Low-voltage reactive power compensation intelligent monitoring control system

Abstract

The application discloses a kind of low-voltage reactive power compensation intelligent monitoring control systems, and the application relates to low-voltage power distribution technical field, the technical problem of lack of operation and maintenance troubleshooting guide of the fault root cause positioning accuracy of existing system is insufficient.It is solved that the system includes root cause search module and guide generation module: root cause search module constructs reactive power compensation fault knowledge graph in advance, when monitoring real-time alarm, alarm device is mapped as graph entity node, and candidate root cause chain is obtained along the relationship edge search;Guide generation module calculates the comprehensive confidence score of each candidate root cause chain, selects the optimal diagnostic result chain from it, extracts the equipment detection item corresponding to the intermediate node, arranges it as a ladder type troubleshooting operation guide along the causal direction, and pushes it to operation and maintenance interactive terminal.The application improves fault diagnosis accuracy, and is suitable for intelligent monitoring and fault operation and maintenance scene of low-voltage reactive power compensation system.

Description

Technical Field

[0001] This invention relates to the field of low-voltage power distribution technology, specifically to a low-voltage reactive power compensation intelligent monitoring and control system. Background Technology

[0002] Low-voltage reactive power compensation systems are core devices for ensuring power quality in distribution networks. Their operational stability directly affects the power supply reliability and energy efficiency of the distribution system. However, with the increasing scale of reactive power compensation equipment deployment and the growing complexity of its operating conditions, existing monitoring and control systems still have the following shortcomings: Firstly, existing low-voltage reactive power compensation monitoring systems are mostly based on single-point parameter threshold alarms, which can only output discrete equipment alarm signals and cannot reveal the causal propagation relationship of faults between equipment, making it difficult to quickly locate the root cause node after a fault occurs. Secondly, the existing system only outputs fault alarm conclusions and does not generate a tiered operation and maintenance troubleshooting guide. As a result, operation and maintenance personnel need to sort out the detection points and operation sequence themselves, which makes the troubleshooting process lack standardized verification logic and is prone to deviation in the troubleshooting path. Therefore, there is an urgent need for a low-voltage reactive power compensation intelligent monitoring and control system. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides an intelligent monitoring and control system for low-voltage reactive power compensation, which solves the problems of inaccurate fault location and lack of operation and maintenance troubleshooting guidance in existing low-voltage reactive power compensation systems.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a low-voltage reactive power compensation intelligent monitoring and control system, comprising: The root cause search module pre-constructs a knowledge graph of reactive power compensation system faults. When a real-time alarm is detected, the alarm device identifier is parsed from the alarm message, mapped to a device entity node in the graph and set as the source node. At the same time, the initial weights corresponding to the relation edges are calculated, and a directed path search is performed along the relation edges starting from the source node to obtain all candidate root cause chains that terminate at the fault entity node and store them in the candidate set. The graph uses device entity nodes and fault mode entity nodes as nodes. The guidance generation module calculates a comprehensive confidence score for each candidate root cause chain in the candidate set, takes the candidate root cause chain with the highest score as the diagnostic result chain, extracts the device detection items associated with the intermediate nodes in the diagnostic result chain except for the first and last nodes, arranges them along the causal direction of the chain to form a tiered troubleshooting operation guide, and pushes it to the operation and maintenance interactive terminal.

[0005] As a further aspect of the present invention, the specific steps for constructing a fault knowledge graph of a reactive power compensation system are as follows: Parse the engineering configuration file of the reactive power compensation system to obtain all equipment instances and their models, create equipment entity nodes for each equipment, and at the same time, for each equipment model, query all fault mode entries listed in the equipment technical manual, create fault mode entity nodes for each fault mode, and establish "representation" relationship edges from the equipment entity nodes to the fault mode entity nodes. Extract the "type of output electrical parameter change caused" recorded in the manual for this fault mode as the "output parameter type" attribute, extract the "type of required external trigger condition parameter" as the "trigger parameter type" attribute, and mark the two attributes on the entity node of this fault mode; Analyze the electrical main wiring diagram of the reactive power compensation system to obtain the active power flow direction and control signal path between the physical nodes of the equipment. Establish "influence" edges between the equipment nodes along the power flow direction and "control" edges along the control signal path. For each upstream and downstream device node connected by an "influence" edge, the "output parameter type" attribute value of the fault mode entity node associated with the upstream device is compared with the "trigger parameter type" attribute value of the fault mode entity node associated with the downstream device. If they match, a "cause" edge is established between the corresponding fault mode entity nodes.

[0006] As a further aspect of the present invention, the "typical rate of change of each electrical parameter caused" recorded in the manual is recorded as an offset vector in the form of a fixed-dimensional vector, and it is marked on the entity node of the fault mode.

[0007] As a further aspect of the present invention, the specific steps for calculating the initial weights corresponding to the relation edges are as follows: For each pair of fault mode nodes connected by the "causing" edge in the graph, the upstream fault mode node is recorded as the predecessor fault and the downstream fault mode node is recorded as the successor fault. All fault event chains that contain both the predecessor fault and the successor fault are extracted from the historical fault records. The occurrence time of the predecessor fault in the fault event chain must be earlier than the occurrence time of the successor fault. Extract the time of occurrence of the preceding failure and the time of occurrence of the subsequent failure in each event chain, take the difference between the two as the causal propagation delay, sort all causal propagation delays in ascending order and take the median, and take the reciprocal of the median as the initial weight of the "causing" edge.

[0008] As a further aspect of the present invention, the fault event chain refers to an ordered sequence formed by arranging multiple fault events belonging to the same fault propagation process in chronological order of their occurrence from the historical fault records of the reactive power compensation system. Specifically, it needs to meet the following conditions: Condition 1, each element in the sequence corresponds to a specific fault event and must include the equipment identifier, fault mode name, and occurrence timestamp of the fault event; Condition 2, based on two adjacent fault events in the sequence, the propagation delay is obtained by subtracting the occurrence time of the previous fault from the occurrence time of the later fault event. If the propagation delay exceeds a preset causal time span, it is determined that they belong to different propagation processes and are not organized in the same chain.

[0009] As a further aspect of the present invention, the specific steps for performing a directed path search along the relation edges starting from the source node are as follows: Starting from the device entity node set as the source node, follow the "representation" relationship edge in its outward direction to enter all the fault mode entity nodes associated with the device, form the current layer set of nodes to be expanded from these fault mode entity nodes, and use each fault mode entity node as the starting point of a path. Starting from each fault mode entity node in the current layer's set of nodes to be expanded, iteratively expand the neighbor nodes hop by hop along the opposite direction of the "causing" relationship edge. For each hop of expansion, add the current node and the "causing" edge traversed to the sequence of the path, and refresh the current layer set with the newly expanded upstream fault mode nodes. When the extension reaches a node in a certain failure mode and there is no incoming "causing" edge, the extension of the path is immediately terminated, and the node is identified as a candidate root cause node. At the same time, the entire path is stored in the candidate set as a candidate root cause chain.

[0010] As a further aspect of the present invention, the specific steps for calculating the comprehensive confidence score of candidate root cause chains are as follows: Following the causal propagation direction from the starting node to the ending root cause node, the offset vector of each fault mode entity node in the candidate root cause chain is read sequentially, and all offset vectors are accumulated item by item according to the same parameter components to generate the predicted overall offset vector corresponding to the chain, and then temporarily stored in memory. Collect the average value of each electrical parameter in the cycle before the real-time alarm occurs and the instantaneous value at the time of the alarm, calculate the actual rate of change of each parameter, and form an actual offset vector with the same dimension as the predicted overall offset vector. For each candidate root cause chain, the cosine similarity between the predicted overall offset vector and the actual offset vector is calculated and recorded as the comprehensive confidence score of that candidate root cause chain.

[0011] As a further aspect of the present invention, the specific rules for extracting the device detection items associated with intermediate nodes (excluding the first and last nodes) in the diagnostic result chain are as follows: Traverse the diagnostic result chain from the beginning to the end, locate each intermediate fault mode node in the chain except for the first and last nodes, and count the device entity nodes associated with these intermediate fault mode nodes. Obtain the installation location and terminal number of the device entity node through the device entity node. For each intermediate failure mode node, read the downstream trigger parameter type name referenced by its inbound "causing" edge and the upstream output parameter type name referenced by its outbound "causing" edge; After merging and deduplicating the two parameter type names, a list of parameter test items is generated. Each test item is recorded as "measure the current value of this parameter type at the terminal corresponding to the equipment installation location". The test items are then arranged in the order of the intermediate fault mode nodes in the chain to form a test item sequence.

[0012] As a further aspect of the present invention, the specific steps for forming a tiered investigation operation guide are as follows: For each "causing" edge in the diagnostic result chain, read the "output parameter change direction mark" of the upstream fault mode and the "trigger parameter required direction mark" of the downstream fault mode corresponding to that edge; If the two direction markers are consistent, the edge is defined as a forward transmission edge, and the upstream fault mode node and the downstream fault mode node connected by the edge are combined into a transmission verification segment. At the same time, the parameter type name of the edge and the direction marker that matches the comparison are written into the data structure of the transmission verification segment. Arrange all the transmission verification segments in the order of each node in the diagnostic result chain, and write the operation instructions in each transmission verification segment: instruct the operation and maintenance personnel to measure the change direction of the corresponding parameter type on the device terminal associated with the upstream fault mode node; The instruction has a branch step. If the measured direction is consistent with the direction mark recorded in the transmission verification segment, the operation instruction of the next transmission verification segment will be transferred. If they are inconsistent, the troubleshooting sequence will be terminated, and the upstream fault mode node of the current transmission verification segment will be marked as a fault propagation blocking point. At the same time, the blocking point information will be pushed to the operation and maintenance interactive terminal.

[0013] This invention provides a low-voltage reactive power compensation intelligent monitoring and control system, which has the following advantages compared with the prior art: (1) This invention constructs a knowledge graph of faults in a reactive power compensation system, uses the alarm device entity node as the starting point to search backwards to obtain candidate root cause chains, and combines the parameter offset features of the fault mode with real-time operating data to calculate a comprehensive confidence score, which can accurately locate the root cause node of the fault and improve the diagnostic accuracy in multi-device cascading fault scenarios. (2) Based on the diagnostic result chain with the best score, this invention extracts the equipment detection items corresponding to the intermediate nodes and arranges them along the causal direction to form a step-by-step troubleshooting operation guide. Subsequently, the troubleshooting process is standardized by segmented transmission verification and branch judgment logic, which effectively solves the problem that the traditional system only outputs alarm conclusions and lacks practical guidance. Attached Figure Description

[0014] Figure 1 This is the system principle block diagram of the present invention; Figure 2 The flowchart for constructing a fault knowledge graph of the reactive power compensation system for this invention is shown. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] like Figure 1 This invention provides a low-voltage reactive power compensation intelligent monitoring and control system; As an embodiment of this application, it includes: The root cause search module pre-constructs a knowledge graph of reactive power compensation system faults. When a real-time alarm is detected, the alarm device identifier is parsed from the alarm message, mapped to a device entity node in the graph and set as the source node. At the same time, the initial weights corresponding to the relation edges are calculated, and a directed path search is performed along the relation edges starting from the source node to obtain all candidate root cause chains that terminate at the fault entity node and store them in the candidate set. The graph uses device entity nodes and fault mode entity nodes as nodes. The guidance generation module calculates a comprehensive confidence score for each candidate root cause chain in the candidate set, takes the candidate root cause chain with the highest score as the diagnostic result chain, extracts the device detection items associated with the intermediate nodes in the diagnostic result chain except for the first and last nodes, arranges them along the causal direction of the chain to form a tiered troubleshooting operation guide, and pushes it to the operation and maintenance interactive terminal.

[0017] As a second embodiment of this application, it is implemented based on the first embodiment, except that this embodiment includes: The root cause search module is used in reactive power compensation systems, which include multiple equipment entities such as capacitors, contactors, controllers, fuses, and busbars. However, when a fault occurs, it is often not a single equipment failure, but a chain reaction formed through electrical connections, control logic, or protection interlocks. For example, a bulging capacitor may cause a fuse to blow, which in turn leads to overload of other branches after the branch is cut off. The fault phenomenon and the root cause usually span multiple equipment levels. Relying on manual troubleshooting at each level is not only time-consuming but also prone to misjudgment. In addition, traditional monitoring systems can only provide discrete alarm signals and cannot reveal the causal relationship between alarms. Therefore, it is necessary to pre-construct a knowledge graph of reactive power compensation system faults and calculate the initial weight for each "causing" edge based on historical fault records. The specific rules are as follows: For each pair of fault mode nodes connected by the "causing" edge in the graph, the upstream fault mode node is recorded as the predecessor fault and the downstream fault mode node is recorded as the successor fault. All fault event chains that contain both the predecessor fault and the successor fault are extracted from the historical fault records. The occurrence time of the predecessor fault in the fault event chain must be earlier than the occurrence time of the successor fault. The fault event chain refers to an ordered sequence formed by arranging multiple fault events belonging to the same fault propagation process from the historical fault records of the reactive power compensation system in chronological order of their occurrence. Specifically, it needs to meet the following conditions: Condition 1: Each element in the sequence corresponds to a specific fault event and must include the device identifier, fault mode name, and timestamp of the occurrence of the fault event; Condition 2: Based on two adjacent fault events in the sequence, the propagation delay is obtained by subtracting the occurrence time of the previous fault from the occurrence time of the subsequent fault event. If the propagation delay exceeds the preset causal time span, it is determined that they belong to different propagation processes and are not organized in the same chain. Extract the time of occurrence of the preceding failure and the time of occurrence of the subsequent failure in each event chain, take the difference between the two as the causal propagation delay, sort all causal propagation delays in ascending order and take the median, and take the reciprocal of the median as the initial weight of the "causing" edge. When a real-time alarm is detected, the alarm device identifier is parsed from the alarm message. This is because in a reactive power compensation system, the alarm message formats and field definitions generated by various protection devices, controllers, and sensors are often different. Some are identified by the device number, while others are identified by the measurement point code or logical name. Without parsing, the system cannot extract the identity information pointing to a specific physical device from heterogeneous messages. The core of the parsing process is to extract and restore the device identification field from the original message string according to the communication protocol of the message (such as IEC 61850, Modbus or proprietary protocol), and convert it into a standardized identifier. The alarm device identifier is a string or number code, which does not have semantic relationships or topological information, while the device entity node in the knowledge graph contains a structured object with multiple attributes such as device type and associated fault mode. Therefore, by mapping the alarm device identifier to the device entity node in the graph and setting it as the source node, the system maps the externally obtained flat identifier to the device node with complete semantics in the graph through the mapping relationship, so that the alarm information is upgraded from a simple label to an entity with reasoning ability. The mapping relationship needs to be established in advance during the graph construction phase. For example, in the project configuration file, the device identifier of "Capacitor #1" is "CAP_001", and the corresponding device entity node name in the graph is "Capacitor #1". The system establishes the binding relationship between the two in advance during initialization. When the alarm message parses out "CAP_001", it directly indexes the node through the mapping mechanism without real-time searching or guessing. Even if the device identifier changes when the system is expanded, only the mapping table needs to be updated. Simultaneously, load the relation edges and their corresponding initial weights, and perform a directed path search along the causal directions of the relation edges, starting from the source node. The specific operation is as follows: Starting from the device entity node set as the source node, follow the "representation" relationship edge in its outward direction to enter all the fault mode entity nodes associated with the device, form the current layer set of nodes to be expanded from these fault mode entity nodes, and use each fault mode entity node as the starting point of a path. Starting from each fault mode entity node in the current layer's set of nodes to be expanded, iteratively expand the neighbor nodes hop by hop along the opposite direction of the "causing" relationship edge. For each hop of expansion, add the current node and the "causing" edge traversed to the sequence of the path, and refresh the current layer set with the newly expanded upstream fault mode nodes. When the extension reaches a node in a certain failure mode and there is no incoming "causing" edge, the extension of the path is immediately terminated, and the node is identified as a candidate root cause node. At the same time, the entire path is stored in the candidate set as a candidate root cause chain.

[0018] The guidance generation module calculates a comprehensive confidence score for each candidate root cause chain in the candidate set. The specific steps are as follows: Following the causal propagation direction from the starting node to the ending root cause node, the offset vector of each failure mode entity node in the candidate root cause chain is read sequentially (obtainable from the constructed knowledge graph), and all offset vectors are accumulated item by item according to the same parameter components to generate the overall predicted offset vector corresponding to the chain, and then temporarily stored in memory. Collect the average value of each electrical parameter in the cycle before the real-time alarm occurs and the instantaneous value at the time of the alarm, calculate the actual rate of change of each parameter, and form an actual offset vector with the same dimension as the predicted overall offset vector. For each candidate root cause chain, calculate the cosine similarity between the predicted overall offset vector and the actual offset vector, and record it as the comprehensive confidence score of that candidate root cause chain; The candidate root cause chain with the highest score is used as the diagnostic result chain, and the device detection items associated with the intermediate nodes in the diagnostic result chain, excluding the first and last nodes, are extracted. The specific rules are as follows: Traverse the diagnostic result chain from the beginning to the end, locate each intermediate fault mode node in the chain except for the first and last nodes, and count the device entity nodes associated with these intermediate fault mode nodes. Obtain the installation location and terminal number of the device entity node through the device entity node. For each intermediate failure mode node, read the downstream trigger parameter type name referenced by its inbound "causing" edge and the upstream output parameter type name referenced by its outbound "causing" edge; After merging and deduplicating the two parameter type names, a list of parameter test items is generated. Each test item is recorded as "measure the current value of this parameter type at the terminal corresponding to the equipment installation location". The test items are then arranged in order of their sequence in the chain according to the order of the intermediate fault mode nodes to form a test item sequence. The diagnostic results are arranged along a causal chain to form a tiered screening procedure. The specific steps are as follows: For each "causing" edge in the diagnostic result chain, read the "output parameter change direction mark" of the upstream fault mode and the "trigger parameter required direction mark" of the downstream fault mode corresponding to that edge; The output parameter change direction mark refers to the change trend of the output electrical parameters relative to the normal operating state when the upstream fault mode actually occurs. This change trend is extracted based on the qualitative description of the fault mode recorded in the equipment technical manual, and the value is one of three: increase, decrease, or no change. For example, if the manual describes a certain fault mode as "causing an increase in line current amplitude", then the output parameter change direction mark of the physical node of the fault mode is marked as "increase". The required direction mark for the triggering parameter refers to the direction of change that the sensitive external electrical parameter should exhibit when the downstream fault mode is triggered. This direction of change is extracted based on the qualitative description of the triggering condition of the fault mode in the equipment technical manual, and the value is one of three: increase, decrease, or no change. For example, if the manual records a fault mode as "triggered when the external current amplitude increases", then the required direction mark for the triggering parameter of the entity node of the fault mode is marked as "increase". If the two direction markers are consistent, the edge is defined as a forward transmission edge, and the upstream fault mode node and the downstream fault mode node connected by the edge are combined into a transmission verification segment. At the same time, the parameter type name of the edge and the direction marker that matches the comparison are written into the data structure of the transmission verification segment. Arrange all the transmission verification segments in the order of each node in the diagnostic result chain, and write the operation instructions in each transmission verification segment: instruct the operation and maintenance personnel to measure the change direction of the corresponding parameter type on the device terminal associated with the upstream fault mode node; The instruction has a branch step. If the measured direction is consistent with the direction mark recorded in the transmission verification segment, the operation instruction of the next transmission verification segment will be transferred. If they are inconsistent, the troubleshooting sequence will be terminated, and the upstream fault mode node of the current transmission verification segment will be marked as a fault propagation blocking point. At the same time, the blocking point information will be pushed to the operation and maintenance interactive terminal.

[0019] As a third embodiment of this application, this embodiment further discloses a method for constructing a fault knowledge graph of a reactive power compensation system based on embodiments one and two, such as... Figure 2 As shown, the specific content includes: Parse the engineering configuration file of the reactive power compensation system to obtain all equipment instances and their models, create equipment entity nodes for each equipment, and at the same time, for each equipment model, query all fault mode entries listed in the equipment technical manual, create fault mode entity nodes for each fault mode, and establish "representation" relationship edges from the equipment entity nodes to the fault mode entity nodes. Extract the "type of output electrical parameter change caused" recorded in the manual for this fault mode as the "output parameter type" attribute, extract the "type of required external trigger condition parameter" as the "trigger parameter type" attribute, and mark the two attributes on the entity node of this fault mode; Meanwhile, the "typical rate of change of each electrical parameter caused" recorded in the manual is recorded as an offset vector in the form of a fixed-dimensional vector and marked on the entity node of the fault mode. For example, the "capacitor internal breakdown" fault mode is described in the manual as causing the line current amplitude to increase to 3 times the rated value (i.e., the rate of change +200%), while simultaneously causing the bus voltage amplitude to decrease to 0.85 times the rated value (i.e., the rate of change -15%). Its offset vector can then be expressed as: offset vector = {"line current amplitude": +200%, "bus voltage amplitude": -15%, ...}, where the ellipsis indicates that the default rate of change for other parameter dimensions is 0; Analyze the electrical main wiring diagram of the reactive power compensation system to obtain the active power flow direction and control signal path between the physical nodes of the equipment. Establish "influence" edges between the equipment nodes along the power flow direction and "control" edges along the control signal path. For each upstream and downstream device node connected by an "influence" edge, the "output parameter type" attribute value of the fault mode entity node associated with the upstream device is compared with the "trigger parameter type" attribute value of the fault mode entity node associated with the downstream device. If they match, a "cause" edge is established between the corresponding fault mode entity nodes.

[0020] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.

[0021] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended 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 methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A low-voltage reactive power compensation intelligent monitoring and control system, characterized in that, include: The root cause search module pre-constructs a knowledge graph of reactive power compensation system faults. When a real-time alarm is detected, the alarm device identifier is parsed from the alarm message, mapped to a device entity node in the graph and set as the source node. At the same time, the initial weights corresponding to the relation edges are calculated, and a directed path search is performed along the relation edges starting from the source node to obtain all candidate root cause chains that terminate at the fault entity node and store them in the candidate set. The graph uses device entity nodes and fault mode entity nodes as nodes. The guidance generation module calculates a comprehensive confidence score for each candidate root cause chain in the candidate set, takes the candidate root cause chain with the highest score as the diagnostic result chain, extracts the device detection items associated with the intermediate nodes in the diagnostic result chain except for the first and last nodes, arranges them along the causal direction of the chain to form a tiered troubleshooting operation guide, and pushes it to the operation and maintenance interactive terminal.

2. The low-voltage reactive power compensation intelligent monitoring and control system according to claim 1, characterized in that, The specific steps for constructing a fault knowledge graph for a reactive power compensation system are as follows: Parse the engineering configuration file of the reactive power compensation system to obtain all equipment instances and their models, create equipment entity nodes for each equipment, and at the same time, for each equipment model, query all fault mode entries listed in the equipment technical manual, create fault mode entity nodes for each fault mode, and establish "representation" relationship edges from the equipment entity nodes to the fault mode entity nodes. Extract the "type of output electrical parameter change caused" recorded in the manual for this fault mode as the "output parameter type" attribute, extract the "type of required external trigger condition parameter" as the "trigger parameter type" attribute, and mark the two attributes on the entity node of this fault mode; Analyze the electrical main wiring diagram of the reactive power compensation system to obtain the active power flow direction and control signal path between the physical nodes of the equipment. Establish "influence" edges between the equipment nodes along the power flow direction and "control" edges along the control signal path. For each upstream and downstream device node connected by an "influence" edge, the "output parameter type" attribute value of the fault mode entity node associated with the upstream device is compared with the "trigger parameter type" attribute value of the fault mode entity node associated with the downstream device. If they match, a "cause" edge is established between the corresponding fault mode entity nodes.

3. The low-voltage reactive power compensation intelligent monitoring and control system according to claim 2, characterized in that, The "typical rate of change of each electrical parameter caused" recorded in the manual is recorded as an offset vector in the form of a fixed-dimensional vector, and it is marked on the entity node of the fault mode.

4. The low-voltage reactive power compensation intelligent monitoring and control system according to claim 2, characterized in that, The specific steps for calculating the initial weights corresponding to relation edges are as follows: For each pair of fault mode nodes connected by each "causing" edge in the graph, the upstream fault mode node is recorded as the predecessor fault and the downstream fault mode node is recorded as the successor fault. All fault event chains that simultaneously contain the predecessor fault and the successor fault are extracted from the historical fault records. The occurrence time of the predecessor fault in the fault event chain must be earlier than the occurrence time of the successor fault. Extract the time of occurrence of the preceding failure and the time of occurrence of the subsequent failure in each event chain, take the difference between the two as the causal propagation delay, sort all causal propagation delays in ascending order and take the median, and take the reciprocal of the median as the initial weight of the "causing" edge.

5. The low-voltage reactive power compensation intelligent monitoring and control system according to claim 4, characterized in that, The fault event chain refers to an ordered sequence formed by arranging multiple fault events belonging to the same fault propagation process in chronological order of their occurrence from the historical fault records of the reactive power compensation system. Specifically, it needs to meet the following conditions: Condition 1: Each element in the sequence corresponds to a specific fault event and must include the equipment identifier, fault mode name, and occurrence timestamp of the fault event; Condition 2: Based on two adjacent fault events in the sequence, the propagation delay is obtained by subtracting the occurrence time of the previous fault from the occurrence time of the later fault event. If the propagation delay exceeds the preset causal time span, it is determined that they belong to different propagation processes and are not organized in the same chain.

6. The low-voltage reactive power compensation intelligent monitoring and control system according to claim 1, characterized in that, The specific steps for a directed path search along the relation edges starting from the source node are as follows: Starting from the device entity node set as the source node, follow the "representation" relationship edge in its outward direction to enter all the fault mode entity nodes associated with the device, form the current layer set of nodes to be expanded from these fault mode entity nodes, and use each fault mode entity node as the starting point of a path. Starting from each fault mode entity node in the current layer's set of nodes to be expanded, iteratively expand the neighboring nodes hop by hop in the opposite direction of the "cause" relationship edge. For each hop of expansion, add the current node and the "cause" edge traversed to the sequence of the path, and refresh the current layer set with the newly expanded upstream fault mode nodes. When the extension reaches a fault mode node where there is no incoming "causing" edge, the extension of the path is immediately terminated, and the node is identified as a candidate root cause node. At the same time, the entire path is stored as a candidate root cause chain in the candidate set.

7. The low-voltage reactive power compensation intelligent monitoring and control system according to claim 1, characterized in that, The specific steps for calculating the overall confidence score for candidate root cause chains are as follows: Following the causal propagation direction from the starting node to the ending root cause node, the offset vector of each fault mode entity node in the candidate root cause chain is read sequentially, and all offset vectors are accumulated item by item according to the same parameter components to generate the predicted overall offset vector corresponding to the chain, and then temporarily stored in memory. Collect the average value of each electrical parameter in the cycle before the real-time alarm occurs and the instantaneous value at the time of the alarm, calculate the actual rate of change of each parameter, and form an actual offset vector with the same dimension as the predicted overall offset vector. For each candidate root cause chain, the cosine similarity between the predicted overall offset vector and the actual offset vector is calculated and recorded as the comprehensive confidence score of that candidate root cause chain.

8. The low-voltage reactive power compensation intelligent monitoring and control system according to claim 1, characterized in that, The specific rules for extracting the device test items associated with intermediate nodes (excluding the first and last nodes) in the diagnostic result chain are as follows: Traverse the diagnostic result chain from the beginning to the end, locate each intermediate fault mode node in the chain except for the first and last nodes, and count the device entity nodes associated with these intermediate fault mode nodes. Obtain the installation location and terminal number of the device entity node through the device entity node. For each intermediate failure mode node, read the downstream trigger parameter type name referenced by its inbound "causing" edge and the upstream output parameter type name referenced by its outbound "causing" edge; After merging and deduplicating the two parameter type names, a list of parameter test items is generated. Each test item is recorded as "measure the current value of this parameter type at the terminal corresponding to the equipment installation location". The test items are then arranged in order of their sequence in the chain according to the intermediate fault mode nodes, forming a test item sequence.

9. The low-voltage reactive power compensation intelligent monitoring and control system according to claim 1, characterized in that, The specific steps for creating a tiered investigation guideline are as follows: For each "causing" edge in the diagnostic result chain, read the "output parameter change direction mark" of the upstream fault mode and the "trigger parameter required direction mark" of the downstream fault mode corresponding to the edge; If the two direction markers are consistent, the edge is defined as a forward transmission edge, and the upstream fault mode node and the downstream fault mode node connected by the edge are combined into a transmission verification segment. At the same time, the parameter type name of the edge and the direction marker that matches the comparison are written into the data structure of the transmission verification segment. Arrange all the transmission verification segments in the order of each node in the diagnostic result chain, and write the operation instructions in each transmission verification segment: instruct the operation and maintenance personnel to measure the change direction of the corresponding parameter type on the device terminal associated with the upstream fault mode node; The instruction has a branch step. If the measured direction is consistent with the direction mark recorded in the transmission verification segment, the operation instruction of the next transmission verification segment will be transferred. If they are inconsistent, the troubleshooting sequence will be terminated, and the upstream fault mode node of the current transmission verification segment will be marked as a fault propagation blocking point. At the same time, the blocking point information will be pushed to the operation and maintenance interactive terminal.

Images