the model parameters. The pulse with the highest fit to the real pulse waveform is determined when the root mean square error is minimized, and the set of all pulse features is output.

[0077] The third layer of analysis uses the Prony algorithm to analyze the waveform after removing obvious pulse components. The waveform is represented as a superposition of several damped sinusoidal components with specific frequencies, attenuation factors, amplitudes, and initial phases. The top three oscillation modes with the largest amplitudes are extracted, and the oscillation mode characteristics are output. In an optional embodiment, a fitting method based on a specific damped sine function basis can also be used to analyze the waveform after removing obvious pulse components; It should be noted that a specific frequency can be the center frequency value of each independent oscillation component constituting the waveform, which can be directly calculated from the actual acquired transient waveform data through mathematical decomposition methods (Prony algorithm or damped sine fitting algorithm).

[0078] S504. Source localization is performed for different features, followed by comprehensive analysis. The specific steps are as follows: S5041. Analyze the output pulse feature set, and determine the dominant feature type in this disturbance by comparing the sum of the total pulse energy extracted from all nodes and the sum of the total amplitude of the oscillation modes extracted from all nodes; if the total change in the total pulse energy or the total amplitude of the oscillation mode is the largest, then it is determined to be the dominant feature type of this disturbance.

[0079] S5042. Perform inversion calculations on the dominant feature type. For feature components belonging to the dominant feature type, extract the theoretical observations of the current component: For harmonics, the observed values ​​are the amplitude change and the phase change; For a pulse, the observed values ​​are the pulse feature set and the pulse arrival time; For oscillations, the observed values ​​are amplitude and initial phase.

[0080] Furthermore, based on the known line resistance, inductance, and capacitance distribution parameters of the current distribution network, a distributed parameter network is constructed to simulate the characteristic components with given amplitude and phase. When injected from any point in the distribution network, the amplitude attenuation and phase change when propagating to other nodes are used as theoretical observations and theoretical phases.

[0081] Furthermore, all nodes in the network are sequentially assumed to be potential sources of disturbance. Using the assumed source locations as injection points, a characteristic component of unit intensity is injected into the network. The theoretical observations and theoretical phases that should be generated when propagating to all remaining observation nodes are calculated using forward propagation, which are then used as the theoretical propagation results. The obtained theoretical observations are compared with the actual observations to calculate the comprehensive error at all observation nodes. It should be noted that, in one optional embodiment, for amplitude and phase, the error is calculated as a weighted root mean square error, and the weight can be inversely proportional to the electrical distance between the observation nodes; in another optional embodiment, for pulse delay, the error is the root mean square error of the arrival time difference.

[0082] After traversing all hypothetical source locations, the location with the smallest comprehensive error is selected as the warning source for the current feature component.

[0083] S5043. Perform inversion calculations on all dominant feature components to obtain the location of the early warning source; If the electrical length of the line between the locations of the pre-warning source locations in physical space is less than 1 kilometer, then the cluster area is determined to be the source area.

[0084] It should be noted that solving the problem of ambiguous positioning and achieving precise defect location; performing multi-feature hierarchical analysis and network inversion calculation on the wide-area transient waveforms related to the early warning can accurately locate the electrical source equipment or location that generates abnormal characteristics, making preventive maintenance measures highly targeted.

[0085] In an embodiment of the present invention, step S600 integrates the event chain, early warning results, and defect sources to generate a dynamic sensing map of the distribution network, including the following steps S601-S602: S601, Receive the event chain, the latent defect warning result, and the warning source.

[0086] S602. To comprehensively reflect the urgency, severity, and spatial impact of risks, the computing device nodes calculate dynamically perceived risk values: , in, This represents the dynamic perceived risk value of device node i; , , These are the global weighting coefficients for historical events, current warnings, and source risks, all of which are positive numbers. The initial values ​​are set by domain experts based on operational experience and can be set to 1.0, 2.0, and 1.5 respectively. It should be noted that, The historical event chain risk component represents the potential risk accumulated from historical abnormal events. Specifically, for all event chain records associated with node i in the last 7 days, its contribution weight is multiplied by the time decay factor, where the time decay factor is equal to 1 / (1+Δt), and Δt is the number of days between the current time and the time of the event. All the decayed contribution values ​​are added together to obtain the historical event chain risk component. The current warning risk component represents the explicit but unaddressed defect risks. Specifically, for all unresolved warnings at node i, the warning level is multiplied by the warning aggravation factor, where the warning aggravation factor is equal to (1+Δ). / 10), Δ It is the number of hours between the current time and the time the warning was generated; add all the products together to get the current warning risk component; The disturbance source risk component represents the proactive hazard risk to the power grid caused by the disturbance source. Specifically, for each report with node i as the source, the spatial attenuation factor of the source's influence is calculated, which is equal to 1 / (1+d), where d is the electrical distance from the source to the affected node. The report confidence, spatial attenuation factor, and temporal attenuation factor (consistent with the temporal attenuation factor in the historical event chain risk component) are multiplied together. The calculation results for all reports are summed to obtain the disturbance source risk component. It should be noted that the report confidence is determined based on the fitting error between the theoretical propagation results and the actual observed characteristics in the inversion calculation. The smaller the fitting error, the higher the confidence.

[0087] Based on the power grid topology diagram, the calculations are performed for each node. Values ​​are mapped onto the color spectrum; In an embodiment of the present invention, when When <20, it is marked in blue; when 20≤ When <40, it is marked in green; when 40≤ When <60, it is marked in yellow; when 60≤ When <80, it is marked in orange; when When the value is ≥80, it is marked in red, and a dynamic sensing map of the distribution network is generated.

[0088] Example 3: This example also provides an electronic device applicable to a power distribution network dynamic sensing method, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the power distribution network dynamic sensing method proposed in the above examples.

[0089] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements a dynamic sensing method for the entire distribution network as proposed in the above embodiments.

[0090] The storage medium proposed in this embodiment belongs to the same inventive concept as the method for realizing a dynamic sensing of the entire power distribution network proposed in the above embodiments. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0091] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0092] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for dynamic sensing across the entire distribution network, characterized in that: include, Construct an active sensing layer and collect transient waveform data of the distribution network based on the active sensing layer; The transient waveform data is fused with the power distribution network operation data to generate a global data association cloud map; When a tripping event occurs, the event chain is obtained by tracing back in reverse using the global data association cloud map. Based on the event chain, a stress accumulation model is constructed to provide early warning of hidden defects; Based on the aforementioned latent defect warning results, the source of the defect is determined; Based on the event chain, early warning results, and defect sources, a dynamic perception map of the power distribution network is generated.

2. The method for dynamic sensing of the entire distribution network as described in claim 1, characterized in that, The construction of the active perception layer includes: Based on the topology of the distribution network, monitoring terminals are deployed at key nodes of the distribution network, and micro-disturbance thresholds and lists of electrically related neighbor nodes are set. When any monitoring terminal detects that the electrical quantity exceeds the micro-disturbance threshold, it sends a synchronization trigger command packet. After receiving the synchronization trigger command packet, it performs command validity verification and parsing to obtain the transient waveform packet. The transient waveform packet is time-aligned and encapsulated to generate a wide-area correlated transient data group.

3. The method for dynamic sensing of the entire distribution network as described in claim 2, characterized in that, The fusion includes: Receive the wide-area correlated transient data group to obtain distribution network operation data; The dynamic topological time-varying correlation degree between any two nodes is calculated by fusing the basic electrical coupling factor, power flow influence factor, and applied switch state attenuation coefficient. The calculated dynamic topological time-varying correlation degree is used for visualization mapping to construct a global data correlation cloud map.

4. The method for dynamic sensing of the entire distribution network as described in claim 3, characterized in that, The reverse backtracking includes: When a tripping event occurs, determine the backtracking root node corresponding to the tripping point, and set the backtracking time window with the tripping time as the endpoint. From the global data association cloud map, extract all historical data events within the backtracking time window that have an association strength exceeding a preset threshold with the backtracking root node, and form association events; Starting from the backtracking root node, the process of event impact propagation is simulated to construct a potential event chain; For each event included in the potential event chain, calculate the contribution weight of the current event to the tripping event and perform normalization processing to generate the event chain.

5. The method for dynamic sensing of the entire distribution network as described in claim 4, characterized in that, The latent defect warning includes: Based on the event chain, the equipment identifiers of the involved equipment are marked, and an operational stress accumulation file is established for all marked equipment. The stress increment is set and merged into a comprehensive cumulative stress value that represents the overall cumulative damage level of the equipment, and the normal aging process of all equipment is depicted as an aging curve; A warning threshold is defined from the aging curve. The comprehensive cumulative stress value is compared with the warning threshold. When the comprehensive cumulative stress value is greater than or equal to the warning threshold, a latent defect warning is generated for the current equipment.

6. The method for dynamic sensing of the entire distribution network as described in claim 5, characterized in that, The determination of the source of the defect includes: When a latent defect warning is received, the wide-area correlated transient data group prior to the warning time is acquired; The current and voltage waveforms of the corresponding nodes in the wide-area correlated transient data group are preprocessed; The preprocessed waveform is subjected to a parallel three-layer analysis process, and harmonic features, pulse features and oscillation mode features are extracted respectively. By comparing the energy and amplitude of different characteristics, the dominant characteristic type of this disturbance is determined, and the source of each dominant characteristic type is located.

7. The method for dynamic sensing of the entire distribution network as described in claim 6, characterized in that, The source location includes: For the dominant feature type, a distributed parameter network is constructed, and the nodes in the network are successively assumed to be disturbance sources to simulate the propagation process of feature components; Calculate the error between theoretical propagation results and actual observed characteristics under different disturbance sources; The node with the smallest error is identified as the source of the early warning.

8. The method for dynamic sensing of the entire distribution network as described in claim 7, characterized in that, The dynamic sensing map of the distribution network includes: For any device node, the event chain, the latent defect warning result, and the warning source are received; Dynamically perceived risk values ​​are calculated by analyzing the risk components of historical event chains, current early warning risk components, and risk components of disturbance sources. The calculated dynamic sensing risk values ​​are mapped onto a color spectrum to obtain the dynamic sensing map of the power distribution network.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the dynamic sensing method for the entire distribution network as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the dynamic sensing method for the entire distribution network as described in any one of claims 1 to 8.

Images