Power distribution network abnormal interval detection method and device and readable storage medium

Abstract

The invention belongs to the technical field of power supply and distribution, and relates to a power distribution network abnormal interval detection method and device and a computer readable storage medium. The method comprises the following steps: constructing a topological model by taking equipment in a to-be-detected power distribution network as nodes and taking a connection relationship between the equipment as edges; acquiring historical fault data, and calculating the probability of associated fault of each downstream node connected with each node when each node in the topology model has direct fault based on the historical fault data by using a maximum likelihood estimation method, so as to construct a Bayesian network model; the method comprises the following steps: acquiring an outage range of a to-be-detected power distribution network in real time, taking each upstream node in a node set corresponding to the outage range as a father node in sequence, and calculating a posterior probability of associated faults of downstream nodes connected with each father node when each father node has a direct fault, so as to screen a target father node and a root node; and obtaining an abnormal equipment interval of the to-be-detected power distribution network based on the target father node, the root node and the line between the target node and the root node.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power supply and power distribution, in particular to a power distribution network abnormal section detection method, device and computer readable storage medium. BACKGROUND

[0002] With the rapid development of the power industry, the scale of the power distribution network is increasingly large, and its structure and operation mode are becoming increasingly complex. How to quickly and accurately identify abnormal equipment in the power distribution network from massive data has become a problem to be solved.

[0003] Traditional power distribution network anomaly detection methods are mostly based on statistical principles or deep learning classification ideas. By obtaining the operating parameters of each device, the operating parameters are directly compared with the set threshold to determine whether the device is abnormal, or the operating feature information of each device is extracted, and the deep learning method is used to classify the operating feature information to determine whether each device is abnormal. However, the power distribution network is not a simple collection of isolated devices, but is formed by connecting devices such as lines, switches, and transformers through a complex topology structure. There is conduction and correlation between devices, and the abnormality of each device is not only affected by the components of the device itself, but also affected by the operating state of other devices connected upstream and downstream. Therefore, by detecting the operating state of each device without considering the influencing factors between devices, the existing technology is prone to misjudgment or omission, and cannot accurately identify abnormal device sections. For example, cable failure of a branch line may cause a circuit breaker to trip, thereby causing a sudden drop in the current of the associated line device. If the current parameter of the circuit breaker is detected in isolation, it will be misjudged as a circuit breaker fault, and the line problem cannot be located, nor can the multi-device coordinated section abnormality problem be identified.

[0004] In summary, the existing power distribution network anomaly detection method can only detect a single abnormal device in the power distribution network, and cannot comprehensively consider the correlation and influence between devices, leading to misjudgment or omission, and cannot accurately identify abnormal device sections. SUMMARY

[0005] To this end, the technical problem to be solved by the present application is to overcome the problem that the existing power distribution network anomaly detection method can only detect a single abnormal device in the power distribution network, and cannot comprehensively consider the correlation and influence between devices, leading to misjudgment or omission, and cannot accurately identify abnormal device sections.

[0006] To solve the above technical problems, the present application provides a power distribution network abnormal section detection method, comprising: A topology model of the power distribution network to be detected is constructed by taking each device in the power distribution network to be detected as a node and a connection relationship between each device as an edge; historical fault data of the power distribution network to be detected is acquired, the historical fault data including a fault device, a fault type, a line on which the fault device is located, and a shutdown range in which the fault device is located; the fault type includes a direct fault and a consequential fault; A conditional probability parameter of each downstream node connected to each node in the topology model is calculated by using a maximum likelihood estimation method based on the historical fault data when the direct fault occurs in each node, and a Bayesian network model is obtained based on each node and the conditional probability parameter of each node; A shutdown range of the power distribution network to be detected is acquired in real time, and a node set in the Bayesian network model corresponding to the shutdown range is acquired; each upstream node in the node set is sequentially taken as a parent node, and a posterior probability of a consequential fault occurring in a downstream node connected to each parent node when the direct fault occurs in each parent node is calculated; A number of nodes whose posterior probability is greater than a preset threshold when the direct fault occurs in each parent node is acquired, and each parent node whose number of nodes is greater than a preset number is taken as a target parent node; each target parent node is taken as a root node when the posterior probability of a consequential fault occurring in a downstream node connected to each target parent node is greater than a preset threshold. An abnormal device interval of the power distribution network to be detected is obtained based on each target parent node, each root node, and a line between each target node and its root node.

[0007] Preferably, after the abnormal device interval of the power distribution network to be detected is obtained, the method further includes: A direct fault probability interval [a, b] of each target parent node is set; A midpoint value c of the direct fault probability interval of each target parent node is substituted into a posterior probability formula, a direct fault posterior probability of each target parent node and a change rate of the direct fault posterior probability are calculated; If the change rate of the direct fault probability of a target parent node is 0, the target parent node is taken as a fault core node, and a target abnormal device interval is obtained based on the target parent node and its root node; If the change rate of the direct fault probability of all target parent nodes is not 0, a target parent node with the maximum direct fault posterior probability is taken as a fault core node, and a target abnormal device interval is obtained based on the target parent node and its root node.

[0008] Preferably, the target parent node with the maximum direct fault posterior probability is taken as the fault core node, and the method includes: If the change rate of the direct fault probability of the target parent node is less than 0, the direct fault probability interval of the target parent node is updated to [a, c]; If the change rate of the direct failure probability of the target parent node is greater than 0, the direct failure probability interval of the target parent node is updated to [c, b]; The midpoint value of the direct failure probability interval of the target parent node is substituted into the posterior probability formula again until the difference between the endpoint values of the direct failure probability interval is less than the preset difference value, and the midpoint value of the current direct failure probability interval is substituted into the posterior probability formula to obtain the optimal posterior probability value of the target parent node. The target parent node with the maximum optimal posterior probability value is taken as the fault core node, and the target abnormal equipment interval is obtained based on the target parent node and its root node.

[0009] Preferably, after obtaining the target abnormal equipment interval, the method further comprises: Obtaining the parent node set of the fault core node, and calculating the posterior probability of the fault core node when each parent node thereof has a direct failure; When the posterior probability of the fault core node having a collateral failure is greater than a preset probability, the corresponding parent node is incorporated into the target abnormal equipment interval.

[0010] Preferably, after obtaining the abnormal equipment interval of the power distribution network to be detected, the method further comprises: Obtaining the resistance, inductance and line length of the abnormal equipment interval, and calculating the transmission delay time of the fault signal when transmitted in the abnormal equipment interval; Based on the time when the power distribution network to be detected has a fault and the sum of the time when the fault occurs and the transmission delay time, a fault time window of the power distribution network to be detected is obtained; Based on the topological coordinates of each electromechanical device in the abnormal equipment interval and the fault time window, a three-dimensional fault cuboid model of the power distribution network to be detected is constructed.

[0011] Preferably, the three-dimensional fault cuboid model is expressed as: , wherein, represents a node; represents a node and the topological distance of the fault core node ; represents the maximum topological distance of the fault influence; represents the time when the power distribution network to be detected has a fault; represents the time when the node has a fault; represents the transmission delay time.

[0012] Preferably, after constructing the three-dimensional fault cuboid model of the power distribution network to be detected, the method further comprises visualizing the three-dimensional fault cuboid model by using the topological graph layer and the probability graph layer.

[0013] Preferably, the calculation formula of the posterior probability is: , Wherein, represents the posterior probability of the set S of downstream nodes connected with the parent node set occurring a direct fault and the set S of downstream nodes connected with the parent node set occurring a direct fault and the set S of downstream nodes connected with the parent node set occurring a direct fault and the set S of downstream nodes connected with the parent node set occurring a direct fault and the set S of downstream nodes connected with the parent node set occurring a direct fault and the set S of downstream nodes connected with the parent node set occurring a direct fault and the set S of downstream nodes connected with the parent node set occurring a direct fault and the set S of downstream nodes connected with the parent node set represents the number of downstream nodes connected with the parent node .

[0014] The application also provides a power distribution network abnormal section detection device, comprising: A model construction and data acquisition module is configured to construct a topological model of a power distribution network to be detected by taking each device in the power distribution network to be detected as a node and taking the connection relationship between each device as an edge, and acquire historical fault data of the power distribution network to be detected, wherein the historical fault data comprises a fault device, a fault type, a line on which the fault device is located, and a shutdown range in which the fault device is located, and the fault type comprises a direct fault and a dependent fault; A conditional probability parameter calculation module is configured to calculate the probability of each downstream node connected with each node in the topological model occurring a dependent fault when the node occurs a direct fault based on the historical fault data by using a maximum likelihood estimation method, and obtain conditional probability parameters of each downstream node, and obtain a Bayesian network model based on each node and the conditional probability parameters thereof; A posterior probability calculation module is configured to acquire a shutdown range of the power distribution network to be detected in real time, acquire a set of nodes in the Bayesian network model corresponding to the shutdown range, take each upstream node in the set of nodes as a parent node in turn, and calculate the posterior probability of each downstream node connected with each parent node occurring a dependent fault when the parent node occurs a direct fault; A node screening module is configured to acquire the number of nodes whose posterior probability is greater than a preset threshold when each parent node occurs a direct fault, and take each parent node whose number of nodes is greater than a preset number as a target parent node, and take each node whose posterior probability is greater than a preset threshold among the downstream nodes connected with the target parent node and occurring a dependent fault as a root node; A section positioning module is configured to obtain an abnormal device section of the power distribution network to be detected based on the target parent node, the root node, and a line between the target node and the root node.

[0015] The application further provides a computer readable storage medium, wherein a computer program is stored on the computer readable storage medium, and the computer program is executed by a processor to implement steps of the power distribution network abnormal section detection method.

[0016] The power distribution network abnormal section detection method provided by the application has the following beneficial effects: The application constructs a topology model by taking equipment as a node and a connection relationship as an edge, directly converts the physical connection of the power distribution network into a mathematically computable network structure, and makes the association relationship of "upstream equipment-downstream equipment" clear and visible; the historical fault data are used to calculate the probability of a collateral fault of a child node (downstream) when a parent node (upstream) directly fails by maximum likelihood estimation, and a Bayesian network is constructed based on the probability, so that the association relationship between equipment is upgraded from qualitative description to quantitative probability, and data support is provided for judging the fault propagation range; then, the outage range is obtained in real time and located in the corresponding node set in the Bayesian network, the range possibly affected by the fault is locked, and indiscriminate analysis of all power grid equipment is avoided, and then each upstream node in the set is taken as a suspected direct fault source (parent node), the posterior probability of a collateral fault of a downstream node when the upstream node directly fails is calculated, the number of downstream nodes whose posterior probability is greater than a threshold when each suspected parent node directly fails is counted, and if the number exceeds a preset value, it is indicated that the fault of the parent node can cause a large range of collateral faults, which meets the characteristics of the section abnormality, and therefore the parent node is defined as a target parent node, the nodes whose posterior probability is greater than the threshold downstream of the target parent node are defined as root nodes, and the nodes are specific equipment affected by the core fault source, and finally, the target parent node, the root nodes and the lines therebetween are combined to form a complete abnormal section of the core fault source+collateral fault equipment+connection line, and the equipment abnormal section positioning is realized. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to make the content of the application more easily understood, the application will be further described in detail below according to specific embodiments of the application and in combination with the drawings, in which: Figure 1 A flowchart of the power distribution network abnormal section detection method provided by the application is shown in FIG. 1; Figure 2 A schematic diagram of the Bayesian network model provided by the application is shown in FIG. 2; Figure 3 A structural schematic diagram of the Bayesian network model provided by the application after the nodes in the Bayesian network model are replaced by power distribution network equipment is shown in FIG. 3; Figure 4 A collateral fault probability distribution schematic diagram of each node when a direct fault of a parent node occurs is shown in FIG. 4; Figure 5 A structural diagram of the power distribution network abnormal section detection device provided by the application is shown in FIG. 5. DETAILED DESCRIPTION

[0018] The application will be further described below in conjunction with the accompanying drawings and specific embodiments so that those skilled in the art can better understand and implement the application, but the embodiments are not intended to limit the application.

[0019] Please refer to Figure 1 , Figure 1 The power distribution network abnormal section detection method provided by the application is shown in the flow chart, and the method specifically comprises: S10: taking each device in the power distribution network to be detected as a node and the connection relationship between each device as an edge, a topological model of the power distribution network to be detected is constructed; historical fault data of the power distribution network to be detected is obtained, and the historical fault data comprises a fault device, a fault type, a line on which the fault device is located, and a shutdown range in which the fault device is located; wherein the fault type comprises a direct fault and a dependent fault.

[0020] For example, the historical fault data comprises an ID of the fault device, a direct fault or a dependent fault, a fault line ID, and a fault shutdown range, and can further comprise a fault time, a fault level, a fault frequency, etc.

[0021] S20: using a maximum likelihood estimation method to calculate, based on the historical fault data, a probability that each downstream node connected to each node in the topological model has a dependent fault when the node has a direct fault, to obtain a conditional probability parameter of each downstream node; and obtaining a Bayesian network model based on each node and the conditional probability parameter thereof.

[0022] Specifically, the implementation steps of step S20 comprise: Step 1-1: encoding and discretization processing of continuous data are performed on the historical fault data; for example, the discretization processing of data can be completed by using a decision tree, an equidistance method, an equal frequency method, and Kmeans.

[0023] Step 1-2: the processed data is divided into training samples and test samples which are partially layered with each other, parameter information related to sample quantity and dimension is initialized, a parent node search is performed on each node by using a Bayesian network, and a directed acyclic graph under the historical fault data is obtained.

[0024] Step 1-3: according to the network structure that has been constructed and the historical fault data, a probability that each child node has a dependent fault when a given parent node has a direct fault is calculated as a conditional probability parameter of each child node.

[0025] Specifically, the Bayesian network structure learning adopts a method based on score search, which first defines a certain score function for judging a specific network structure, and measures the matching degree of the independent and dependent relationships contained in the network structure and the data samples. In this embodiment, the Bayesian information criterion (BIC) is used as the score function, and the calculation formula is: .

[0026] The BIC scoring function is decomposable, and its decomposition is expressed as: That is, the BIC score of the entire Bayesian network is obtained by adding the scores of each node, where, Indicates the number of nodes; This represents the combination of prior probabilities that the i-th node will fail when different parent nodes experience direct failures. This indicates the fault type of the i-th node; This represents the number of historical fault data points where the i-th node is of the k-th fault type when the j-th parent node experiences a direct failure. This represents the number of historical failure data points that occurred when the i-th node experienced a direct failure in its j-th parent node. Indicates the number of historical fault data; This represents the BIC score of the i-th node.

[0027] Steps 1-4: Obtain the parameters of the Bayesian network based on the conditional probability parameters of each node, thereby constructing the Bayesian network model, such as... Figure 2 As shown.

[0028] Specifically, the parameter learning of Bayesian networks uses the maximum likelihood estimation method as a criterion. Its basic principle is to find a parameter that maximizes the likelihood function value. Based on historical fault data and the Bayesian network structure, the parameters in the Bayesian network can be calculated. Maximum likelihood estimation: .

[0029] S30: Real-time acquisition of the outage range of the distribution network to be detected, and acquisition of the node set in the Bayesian network model corresponding to the outage range; taking each upstream node in the node set as the parent node in turn, calculate the posterior probability of the downstream nodes connected to each parent node experiencing a direct failure.

[0030] Specifically, the formula for calculating the posterior probability is: , in, This represents the set of parent nodes within the current outage area. The posterior probability that the set of downstream nodes S connected to it will experience a cascading failure when a direct failure occurs; Indicates at the parent node When a direct failure occurs, its connected downstream nodes The conditional probability of a cascading failure occurring; Indicates at the parent node When a direct failure occurs, its connected downstream nodes The conditional probability of a failure occurring; Indicates the relationship with the parent node The number of downstream nodes connected.

[0031] S40: Obtain the number of nodes whose posterior probability is greater than a preset threshold when each parent node experiences a direct failure, and take the parent nodes whose number of nodes is greater than the preset number as target parent nodes; take the nodes whose posterior probability of cascading failures among the downstream nodes connected to each target parent node when it experiences a direct failure is greater than the preset threshold as its root node.

[0032] S50: Based on each target parent node, each root source node, and the lines between each target node and its root source node, the abnormal equipment range of the distribution network to be detected is obtained.

[0033] For example, such as Figure 3 The diagram shown is a structural schematic after replacing the nodes in the Bayesian network model with distribution network equipment. Figure 4 The diagram shows the probability distribution of cascading failures when a parent node experiences a direct failure.

[0034] Furthermore, in some embodiments of this application, after obtaining the abnormal equipment interval of the distribution network to be detected, the method further includes using a binary search to find the fault core node, which specifically includes: Step 2-1: Define the direct failure probability range [a, b] for each target parent node. For example, the direct failure probability range can be [0, 1], where 0 represents the probability that the target parent node has no direct failure and 1 represents the probability that the target parent node has a complete failure.

[0035] Step 2-2: Substitute the midpoint value 'c' of the direct failure probability interval of each target parent node into the posterior probability formula to calculate the direct failure posterior probability of each target parent node, as well as the rate of change of the direct failure posterior probability. Specifically, c = (a + b) / 2, by calculating... This allows us to obtain the direct posterior probability of failure for the target parent node. The derivative is the rate of change of the posterior probability of the direct fault.

[0036] Steps 2-3: If the rate of change of the direct failure probability of the target parent node is 0, then the target parent node is taken as the core fault node, and the target abnormal equipment range is obtained based on the target parent node and its root node.

[0037] Steps 2-4: If the rate of change of the direct failure probability of all target parent nodes is not 0, then the target parent node with the highest direct failure posterior probability is taken as the core fault node, and the target abnormal equipment interval is obtained based on the target parent node and its root node.

[0038] Further, the target parent node with the maximum direct fault posterior probability is taken as the fault core node, comprising: Step 3-1: If the change rate of the direct fault probability of the target parent node is less than 0, the direct fault probability interval of the target parent node is updated to [a, c].

[0039] Step 3-2: If the change rate of the direct fault probability of the target parent node is greater than 0, the direct fault probability interval of the target parent node is updated to [c, b].

[0040] Step 3-3: The midpoint value of the direct fault probability interval of the target parent node is substituted into the posterior probability formula again until the difference between the endpoint values of the direct fault probability interval is less than the preset difference value. The midpoint value of the current direct fault probability interval is substituted into the posterior probability formula to obtain the optimal posterior probability value of the target parent node.

[0041] Step 3-4: The target parent node with the maximum optimal posterior probability value is taken as the fault core node, and the target abnormal device interval is obtained based on the target parent node and its root node.

[0042] Further, after obtaining the target abnormal device interval, the abnormal device interval is further expanded by backtracking the parent nodes from the fault core node, which specifically comprises: Step 4-1: Obtain the parent node set of the fault core node, and calculate the posterior probability of the fault core node when each parent node of the fault core node has a direct fault.

[0043] Step 4-2: When the posterior probability of the fault core node having a collateral fault is greater than a preset probability, the corresponding parent node is incorporated into the target abnormal device interval.

[0044] Specifically, since the fault influence of a parent node on a child node is usually not less than 50%, the parent node with a greater fault influence on the fault core node is selected into the device abnormal interval by setting the preset probability.

[0045] Further, after obtaining the abnormal device interval of the power distribution network to be detected, the abnormal device interval is fused with the fault time dimension to generate a three-dimensional fault cube, i.e., the abnormal device coordinates (x, y) + fault time, by calibrating the time and space dimensions and combining the time and space characteristics of the power distribution network, which specifically comprises: Step 5-1: Obtain the resistance, inductance, and line length of the abnormal device interval, and calculate the transmission delay time of the fault signal when transmitted in the abnormal device interval.

[0046] Specifically, the calculation formula of the transmission delay time is: , wherein L represents the line length, L represents the inductance, Indicates resistance. For example, a 10kV cable line... , The transmission delay coefficient is 0.002s / km; for 35KV overhead lines... , The transmission delay coefficient is 0.0015s / km.

[0047] Step 5-2: Based on the time of the fault in the distribution network under test and the sum of the time of the fault and the transmission delay, obtain the fault time window of the distribution network under test.

[0048] Specifically, based on the fault occurrence time and combined with the distribution network fault response specifications (e.g., the fault location time must be less than or equal to 15 minutes), a fault time window is set. Optionally, for long-distance lines with large transmission delays, the fault time window can be appropriately extended to avoid missing fault time due to delays.

[0049] Step 5-3: Based on the topological coordinates of each electromechanical component within the abnormal equipment interval and the fault time window, construct a three-dimensional fault cube model of the distribution network to be inspected.

[0050] Specifically, the three-dimensional fault cube model Represented as: , in, Represents a node; Represents a node With the faulty core node Topological distance; Indicates the maximum topology distance affected by the fault; Indicates the time when the fault occurred in the distribution network to be tested; Represents a node The time of the failure; This indicates the transmission delay time.

[0051] Furthermore, after constructing the three-dimensional fault cube model of the distribution network to be detected, the method also includes: visualizing the three-dimensional fault cube model using topology layers and probability layers.

[0052] Specifically, the visualization steps include: Step 6-1: Using Neo4j graph database as the storage and rendering core, construct a physical topology visualization model of the power distribution network equipment, specifically including: Different shapes represent different types of equipment (e.g., round - switch, square - circuit breaker, hexagon - transformer, triangle - tower), node size maps failure probability (e.g., diameter range 5-20px, diameter ≥15px when failure probability ≥0.8, 10-15px when 0.5≤ failure probability <0.8, 5-10px when failure probability <0.5), red, orange, yellow, and green colors represent the severity of the fault (e.g., red (RGB: 255,0,0) corresponds to failure probability ≥0.9 (emergency fault), orange (RGB: 255,165,0) corresponds to 0.7≤ failure probability <0.9 (important fault), yellow (RGB: 255,255,0) corresponds to 0.5≤ failure probability <0.7 (general fault), green (RGB: 0,255,0) corresponds to failure probability <0.5 (normal state)), node interaction information: mouse hover displays device details popup, including "fault device ID, fault type, fault time, posterior probability, associated device list", click on the node to jump to the historical fault record page of the device (link to distribution network equipment operation and maintenance management system).

[0053] Edge type and weight: solid line represents main line of distribution network (10kV and above), dashed line represents branch line (10kV and below); edge width maps failure propagation confidence (width range 1-5px, width ≥4px when confidence ≥0.8, 2-4px when 0.5≤ confidence <0.8, 1-2px when confidence <0.5); dynamic effect of edge: add "flickering animation" (flickering frequency 2 times / sec) to the edges on the fault propagation path (i.e., the connection edges between core fault nodes and nodes within the abnormal interval), which intuitively identifies the direction of fault diffusion; if the confidence of an edge ≥0.9 and the corresponding line is an "important power supply line" (such as the line supplying power to a hospital or substation), the color of the edge is superimposed with a red halo, highlighting the urgency.

[0054] Region annotation: In the topological region corresponding to the three-dimensional fault cube, add a semi-transparent polygon annotation (transparency 60%, color consistent with the color of the core node within the interval), which displays "abnormal interval number, number of devices included, number of users affected by the fault (based on distribution network user account statistics), estimated recovery time (initial value set to 2 hours, which can be updated according to the repair progress)", helping operation and maintenance personnel quickly grasp the overall situation of the interval.

[0055] Step 6-2: Based on the conditional probability distribution obtained by parameter learning, construct a multi-dimensional probability information display module, the specific content is as follows: 1. Conditional probability table (CPT) display: The conditional probability table of the core fault nodes (the top 5 high-probability nodes) is presented in a table form. The table columns include "parent node combination (such as switch state = abnormal + line load = overload), sub-node fault probability (such as circuit breaker fault probability = 0.85), sample support number (i.e. the number of historical fault samples corresponding to the parent node combination), confidence level (sample support number / total sample number)". The table supports ordering by "confidence level in descending order", facilitating the focus on statistically significant probability relationships.

[0056] 2. Probability heat map: The fault probability heat map is generated according to the "device type - region" two-dimensional dimension. The horizontal axis is the power distribution network region division (such as "eastern urban area, western urban area, southern urban area"), and the vertical axis is the device type (such as "switch, circuit breaker, transformer, line"). The heat map color depth maps the average fault probability of the "region - device type" (the color gradient is consistent with the topology layer). The heat map supports "drilling view". Clicking on a cell can display the specific fault list of the device type in the region (including device ID, fault time, probability value).

[0057] 3. Probability trend curve: Select the core equipment in the abnormal interval to display the "12-hour fault probability change curve" (horizontal axis: time, interval 1 hour; vertical axis: fault probability). The curve marks the "fault occurrence time point" (red dot) and the "probability mutation point" (such as the time point when the probability suddenly rises from 0.2 to 0.7, marked with an orange triangle), helping the operation and maintenance personnel analyze the fault development law (such as whether there is a "slow accumulation - sudden outbreak" feature).

[0058] Step 6-3: Based on the abnormal interval deduction result, combined with the power distribution network operation and maintenance specification and historical maintenance data, generate targeted decision support scheme, including: Hierarchical maintenance strategy: calculate the maintenance priority according to the fault probability, device importance weight (core device weight 1.2, ordinary device weight 1.0, auxiliary device weight 0.8), formula as follows: Priority = fault probability * device importance weight * influence user number weight; among them, the influence user number weight is set to 1.5 when "influence user number ≥1000 households", 1.2 when "100 ≤ influence user number <1000 households", and 1.0 when "influence user number <100 households"; According to the priority, the maintenance tasks are divided into three levels: first-level tasks (Priority≥1.8): urgent faults, need to be assigned within 30 minutes, and on-site maintenance within 2 hours (such as main transformer failure affecting substation power supply); second-level tasks (1.2≤Priority<1.8): important faults, assigned within 1 hour, and on-site maintenance within 4 hours (such as branch line failure affecting the power supply of the residents in the area); third-level tasks (Priority<1.2): general faults, assigned within 2 hours, and on-site maintenance within 8 hours (such as ordinary pole tower failure without user power outage). At the same time, "standard maintenance process" is provided for each level of task, such as "breaker failure maintenance process" for first-level task: ① remotely operate adjacent switches to isolate fault interval (within 5 minutes); ② send live working vehicle to the site (within 30 minutes); ③ detect faulty components (such as arc-extinguishing chamber, operating mechanism); ④ replace faulty components and debug; ⑤ restore power supply and monitor for 15 minutes.

[0059] Resource scheduling suggestions: Based on the location, equipment type and maintenance priority of the abnormal interval, generate maintenance resource scheduling scheme: personnel scheduling: according to the skill certification of operation and maintenance personnel (such as "high voltage equipment maintenance certification", "line live working certification") and current location, match personnel with corresponding skills and closest to the fault interval (such as breaker failure needs to match "high voltage equipment maintenance certification" personnel, preferentially schedule personnel within 10km distance); Equipment scheduling: list the spare parts (such as breaker arc-extinguishing chamber, transformer bushing) and tools (such as insulated boom truck, high voltage detector) required for maintenance, based on the inventory of distribution network spare parts warehouse (real-time docking with warehouse management system), recommend the nearest spare parts distribution point (such as the fault interval is located in the eastern part of the city, preferentially dispatch spare parts from the eastern warehouse, if the inventory is insufficient, dispatch from the central warehouse); Power outage plan: if maintenance requires power outage, generate power outage notification scheme according to the user type of the fault interval (important users such as hospitals, ordinary residential users), including "notification method (SMS, APP push), notification advance (important users 1 hour in advance, ordinary users 30 minutes in advance), power outage duration estimation", and mark "backup power supply points (such as emergency generator location) that can provide temporary power supply".

[0060] Fault prevention suggestions: Based on the deduction results of this abnormal interval, combined with historical fault data, propose targeted preventive measures: equipment maintenance suggestions: for equipment with fault probability≥0.7 in the abnormal interval, recommend "shortening the maintenance period" (such as original maintenance period of 6 months, shorten to 3 months), and specify the maintenance focus (such as transformer failure needs to focus on detecting oil level, insulation resistance; line failure needs to focus on checking icing, tree barriers); Operation parameter adjustment suggestion: for the operation parameters associated with the fault (such as line load rate, voltage deviation), the threshold value is proposed (such as the original line load rate threshold value is 80%, and it is adjusted to 75% to reduce the risk of overload), and it is suggested to realize the real-time monitoring and early warning of the parameters through the distribution network automation system (such as the DMS system); Environmental response suggestion: if the fault is related to environmental factors (such as rainstorm leading to line short circuit, high temperature leading to equipment overheating), targeted protection measures (such as installing rain cover for line, installing cooling fan for transformer) are recommended, and meteorological warning system is associated to realize "advance inspection before bad weather".

[0061] Step 6-4: The whole process data of the abnormal interval deduction is arranged and archived according to the "distribution network data archiving specification", including: Basic data: original fault data (data set before and after cleaning), Bayesian network model file (structure file, conditional probability table file); inference data: posterior probability calculation result, fault node positioning process data (iteration interval, precision threshold), three-dimensional fault cube data; output data: visualization configuration file (node / edge style parameter), maintenance decision scheme (priority list, dispatching scheme); The archived data adopts the naming rule of "time+interval number" (such as "20250828_abnormal interval 001_basic data.zip"), is stored to the historical database (such as HadoopHDFS) of the distribution network big data platform, and the archiving path, data size and archiving person in charge are recorded, so as to facilitate subsequent tracing and review.

[0062] Based on the abnormal interval detection method of the distribution network provided in the above embodiment, the embodiment of the present application also provides a kind of abnormal interval detection device of distribution network, as shown in Figure Figure 5 The device comprises: Model construction and data acquisition module 10, for each device in the distribution network to be detected as node, with the connection relationship between each device as edge, constructs the topological model of the distribution network to be detected;Obtain the historical fault data of the distribution network to be detected, and the historical fault data includes fault device, fault type, line of fault device, outage range of fault device;Wherein, the fault type includes direct fault and incidental fault.

[0063] Conditional probability parameter calculation module 20, for calculating the probability of each downstream node connected to each node in the topological model occurring incidental fault when the node occurs direct fault based on historical fault data using maximum likelihood estimation method, to obtain the conditional probability parameters of each downstream node;Based on each node and its conditional probability parameters, a Bayesian network model is obtained.

[0064] The posterior probability calculation module 30 is configured to acquire a outage range of the power distribution network to be detected in real time, and acquire a node set in the Bayesian network model corresponding to the outage range; take each upstream node in the node set as a parent node in turn, and calculate a posterior probability of a downstream node connected to the parent node to have a collateral fault when the parent node has a direct fault.

[0065] The node screening module 40 is configured to acquire a number of nodes whose posterior probability is greater than a preset threshold when each parent node has a direct fault, and take a parent node whose number is greater than a preset number as a target parent node; take a node whose posterior probability is greater than a preset threshold among downstream nodes connected to the target parent node as a root node when the target parent node has a direct fault.

[0066] The interval positioning module 50 is configured to obtain an abnormal equipment interval of the power distribution network to be detected based on the target parent node, the root node, and a line between the target node and the root node.

[0067] The embodiment of the application further provides a computer readable storage medium, and the computer readable storage medium stores a computer program. The computer program is executed by a processor to realize the steps of the power distribution network abnormal interval detection method.

[0068] Those skilled in the art should understand that the embodiments of the application can be provided as a method, a system, or a computer program product. Therefore, the application can be in the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the application can be in the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0069] The application is described with reference to the flowcharts and / or block diagrams according to the embodiments of the application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to produce a machine, so that the instructions executed by the computer or other programmable data processing devices produce a device for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The device for implementing the functions specified in one or more flows and / or blocks. Figure 1 The device for implementing the functions specified in one or more flows and / or blocks.

[0070] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the flow Figure 1 The flow or flows and / or blocks Figure 1 The flow or flows and / or blocks

[0071] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions that execute on the computer or other programmable apparatus provide steps for implementing the flow Figure 1 The flow or flows and / or blocks Figure 1 The flow or flows and / or blocks

[0072] Obviously, the above-described embodiments are only examples and are not intended to limit the present application. Other variations and modifications can be made based on the above description and illustrations, and such variations and modifications are also within the scope of the present application. It is not necessary to recite all the embodiments of the present application. The obvious variations and modifications that are derived from the present application are also within the scope of the present application.

Claims

1. A method for detecting an abnormal section of a power distribution network, characterized by, The application relates to a method for detecting an abnormal equipment interval in a power distribution network. The method comprises the following steps: a topology model of the power distribution network to be detected is constructed by taking each device in the power distribution network to be detected as a node and taking the connection relationship between the devices as an edge; historical fault data of the power distribution network to be detected is acquired, the historical fault data comprising a fault device, a fault type, a line where the fault device is located and a shutdown range where the fault device is located; the fault type comprises a direct fault and a consequential fault; the conditional probability parameters of each downstream node are calculated by using a maximum likelihood estimation method based on the historical fault data when each node in the topology model has a direct fault, and a Bayesian network model is obtained based on each node and the conditional probability parameters thereof; the shutdown range of the power distribution network to be detected is acquired in real time, and a node set in the Bayesian network model corresponding to the shutdown range is acquired; each upstream node in the node set is taken as a parent node in turn, and the posterior probability of a downstream node connected to the parent node having a consequential fault when the parent node has a direct fault is calculated; the number of nodes having a posterior probability greater than a preset threshold when each parent node has a direct fault is acquired, and each parent node having a node number greater than a preset number is taken as a target parent node; each target parent node having a posterior probability greater than a preset threshold when a downstream node connected to the target parent node has a consequential fault is taken as a root node of the target parent node; 2. The power distribution network abnormal section detection method according to claim 1, characterized by, an abnormal equipment interval of the power distribution network to be detected is obtained based on each target parent node, each root node and a line between each target node and the root node thereof. After the abnormal equipment interval of the power distribution network to be detected is obtained, the following steps are further included: a direct fault probability interval [a, b] of each target parent node is set; a midpoint value c of the direct fault probability interval of each target parent node is substituted into a posterior probability formula, the direct fault posterior probability of each target parent node and the change rate of the direct fault posterior probability are calculated; if the change rate of the direct fault probability of a target parent node is 0, the target parent node is taken as a fault core node, and a target abnormal equipment interval is obtained based on the target parent node and the root node thereof; 3. The power distribution network abnormal section detection method according to claim 2, characterized by, if the change rate of the direct fault probability of all target parent nodes is not 0, a target parent node having a maximum direct fault posterior probability is taken as a fault core node, and a target abnormal equipment interval is obtained based on the target parent node and the root node thereof. The target parent node having the maximum direct fault posterior probability is taken as the fault core node, and the following steps are included: if the change rate of the direct fault probability of the target parent node is less than 0, the direct fault probability interval of the target parent node is updated as [a, c]; if the change rate of the direct fault probability of the target parent node is greater than 0, the direct fault probability interval of the target parent node is updated as [c, b]; the midpoint value of the direct fault probability interval of the target parent node is substituted into the posterior probability formula again until the difference between the endpoint values of the direct fault probability interval is less than a preset difference value, the midpoint value of the current direct fault probability interval is substituted into the posterior probability formula, and an optimal posterior probability value of the target parent node is obtained; the target parent node having the maximum optimal posterior probability value is taken as the fault core node, and a target abnormal equipment interval is obtained based on the target parent node and the root node thereof.

4. The power distribution network abnormal section detection method according to claim 2 or 3, characterized by, The obtaining of the target abnormal equipment interval further comprises: obtaining a parent node set of the fault core node, and calculating a posterior probability of the fault core node in a collateral fault when each parent node of the fault core node has a direct fault; incorporating a parent node corresponding to the posterior probability of the fault core node in the collateral fault being greater than a preset probability into the target abnormal equipment interval.

5. The power distribution network abnormal section detection method according to claim 1, characterized by, The obtaining of the abnormal equipment interval of the power distribution network to be detected further comprises: obtaining resistance, inductance and line length of the abnormal equipment interval, and calculating a transmission delay time of a fault signal in the abnormal equipment interval; obtaining a fault time window of the power distribution network to be detected based on a time of the power distribution network to be detected in a fault and a sum of the time of the power distribution network to be detected in the fault and the transmission delay time; constructing a three-dimensional fault cubic model of the power distribution network to be detected based on a topological coordinate of each electromechanical in the abnormal equipment interval and the fault time window.

6. The power distribution network abnormal section detection method according to claim 5, characterized by, Three-dimensional fault cube model is represented as: ， wherein, represents a node; represents a node topological distance from a faulty core node topological distance from a faulty core node represents the maximum topological distance of a fault impact; represents the time at which a fault occurs in the power distribution network to be detected; represents a node the time at which a fault occurs; represents a transmission delay time.

7. The power distribution network abnormal section detection method according to claim 5, characterized by, The three-dimensional fault cubic model of the power distribution network to be detected is further constructed, and the three-dimensional fault cubic model is visualized by using the topological graph layer and the probability graph layer.

8. The power distribution network abnormal section detection method according to claim 1, characterized by, The posterior probability is calculated according to the following formula: ， wherein, represents the set of child nodes connected to the parent node occurs a direct failure of the parent node represents the set of child nodes connected to the parent node occurs a direct failure of the parent node represents the conditional probability that the set of child nodes connected to the parent node occurs a direct failure of the parent node represents the conditional probability that the set of child nodes connected to the parent node represents the number of child nodes connected to the parent node 9. An abnormal section detection device for a power distribution network, characterized by comprising: The method comprises: a model construction and data acquisition module, configured to construct a topological model of the power distribution network to be detected by taking each device in the power distribution network to be detected as a node and taking a connection relationship between the devices as an edge, and configured to acquire historical fault data of the power distribution network to be detected, the historical fault data comprising a fault device, a fault type, a line on which the fault device is located and a shutdown range in which the fault device is located, wherein the fault type comprises a direct fault and a collateral fault; a conditional probability parameter calculation module, configured to calculate, based on the historical fault data, a probability of each downstream node connected to each node in the topological model in a collateral fault when the node has a direct fault by using a maximum likelihood estimation method, to obtain a conditional probability parameter of each downstream node, and to obtain a Bayesian network model based on each node and the conditional probability parameter of each node; a posterior probability calculation module, configured to acquire a shutdown range of the power distribution network to be detected in real time, to acquire a node set in the Bayesian network model corresponding to the shutdown range, to take each upstream node in the node set as a parent node in turn, and to calculate a posterior probability of a downstream node connected to each parent node in a collateral fault when the parent node has a direct fault; a node screening module, configured to acquire a number of nodes having a posterior probability greater than a preset threshold when each parent node has a direct fault, and to take a parent node having a number of nodes greater than a preset number as a target parent node, and configured to take a node having a posterior probability greater than the preset threshold in a collateral fault among downstream nodes connected to the target parent node as a root node; an interval positioning module, configured to obtain an abnormal equipment interval of the power distribution network to be detected based on the target parent node, the root node and a line between the target node and the root node.

10. A computer-readable storage medium, characterized in that, The computer program is stored on the computer readable storage medium and is executed by the processor to implement the steps of the power distribution network abnormal interval detection method in any one of claims 1 to 8.

Images