Knowledge graph-based dynamic power question and answer system optimization method and system

Abstract

The application provides a dynamic power question and answer system optimization method and system based on a knowledge graph, wherein the method comprises: collecting heterogeneous operation data of a power equipment at multiple positions in real time, including rule document rule constraint items, sensor time sequence state signals and historical maintenance event association items. The rule constraint items and the event association items are mapped to a unified semantic space and semantic cross interference is eliminated, and a topological structure is constructed. Then the topological structure is dynamically coupled with real-time time sequence signals to form a power equipment state evolution graph. Based on the graph and the current state verification parameters uploaded by the distributed sensors, the basic association strength values between nodes are dynamically calibrated, and boundary condition constraints are applied through the rule constraint items. Finally, the event association items are matched with the calibrated association strength values to generate an abnormal reasoning path optimization strategy for question and answer requests. The application improves the accuracy and real-time performance of power equipment abnormal diagnosis.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of knowledge graph, and in particular to a dynamic power question and answer system optimization method and system based on knowledge graph. BACKGROUND

[0002] With the advancement of smart grid construction, the power system needs to process real-time multi-source heterogeneous data of device operation state, including sensor monitoring data, operation and maintenance procedure documents and historical fault records. The traditional manual question and answer mode cannot meet the demand for rapid positioning and reasoning decision of device abnormalities, and an automatic system that can integrate multi-modal data, dynamically update knowledge graph and support intelligent question and answer is urgently needed to realize accurate fault diagnosis and optimization of disposal scheme.

[0003] The current existing scheme adopts a power question and answer system based on static knowledge graph, which matches the graph nodes by combining sensor real-time data with the pre-constructed graph model containing device attributes and relationships for question and answer reasoning. The system uses a graph database to store device topology relationships, and triggers a pre-defined rule chain to complete abnormality judgment when the sensor data exceeds the threshold.

[0004] The static knowledge graph is difficult to adapt to the dynamic change characteristics of device operation state, and the simple threshold matching method of sensor data and graph nodes ignores the spatio-temporal correlation characteristics between multi-modal data, resulting in deviation of abnormal reasoning results from actual fault propagation path, and the disposal strategy output by the question and answer system often lacks predictability for complex chain faults. SUMMARY

[0005] The present application provides a dynamic power question and answer system optimization method and system based on knowledge graph, to solve the problems of low accuracy and poor real-time performance of power device abnormality diagnosis in the prior art.

[0006] In a first aspect, the present application provides a dynamic power question and answer system optimization method based on knowledge graph, comprising:

[0007] Real-time collection of heterogeneous operation data at different positions of power equipment, the heterogeneous operation data including rule constraint items of procedure documents, time sequence state signals continuously generated by sensors, and event association items in historical maintenance records;

[0008] Mapping the rule constraint items and the event association items into a unified semantic space, eliminating the cross interference of repetitive semantic units in the semantic space, to construct a topology structure;

[0009] Cross-modal dynamic coupling of the topology structure and the time sequence state signal to generate a power equipment state evolution graph;

[0010] The basic correlation strength values between nodes in the topology structure are dynamically calibrated based on the power equipment state evolution graph and the current equipment state verification parameters uploaded by each node in the distributed sensor network deployed in the power equipment, and the calibrated correlation strength values are subjected to boundary condition constraints through the rule constraint term;

[0011] The event correlation term is matched with the correlation strength values subjected to boundary condition constraints to generate an abnormal reasoning path optimization strategy for the question and answer system request.

[0012] Optionally, the current equipment state verification parameters include spatial position identifiers and real-time monitoring values of the distributed sensor nodes.

[0013] The basic correlation strength values between nodes in the topology structure are dynamically calibrated based on the power equipment state evolution graph and the current equipment state verification parameters uploaded by each node in the distributed sensor network deployed in the power equipment, and the calibrated correlation strength values are subjected to boundary condition constraints through the rule constraint term, including:

[0014] According to the matching result of the spatial position identifier and the corresponding entity node in the power equipment state evolution graph, the real-time monitoring value is subjected to deviation amount calculation with the historical state transition path of the corresponding entity node.

[0015] When the deviation result exceeds the preset correlation strength allowable fluctuation range, the basic correlation strength values between nodes on the backtracking path are calibrated by backtracking along the state transition path network in the topology structure to the deviation starting node.

[0016] The operation critical value defined in the rule constraint term is taken as the boundary condition constraint, and when the calibrated correlation strength value is greater than the operation critical value, the operation critical value is taken as the correlation strength value subjected to boundary condition constraints.

[0017] Optionally, the basic correlation strength values between nodes on the backtracking path are calibrated by backtracking along the state transition path network in the topology structure to the deviation starting node, including:

[0018] The difference direction between the deviation result and the preset correlation strength allowable fluctuation range is detected, and the forward or reverse backtracking mode along the state transition path network is determined according to the difference direction.

[0019] According to the backtracking mode, the deviation starting node is backtracked from the current entity node along the direction of the connection line in the state transition path network, and the time stamp interval length and state transfer amount cumulative value of the current entity node in the backtracking path are recorded.

[0020] According to the time stamp interval length and the state transmission amount cumulative value, a target correlation strength attenuation coefficient between each two adjacent entity nodes in the backtracking path is calculated;

[0021] On the backtracking path, the target correlation strength attenuation coefficient is multiplied by a basic correlation strength value between corresponding entity nodes in a direction from the deviation starting node to a current entity node to generate a calibrated correlation strength value.

[0022] Optionally, the calculation of the target correlation strength attenuation coefficient between each two adjacent entity nodes in the backtracking path according to the time stamp interval length and the state transmission amount cumulative value comprises:

[0023] The time stamp interval length is span compared with a preset reference time window to generate a time span influence factor;

[0024] The state transmission amount cumulative value is amplitude compared with a preset reference transmission amount to generate a transmission amount influence factor;

[0025] According to an inverse ratio relationship between the time span influence factor and the transmission amount influence factor, an initial correlation strength attenuation coefficient is generated;

[0026] When a value of the initial correlation strength attenuation coefficient exceeds a preset attenuation allowed range, upper and lower limit values of the attenuation allowed range are taken as forced constraint boundaries, and the initial correlation strength attenuation coefficient is subjected to boundary truncation processing to generate a target correlation strength attenuation coefficient.

[0027] Optionally, the matching of the event correlation item with the correlation strength value subjected to boundary condition constraint to generate an abnormal reasoning path optimization strategy for a question and answer system request comprises:

[0028] A feature parameter set of a historical abnormal event is extracted from the event correlation item, and the feature parameter set includes an event trigger node identifier, a correlation strength mutation threshold value, and an influence path length;

[0029] A key entity node matching the event trigger node identifier is searched in the power equipment state evolution graph;

[0030] When a corresponding correlation strength value subjected to boundary condition constraint between the key entity node and an adjacent node reaches the correlation strength mutation threshold value, the key entity node is marked as an abnormal trigger source;

[0031] An abnormal reasoning path is traced in a direction of correlation strength attenuation in the power equipment state evolution graph from the abnormal trigger source as a starting point;

[0032] Sort all nodes in the dependency path according to a spatial decay mode to generate an ordered path chain from the anomaly trigger source to the set threshold node, and generate an anomaly reasoning path optimization strategy according to the ordered path chain and the question and answer system request.

[0033] Optionally, the rule constraint item and the event association item are mapped into a unified semantic space, cross interference of repetitive semantic units in the semantic space is eliminated, and a topology structure is constructed, including:

[0034] Text description units in the rule constraint item and historical event identification units in the event association item are converted into semantic vector units;

[0035] Repetitive semantic unit groups for describing the same power equipment entity or the same operation behavior in the semantic vector are identified, semantic superposition is performed on all repetitive semantic units in each group of repetitive semantic unit groups, and semantic vector expression units are generated;

[0036] A dependent relationship between the semantic vector expression units is determined, and a unidirectional dependent link between power equipment entities in the unified semantic space is established according to the dependent relationship;

[0037] A topology structure is formed through hierarchical expansion of the unidirectional dependent link.

[0038] Optionally, the topology structure and the time sequence state signal are cross-modally and dynamically coupled to generate a power equipment state evolution graph, including:

[0039] A power equipment state parameter set at each collection time point in the time sequence state signal is extracted;

[0040] The power equipment state parameter set is bound to a corresponding power equipment entity node in the topology structure;

[0041] At consecutive collection time points, a state transfer amount between adjacent nodes is calculated according to the device state parameter set along a connection line direction of the topology structure;

[0042] When the state transfer amount exceeds a preset transfer threshold, a state transition marker from the current collection time point to the next collection time point is added in the topology structure;

[0043] A state transition path network with a time stamp is formed in the topology structure and a basic correlation strength value between adjacent nodes is generated by aggregating state transition markers at all collection time points;

[0044] A power equipment state evolution graph is constructed according to the state transition path network and the basic correlation strength value.

[0045] In a second aspect, the application provides a dynamic power question-answering system optimization system based on a knowledge graph, comprising:

[0046] A collection module is configured to collect heterogeneous operation data at different positions of a power device in real time, wherein the heterogeneous operation data comprises rule constraint items of a procedure document, time-series state signals continuously generated by sensors, and event association items in historical maintenance records;

[0047] A construction module is configured to map the rule constraint items and the event association items into a unified semantic space, eliminate cross interference of repetitive semantic units in the semantic space, and construct a topological structure;

[0048] A generation module is configured to cross-modally and dynamically couple the topological structure and the time-series state signals to generate a power device state evolution graph;

[0049] A calibration module is configured to dynamically calibrate basic association strength values between nodes in the power device state evolution graph based on current device state verification parameters uploaded on each node of a distributed sensor network deployed in the power device, and impose boundary condition constraints on the calibrated association strength values through the rule constraint items;

[0050] A matching module is configured to match the event association items with the association strength values subjected to the boundary condition constraints to generate an abnormal reasoning path optimization strategy for a question-answering system request.

[0051] In a third aspect, the application provides a computing device comprising a processor and a memory, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the dynamic power question-answering system optimization method based on a knowledge graph according to any one of the first aspect.

[0052] In a fourth aspect, the application provides a computer storage medium storing computer program instructions, wherein the computer program instructions are executed by a processor to implement the dynamic power question-answering system optimization method based on a knowledge graph according to any one of the first aspect.

[0053] In the application, a dynamic power question and answer system optimization method based on a knowledge graph is provided, the method comprising: collecting heterogeneous operation data at different positions of power equipment in real time, wherein the heterogeneous operation data comprises rule constraint items of procedure documents, time sequence state signals continuously generated by sensors, and event association items in historical maintenance records; mapping the rule constraint items and the event association items into a unified semantic space, eliminating cross interference of repetitive semantic units in the semantic space, to construct a topological structure; cross-modal dynamic coupling of the topological structure and the time sequence state signals to generate a power equipment state evolution graph; based on the power equipment state evolution graph and current equipment state verification parameters uploaded on each node in a distributed sensor network deployed in the power equipment, dynamically calibrating basic association strength values between nodes in the topological structure, and applying boundary condition constraints on the calibrated association strength values through the rule constraint items; matching the event association items with the association strength values subjected to the boundary condition constraints, to generate an abnormal reasoning path optimization strategy for a question and answer system request.

[0054] The technical scheme provided by the application has the following beneficial effects:

[0055] The application fully acquires multi-dimensional characteristics of power equipment operation, provides a comprehensive data basis for knowledge graph construction, eliminates semantic conflicts between multi-source data, establishes accurate device entity relationship expression, realizes deep fusion of static topological structure and dynamic time sequence signals, forms device state representation with time and space evolution characteristics, adaptively adjusts graph node relationship based on real-time sensor data, ensures timeliness of state evaluation, converts procedure requirements into graph parameter constraints, ensures that system output conforms to power safety specifications, and outputs disposal schemes conforming to actual fault propagation rules by combining historical event modes and real-time state matching.

[0056] Further, the application also realizes accurate matching of monitoring data and graph nodes through spatial position identification of distributed sensor nodes, calculates deviation of real-time values and historical states, calibrates association strength along a state transition path when exceeding an allowed range, and takes operation critical values defined by procedures as mandatory constraint boundaries, to ensure that the calibrated association strength values conform to safety specifications.

[0057] Moreover, the application realizes upgrading of power equipment state evaluation from static threshold judgment to dynamic path tracing, through a dual mechanism of spatial positioning matching and time sequence backtracking calibration, abnormal diagnosis results reflect real-time state changes and conform to procedure constraints, and the accuracy of complex fault positioning and the compliance of disposal schemes are improved.

[0058] These and other aspects of the application will become more apparent from the following description of the embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0059] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0060] Figure 1 A flow chart of a dynamic power question and answer system optimization method based on a knowledge graph provided by an embodiment of the present application;

[0061] Figure 2 A structural schematic diagram of a dynamic power question and answer system optimization system based on a knowledge graph provided by an embodiment of the present application;

[0062] Figure 3 A structural schematic diagram of a computing device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0063] In order to enable personnel in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application.

[0064] In some of the processes described in the specification and claims of the present application and in the above-described drawings, a plurality of operations appear in a specific order, but it should be clearly understood that these operations can be executed or performed in parallel or in a different order from that in which they appear in this text. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes can include more or fewer operations, and the operations can be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this text are used to distinguish different messages, devices, modules, etc., and do not represent the order of precedence. Also, "first" and "second" are not of different types.

[0065] In the current power system intelligent question and answer technology, the scheme based on static knowledge graph has limitations: the preset graph node relationship cannot adapt to the dynamic evolution characteristics of the device state, the simple threshold matching mechanism of the sensor data and the graph breaks the spatio-temporal correlation between the multi-modal data, resulting in systematic deviation of the abnormal reasoning result and the actual fault propagation path. This defect is due to the neglect of the time-varying characteristics of the power device operating state and the chain reaction law by the static modeling method, so that the disposal strategy output by the question and answer system often lacks the ability to predict the complex fault evolution.

[0066] To solve the above problems, the application provides a dynamic power question and answer system optimization method based on a knowledge graph, which realizes deep fusion of multi-modal data by constructing a device state graph with time-space evolution characteristics. The method first maps the regulation documents, sensor signals and historical events to a unified semantic space to construct a topological structure, then generates an evolution graph reflecting the device state migration path by dynamically coupling the time sequence signals, and calibrates the node association strength based on real-time sensor data, and finally generates an abnormal reasoning path optimization strategy combined with the historical event mode. The scheme breaks through the rigid modeling limitation of the static graph, accurately captures the fault propagation path through the dynamic association strength calibration mechanism, ensures compliance of the disposal scheme by using boundary condition constraints, and effectively solves the problems of abnormal reasoning deviation and insufficient strategy predictability in the prior art.

[0067] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the application.

[0068] Figure 1 A flowchart of a dynamic power question and answer system optimization method based on a knowledge graph provided by the embodiments of the application is shown in Figure 1 The method comprises the following steps.

[0069] Step 101: Real-time acquisition of heterogeneous operation data at different positions of the power equipment, wherein the heterogeneous operation data comprises rule constraint items of regulation documents, time sequence state signals continuously generated by sensors, and event association items in historical maintenance records.

[0070] In step 101, the heterogeneous operation data refers to a collection of operation data of different sources and different formats in the power system. The rule constraint items of the regulation documents represent the device operation specification clauses extracted from the power industry standard documents, such as the provisions "the main transformer oil temperature shall not exceed 70℃" in the "substation operation regulation". The time sequence state signals of the sensors represent the continuous monitoring data collected by the temperature, current, voltage and other sensors deployed on the power equipment at fixed time intervals. The event association items of the historical maintenance records represent the three-tuple information of device fault phenomena, treatment measures and consequences recorded in the past maintenance work orders.

[0071] In the embodiments of the present application, the system collects real-time monitoring data such as temperature and current through the sensor network deployed on power equipment such as transformers and circuit breakers at a preset sampling period, extracts the latest version of the equipment operation procedure document from the power enterprise document management system, and retrieves the historical fault records of the past few years from the maintenance work order database. After the three types of data are standardized in format, they are stored in different partitions of the distributed data warehouse, among which the sensor data is accompanied by device location encoding and timestamp, the procedure document data is labeled with the applicable equipment type, and the maintenance record data is supplemented with fault level labels, providing structured input for subsequent processing.

[0072] For example, the temperature data of the main transformer oil temperature sensor of a certain 220kV substation is collected every 5 minutes, while the system extracts the clause "the normal value of the top oil temperature shall not exceed 85℃" from the "Operation and Maintenance Regulations for Oil-immersed Transformers" from the knowledge base, and associates it with the fault event "abnormal oil temperature leading to light gas action" in the past three years of maintenance records of the station. The three types of data are respectively labeled as time series signal data of equipment number B01, rule clause R01 and event record E01, and transmitted to the processing module through the data bus.

[0073] Step 102: mapping the rule constraint items and the event association items into a unified semantic space, eliminating the cross interference of repetitive semantic units in the semantic space, and constructing a topological structure.

[0074] In step 102, the unified semantic space representation converts different source data into a mathematical space with the same dimensional feature vector. The semantic unit cross interference represents the semantic ambiguity or conflict in the description of the same device property by different data sources. The topological structure represents the graph structure describing the relationship between power equipment entities, with nodes representing equipment entities and edges representing the functional dependency relationship between equipment.

[0075] In the embodiments of the present application, the procedure clause text and the historical event description are converted into feature vectors through natural language processing technology, wherein the procedure clause is decomposed into a vector representation of three dimensions of equipment type, operation action and limitation condition, and the historical event is converted into a vector representation of three dimensions of equipment number, fault type and treatment measure. By calculating the cosine similarity between vectors, the repetitive vector groups describing the same device property are identified, and the representative vector is generated by taking the weighted average value of each repetitive vector group. According to the dependency relationship between the representative vectors, a directed edge is established, such as the "transformer" vector pointing to the "cooler" vector and labeling the "control" relationship type, and finally forming a topological structure with a weighted directed graph structure.

[0076] For example, rule clause R01 is converted to vector [transformer, oil temperature monitoring, ≤ 85℃], event record E01 to vector [B01, oil temperature anomaly, light gas reset]. The calculation finds that both are similar in device type dimension over the threshold, merging to generate a new vector [B01 transformer, oil temperature monitoring anomaly, ≤ 85℃ or light gas action]. The vector establishes a "protection action" relationship edge with the station circuit breaker vector, forming a topology structure containing device nodes such as main transformer, circuit breaker, cooler, etc. and their relationships.

[0077] Step 103: cross-modal dynamic coupling of the topology structure with the time sequence state signal to generate a power equipment state evolution graph.

[0078] In step 103, cross-modal dynamic coupling means fusing the change characteristics of the time sequence signal with the static relationship of the topology structure. The power equipment state evolution graph represents a dynamic graph structure formed by superimposing the time dimension on the basic topology, with edge weights changing over time.

[0079] In the embodiments of the present application, the sensor time sequence data is matched with the corresponding nodes in the topology structure according to the device number, and the feature values at each time point are extracted to form a state snapshot. The difference between the state values at adjacent time points is calculated, and the state change amount is transmitted along the direction of the edges of the topology structure, such as the oil temperature rise amount of the main transformer to the temperature adjustment demand of the associated cooler node. The edge weight is adjusted according to the transmission amount, and when the transmission amount of a plurality of consecutive time points exceeds the threshold, a time stamp is marked on the corresponding edge. Finally, an evolution graph containing the initial topology relationship and the time-varying state transition path is formed.

[0080] For example, the oil temperature of main transformer B01 rises from 65℃ to 78℃ during 08:00-08:30, and the temperature rise difference value every 5 minutes is calculated and transmitted to the cooler node along the topology edge. When the temperature rise difference value at 08:25 exceeds the threshold of 0.5℃ per minute, the "08:25 emergency heat dissipation" event is marked on the edge from B01 to the cooler, forming an evolution subgraph reflecting the temperature anomaly propagation path.

[0081] Step 104: based on the power equipment state evolution graph and the current device state verification parameters uploaded on each node in the distributed sensor network deployed in the power equipment, dynamically calibrate the basic correlation strength values between the nodes in the topology structure, and impose boundary condition constraints on the calibrated correlation strength values through the rule constraint term.

[0082] In step 104, the current device state verification parameter represents the latest uploaded monitoring data with position identification of the sensor. The basic correlation strength value represents the initial setting value of the edge weight in the topology structure. The graph node represents the power device entity and its state attribute (such as transformer, circuit breaker, etc.), and the network node refers to the physical sensor hardware unit, which are associated through the mapping relationship between the sensor data and the graph entity. The boundary condition constraint represents the value range limitation set by the regulation clause on the correlation strength. The rule constraint item is the explicit rule extracted from the regulation document (such as "temperature exceeding threshold must be shut down"), and the boundary condition constraint is the dynamic limitation for the calculation of the correlation strength by applying these rules (such as forcibly cutting off the graph connection of the over-temperature node). The former is the rule source, and the latter is the rule execution form.

[0083] In the embodiment of the present application, the monitoring data uploaded by the sensor in real time is received, and the corresponding device node is located in the evolution graph. The deviation amount of the current monitoring value from the historical state average value of the node is calculated, and when the deviation exceeds the allowed fluctuation range, the source node of the deviation is found along the time sequence marked by the state transition. Starting from the source node, the correlation strength values of the edges on the path are re-calculated in a forward direction in time, and a recursive formula is used to make the strength values decay with the transmission distance. The calculated value is compared with the limit value specified by the regulation clause, and the part exceeding the limit value is truncated to generate the final correlation strength conforming to the safety specification.

[0084] For example, at 08:30, alarm data of B01 oil temperature 82℃ is received, and the deviation from the historical average value 70℃ is calculated. It is found that the 08:25 cooler performance decline event is the source by backtracking along the evolution graph, and the correlation strength value from the cooler node to B01 is re-calculated. According to the upper limit of 85℃ of regulation clause R01, the final correlation strength is limited within the safe range.

[0085] Step 105: matching the event correlation item with the correlation strength value subjected to the boundary condition constraint to generate an abnormal reasoning path optimization strategy for the question and answer system request.

[0086] In step 105, the abnormal reasoning path optimization strategy represents a disposal suggestion sequence generated for a specific abnormal scenario.

[0087] In the embodiment of the present application, the fault propagation mode in the historical event is matched with the correlation strength distribution of the current evolution graph for similarity, and the highly matched historical mode is found as a reference. The current abnormal source node is identified according to the matching result, and the influence range is traced along the correlation strength decay direction. The nodes on the path are sorted according to the strength value, and the devices corresponding to the strength mutation points are processed preferentially. Combined with the operation requirements of each device in the regulation clause, an optimization strategy containing the inspection sequence and disposal measures is generated.

[0088] For example, the E01 event pattern is matched to the current B01 oil temperature anomaly, and the cooler is determined to be the source. Tracing along the atlas finds that the associated circuit breaker node has a strong intensity mutation, generating a treatment strategy of "first check the cooler fan power supply, then test the circuit breaker protection setting value", which is output after verification by the regulation clause.

[0089] The method constructs a dynamic evolution atlas through multi-source data fusion, realizes spatio-temporal correlation analysis of power equipment state, and upgrades abnormal diagnosis from single-point judgment to system-level path tracing. The correlation strength mechanism calibrated based on real-time data accurately reflects the fault propagation law, and the treatment strategy generated in combination with the regulation constraint improves the accuracy and safety of operation and maintenance response, providing reliable decision support for the power question and answer system.

[0090] To solve the accuracy problem of dynamic evaluation of power equipment state, in some embodiments, step 104: the current equipment state verification parameter includes the spatial position identification and real-time monitoring value of the distributed sensor node.

[0091] Based on the power equipment state evolution atlas and the current equipment state verification parameter uploaded to each node in the distributed sensor network deployed in the power equipment, the basic correlation strength value between nodes in the topology structure is dynamically calibrated, and the calibrated correlation strength value is subjected to boundary condition constraint through the rule constraint term, including:

[0092] Step 201: According to the matching result of the spatial position identification and the corresponding entity node in the power equipment state evolution atlas, the real-time monitoring value is calculated with the deviation amount of the historical state transition path of the corresponding entity node.

[0093] In step 201, the spatial position identification is the physical installation position code of the distributed sensor node on the power equipment, which is used to establish a spatial mapping relationship with the atlas node. The matching result refers to the accurate comparison between the spatial position identification (such as device number ST01) of the distributed sensor node and the spatial code of the entity node in the power equipment state evolution atlas, to determine the mapping relationship of the sensor monitoring data to the specific equipment node (such as the main transformer B01 node). The result ensures that the real-time monitoring value can be accurately associated with the target analysis object in the atlas. The real-time monitoring value is the latest equipment operation parameter collected by the sensor. The historical state transition path is formed by the accumulation of state transition markers, that is, the path set generated by aggregating the device state parameter change trajectory at all time points. The deviation amount calculation is a quantitative process of comparing the deviation degree of the current monitoring value from the historical state trend.

[0094] In the embodiment of the present application, the system receives real-time data uploaded by sensors with position encoding, finds a device node with matching position encoding in the evolution map, extracts state values of the node in the past several sampling periods to form a historical trajectory curve, calculates the perpendicular distance between the real-time data point and the trajectory curve as a deviation, and converts the deviation into a standardized score for subsequent judgment.

[0095] Step 202: When the deviation result exceeds the preset allowable fluctuation range of correlation strength, backtrack to the deviation starting node along the state transition path network in the topology structure, and calibrate the basic correlation strength values between nodes on the backtracking path.

[0096] In step 202, the correlation strength allowable fluctuation range is a reasonable state change interval preset according to the device type. The state transition path network is a dynamic subgraph recording the propagation direction of device state changes in the topology structure, which is generated by coupling the time sequence state signal and the topology structure, and the edges represent the transmission direction of state changes (such as abnormal oil temperature from the cooler node to the transformer node). The connection line between nodes is marked with a time stamp and a transmission amount parameter, which is used for backtracking the abnormal propagation path. Backtracking refers to tracking in reverse time along the state transition mark. The deviation starting node is the device node where the state anomaly first occurs. The backtracking path specifically refers to a dependent link that starts from the latest abnormal data node reported by the distributed sensor and backtracks along the correlation strength attenuation direction in the power device state evolution map. The current entity node path specifically refers to a dynamic backtracking path in the correlation strength attenuation direction in the state transition path network, which is essentially different from the static one-way dependent link when the topology structure is constructed. The basic correlation strength value reflects the inherent relationship strength between devices.

[0097] In the embodiment of the present application, when the standardized deviation score exceeds the fluctuation range threshold, the earliest node where the abnormal feature occurs is found as the starting point by inversely searching the time stamp of the state transition mark from the current node. The correlation strength between each pair of nodes on the path is recalculated in forward time from the starting point. A recursive attenuation algorithm is used to make the strength value decrease with the increase of transmission distance, forming a new correlation strength distribution.

[0098] Step 203: Use the operation critical value defined in the rule constraint term as a boundary condition constraint. When the calibrated correlation strength value is greater than the operation critical value, use the operation critical value as the correlation strength value after boundary condition constraint.

[0099] In step 203, the operation critical value is a safe limit value of the device parameter specified in the procedure document. The boundary condition constraint is a mandatory rule for limiting the correlation strength within a safe range.

[0100] In the embodiments of the present application, the recalculated correlation strength value is compared with the parameter limit value of the corresponding device in the regulation clause, and for the strength value exceeding the limit value, it is forced to adjust to the limit value, to ensure that the final output correlation strength reflects the actual state change and meets the safety specification.

[0101] The following is a specific example:

[0102] The oil temperature sensor ST01 of the main transformer B01 of a certain 220kV substation encodes the spatial position ST01 and the real-time temperature value 86℃ at 08:35. The system matches the position coding to the B01 node in the power equipment state evolution graph. By comparing 86℃ with the historical state transition path formed by the past 24 hours of oil temperature data of the B01 node, the deviation amount of the current temperature from the historical average is calculated to be 16℃, which is obtained by subtracting the historical 24-hour average temperature 70℃ from the current temperature. Since the preset correlation strength allows a fluctuation range of ±10℃, the system determines that the deviation result exceeds the threshold. Backtracking along the state transition path network marked in the graph, it is found that the cooler node C01 at 08:25 has an abnormal mark of "30% reduction in heat dissipation efficiency", which is determined as the deviation starting node. Starting from the C01 node, the correlation strength value to the B01 node is recalculated according to the state transmission direction, and the new correlation strength value is calculated to be 1.8 by using the recursive formula, which is obtained by multiplying the base correlation strength 1.2 by the temperature rise influence coefficient 1.5. Comparing the maximum allowed correlation strength 1.5 corresponding to the upper limit of oil temperature 85℃ specified in the regulation clause R01, the calibrated 1.8 is forced to adjust to 1.5 to ensure compliance with the safety specification.

[0103] In the embodiments of the present application, the dynamic correlation of sensor data and graph nodes is realized by accurate spatial position matching, the abnormality recognition sensitivity is improved based on the deviation detection of historical trajectory, the evaluation result is safe and reliable by combining the strength calibration of regulation constraints, a closed-loop dynamic state evaluation mechanism is formed, and the accuracy and safety of power equipment abnormality diagnosis are effectively improved.

[0104] In order to further improve the accuracy of power equipment abnormal state backtracking, in some embodiments, step 202: backtracking to the deviation starting node along the state transition path network in the topology structure, and calibrating the base correlation strength value between each node on the backtracking path, comprises:

[0105] Step 301: detecting the difference direction between the deviation result and the preset correlation strength allowed fluctuation range, and determining the forward or reverse backtracking mode along the state transition path network according to the difference direction.

[0106] In step 301, the difference direction refers to the comparison result direction of the current monitoring value and the preset fluctuation range, and the positive difference indicates exceeding the upper limit, and the reverse difference indicates being lower than the lower limit. For example, when the power question and answer system monitors the time sequence state signal of the oil temperature sensor of the main transformer of a certain substation and shows that the current oil temperature value is 75 degrees Celsius, first, the system queries the preset associated strength fluctuation range of the main transformer node in the power equipment state evolution graph, which is 65 to 70 degrees Celsius, calculates the deviation of the current value 75 and the upper limit 70 of the allowed range as +5 degrees Celsius, since the deviation calculation result is positive and exceeds the allowed range upper limit threshold value 2 degrees Celsius, which is derived from the safety margin coefficient 1.2 defined in the power equipment regulation document multiplied by the sensor measurement error 1.67 degrees Celsius, the system determines that the difference direction is positive difference, and accordingly determines to use the reverse backtracking mode to track the fault source, wherein the definition standard of the positive difference is that the measured value exceeds the upper limit of the allowed range and the deviation is greater than the threshold value 2 degrees Celsius, and the reverse difference is that the measured value is lower than the lower limit of the allowed range and the absolute value of the deviation is greater than 1.5 degrees Celsius, which is derived from the optimal interval of fault warning effectiveness in the historical maintenance record. The backtracking mode is divided into positive backtracking and reverse backtracking, and the positive backtracking finds the source along the state transfer direction, and the reverse backtracking finds the influence range in the reverse state transfer direction.

[0107] In the embodiment of the present application, the system first determines whether the deviation is positive or negative deviation from the allowed range, if it is positive deviation, the reverse backtracking mode is selected to track the fault source, if it is negative deviation, the positive backtracking mode is selected to track the affected equipment, so as to ensure that the backtracking direction is consistent with the actual fault propagation direction.

[0108] Step 302: According to the backtracking mode, starting from the current entity node, the direction of the connection line in the state transition path network is tracked to the deviation starting node, and the time stamp interval length and state transfer amount cumulative value of the current entity node in the backtracking path are recorded.

[0109] In step 302, the time stamp interval length is the time difference between adjacent state transition marks. The state transfer amount cumulative value is the cumulative value of the state change amount of each node along the backtracking path.

[0110] In the embodiment of the present application, starting from the current abnormal node, the precursor node is found along the time sequence of the state transition mark according to the selected backtracking mode, the time interval and state change amount cumulative value between each node are recorded, and a complete backtracking path and its space-time characteristic parameters are formed.

[0111] Step 303: According to the time stamp interval length and the state transfer amount cumulative value, the target associated strength attenuation coefficient between every two adjacent entity nodes in the backtracking path is calculated.

[0112] In step 303, the target correlation strength attenuation coefficient is a weight adjustment parameter calculated according to the space-time feature, reflecting the attenuation degree of state change on the path.

[0113] In the embodiment of the present application, the time interval length and the state transmission quantity cumulative value are input into the attenuation model, the longer the time interval and the smaller the transmission quantity, the greater the attenuation coefficient, and the dynamic attenuation coefficient between each pair of adjacent nodes on the backtracking path is calculated.

[0114] Step 304: On the backtracking path, the target correlation strength attenuation coefficient is multiplied by the basic correlation strength value between the corresponding entity nodes in the direction from the deviation starting node to the current entity node to generate the calibrated correlation strength value.

[0115] In step 304, the calibrated correlation strength value is the dynamic relationship strength adjusted by the space-time feature.

[0116] In the embodiment of the present application, starting from the deviation starting node, the basic correlation strength between each pair of nodes is multiplied by the corresponding attenuation coefficient along the backtracking path to generate the calibrated value reflecting the actual state propagation strength, and the strength update of the entire path is completed.

[0117] The following is a specific example:

[0118] The oil temperature sensor ST01 of the main transformer B01 of a certain 220kV substation transmits a real-time temperature value of 86℃ at 08:35, and the system calculates the deviation of 16℃ between this value and the historical 24-hour average temperature of 70℃, which exceeds the preset allowable fluctuation range of ±10℃. Since the deviation is a positive difference, the system determines to use the reverse backtracking mode to trace back to the fault source. Starting from the B01 node, the connection line in the state transition path network is directed back to backtrace, and it is found that the cooler node C01 at 08:25 has an abnormal label of a 30% decrease in heat dissipation efficiency, and the time interval length from B01 to C01 is recorded as 10 minutes and the state transmission quantity cumulative value is 16℃. According to the inverse relationship between the time interval and the transmission quantity, the attenuation coefficient calculation formula α=1 / (1+Δt×ΔQ) is used, where Δt is the time interval of 10 minutes and ΔQ is the transmission quantity of 16℃, and the target correlation strength attenuation coefficient between B01 and C01 is calculated as 0.8. Starting from the deviation starting node C01, the basic correlation strength value from C01 to B01 is multiplied by the attenuation coefficient 0.8 to generate the calibrated correlation strength value 0.96.

[0119] In the embodiment of the present application, the difference direction determination ensures that the backtracking path conforms to the fault propagation rule, the attenuation coefficient calculation based on the space-time feature accurately reflects the state change attenuation characteristics, and the dynamically calibrated correlation strength value more truly presents the state influence relationship between devices, providing accurate quantitative basis for abnormal diagnosis.

[0120] To further improve the accuracy of the correlation strength decay coefficient calculation, in some embodiments, step 303: calculating the target correlation strength decay coefficient between each two adjacent entity nodes in the backtracking path according to the timestamp interval length and the state transition amount cumulative value, comprises:

[0121] Step 401: span comparison between the timestamp interval length and the preset reference time window to generate a time span influence factor.

[0122] In step 401, the reference time window is the standard state transition time preset according to the device type. The time span influence factor reflects the deviation degree of the actual transition time from the standard time.

[0123] In the embodiments of the present application, the system divides the actual recorded time interval between adjacent nodes on the backtracking path by the standard state transition time window of the device of this type, to generate a time span influence factor representing time efficiency. The longer the time interval, the smaller the factor value.

[0124] Step 402: amplitude comparison between the state transition amount cumulative value and the preset reference transition amount to generate a transition amount influence factor.

[0125] In step 402, the reference transition amount is the standard state change amount when the device is normally running. The transition amount influence factor reflects the proportional relationship between the actual transition amount and the standard amount.

[0126] In the embodiments of the present application, the actual cumulative transition amount is divided by the standard transition amount of the device in the reference state to generate a transition amount influence factor representing the state change amplitude. The larger the transition amount, the larger the factor value.

[0127] Step 403: generating an initial correlation strength decay coefficient according to the inverse proportional relationship between the time span influence factor and the transition amount influence factor.

[0128] In step 403, the inverse proportional relationship refers to the inverse action mechanism of the time span influence factor and the transition amount influence factor on the decay coefficient, i.e. the longer the time interval (the smaller the factor value) or the smaller the transition amount (the smaller the factor value) will cause the decay coefficient to increase. This relationship is derived from the physical law of power equipment fault propagation - state change naturally decays with time extension and path loss in the transition process. Through analysis of the quantitative correlation between time and transition amount on the final impact degree in historical fault events, it is ensured that the calculation model conforms to the actual physical process. The initial correlation strength decay coefficient is a preliminary adjustment parameter generated according to the space-time characteristics.

[0129] In the embodiment of the present application, the initial attenuation coefficient is obtained by dividing the time span influence factor by the transfer amount influence factor. The time factor proportion is not less than 60%, which ensures that the time factor plays a dominant role in the attenuation calculation.

[0130] Step 404: When the value of the initial correlation strength attenuation coefficient exceeds the preset attenuation allowed range, the upper and lower limit values of the attenuation allowed range are used as the forced constraint boundary, and the initial correlation strength attenuation coefficient is subjected to boundary truncation processing to generate a target correlation strength attenuation coefficient.

[0131] In step 404, the attenuation allowed range is a coefficient boundary preset according to the safe operation requirement of the device. The boundary truncation processing is a process of forcibly adjusting the coefficient exceeding the range to the boundary value.

[0132] In the embodiment of the present application, when the initial attenuation coefficient is less than the lower limit, it is adjusted to the lower limit value, and when it is greater than the upper limit, it is adjusted to the upper limit value, which ensures that the final coefficient reflects the actual state and meets the safety constraints.

[0133] The following is a specific example:

[0134] In the state transfer analysis between the main transformer B01 and the cooler C01 of a certain 220kV substation, the system obtains that the time stamp interval length between the two is 10 minutes, the reference time window of this device type is preset to be 8 minutes, and the time span influence factor 1.25 is obtained by dividing the actual interval 10 minutes by the reference 8 minutes. At the same time, the state transfer amount cumulative value is recorded as 16℃, the reference transfer amount of the device is preset to be 12℃, and the transfer amount influence factor 1.33 is obtained by dividing the actual transfer amount 16℃ by the reference 12℃. According to the inverse ratio relationship formula α = time span influence factor / transfer amount influence factor × 0.6 + 0.4 × time span influence factor, the time factor weight is taken as 60%, and the initial correlation strength attenuation coefficient is calculated as 1.25 / 1.33 × 0.6 + 0.4 × 1.25 = 1.06. The system presets the attenuation allowed range to be 0.8 to 1.2, and since 1.06 is within the allowed range, the final target correlation strength attenuation coefficient remains unchanged at 1.06.

[0135] In the embodiment of the present application, the dual consideration of time and space characteristics ensures that the attenuation coefficient accurately reflects the state transfer characteristics, and the boundary constraint mechanism guarantees the rationality of the calculation result, so that the calibrated correlation strength meets the actual operation state of the device and meets the safety specification requirements.

[0136] In order to further improve the accuracy of the abnormal reasoning path, in some embodiments, step 105: the event correlation item is matched with the correlation strength value subjected to boundary condition constraint to generate an abnormal reasoning path optimization strategy for the question and answer system request, which includes:

[0137] Step 501: Extract a feature parameter set of a historical abnormal event from the event correlation item, the feature parameter set containing an event trigger node identifier, a correlation strength mutation threshold, and an influence path length.

[0138] In step 501, the historical abnormal event is a recorded failure event extracted from historical maintenance records, and an event template library containing a failure node, a strength mutation feature, and an influence range is formed through structured processing. The feature parameter set is a typical failure feature extracted from the historical event. The event trigger node identifier is the encoding of the first-occurrence device of the failure. The correlation strength mutation threshold is the critical change value of the node relationship strength when the failure occurs. The influence path length is the maximum device range involved in the failure propagation.

[0139] In the embodiment of the application, the system screens historical failure cases matching the current device type from the maintenance record library, extracts the first-occurrence device number, the correlation strength jump amplitude, and the number of affected devices in each case, and forms a standardized feature template set.

[0140] Step 502: Search for a key entity node matching the event trigger node identifier in the power device state evolution graph.

[0141] In step 502, the key entity node is a candidate device node in the evolution graph that matches the historical failure feature.

[0142] In the embodiment of the application, the device number in the historical failure feature is accurately matched with the node identifier in the evolution graph to locate the candidate device node that may reproduce the same type of failure, while considering the similarity of device type and topological position.

[0143] Step 503: When the correlation strength value between the key entity node and the adjacent node after the boundary condition constraint reaches the correlation strength mutation threshold, mark the key entity node as an abnormal trigger source.

[0144] In step 503, the abnormal trigger source is the failure starting node determined through double verification, which is achieved by comparing the current correlation strength value with the historical mutation threshold.

[0145] In the embodiment of the application, it is checked whether the calibrated correlation strength value between the candidate node and the adjacent node reaches the mutation threshold of the historical same type of failure, and when a plurality of adjacent relationships successively satisfy the threshold condition, the node is determined as the abnormal trigger source.

[0146] Step 504: Reverse trace the dependency path along the direction of correlation strength attenuation in the power device state evolution graph starting from the abnormal trigger source.

[0147] In step 504, the correlation strength attenuation direction reflects the fault influence diffusion direction. "Backtracking to the deviation starting node" is used to calibrate the correlation strength, and "reverse tracking of the dependency path" is used to locate the abnormal source. The operation objects are the same, but the purposes are different. Both are technical actions of reverse tracking along the path. The correlation between the tracking dependency path and the backtracking path: the tracking dependency path specifically refers to the process of reverse searching the fault propagation path from the abnormal trigger source along the correlation strength attenuation direction, while the backtracking path is the path of reverse searching the deviation source from the current abnormal node. The two have the same technical essence but different application stages. The tracking dependency path is used for final abnormal positioning, and the backtracking path is used for data calibration in the intermediate process. Moreover, the generation of the tracking dependency path directly depends on the correlation strength attenuation rule established in the backtracking path stage.

[0148] In the embodiment of the present application, starting from the trigger source node, the affected device nodes are gradually searched along the direction of decreasing correlation strength, and the strength change of each node passed is recorded to form a complete fault propagation path.

[0149] Step 505: Sort all nodes in the tracking dependency path according to the spatial attenuation mode to generate an ordered path chain from the abnormal trigger source to the set threshold node, and generate an abnormal reasoning path optimization strategy according to the ordered path chain and the question and answer system request.

[0150] In step 505, the spatial attenuation mode describes the numerical decrease rule of the correlation strength value when it propagates outward from the trigger node in the abnormal event. This mode is obtained by analyzing historical maintenance record fault event data statistics. Specifically, it is characterized by taking the fault source as the center, and the correlation strength value decreases in inverse proportion to the square of the physical distance between devices, and superimposes the influence factor of the power network topology to form the final spatial attenuation coefficient curve. The ordered path chain is a sequence of key devices sorted by fault influence degree. The question and answer system request refers to the abnormal diagnosis requirements input by the power operation and maintenance personnel through the human-computer interaction interface. Its sources include two categories: one is the real-time alarm request automatically triggered by the system when the monitoring data exceeds the threshold; the other is the query request actively submitted by the human, such as the diagnosis requirements proposed by the operation and maintenance personnel for the abnormal state of a specific device. The request content usually includes target device identification, abnormal phenomenon description and required diagnosis depth parameter. The system matches these structured request parameters with the abnormal reasoning path to generate an optimization strategy that meets the actual requirements.

[0151] In the embodiment of the present application, the path nodes tracked are sorted in descending order of correlation strength value, and devices with obvious strength mutations are preferentially processed. Combined with the disposal resource limit in the question and answer request, the optimal inspection and maintenance sequence is generated.

[0152] The following is a specific example:

[0153] A 220 kV substation system detects that the main transformer B01 node oil temperature reaches 88℃, exceeding the 85℃ limit specified in the procedure R01, triggering an automatic question and answer request. The system extracts the feature parameters of event record E01 from the historical event library, obtains the cooler node C01 as the typical trigger source, the correlation strength mutation threshold is 1.8, and the influence path length is 3. In the evolution map, the correlation strength value between the current C01 node and the B01 node is 2.1, the calculation formula is 18℃, the current oil temperature deviation value divided by the reference temperature difference 10℃, 1.8, and after superimposing the cooling efficiency decline coefficient 0.3, 2.1 is obtained. Since the value exceeds the historical threshold 1.8, it is determined that C01 is an abnormal trigger source. Tracing the path in the direction of strength attenuation, it is found that the strength value from C01 to B01 decreases from 2.1 to 1.5, and continues to trace to the cooling water pump P01 node, the strength value is 1.2, forming a complete path P01 to C01 to B01. According to the strength value sorting, the ordered path chain C01-B01-P01 is generated, combined with the requirement of "prior processing of main fault points" in the question and answer request, the inspection order is optimized as "firstly processing C01 heat dissipation system, secondly reviewing B01 oil temperature protection, and finally checking P01 water supply pipeline". The strategy is verified to be consistent with the provision of "oil temperature anomaly priority checking cooling system" in the procedure R01, and is output to the operation and maintenance terminal as the final optimization strategy.

[0154] In the embodiments of the present application, the dual matching of historical fault features and real-time state ensures accurate positioning of abnormal sources, the path tracing based on the strength attenuation law truly reflects the fault propagation process, and the generated optimization strategy not only meets the actual change characteristics of the device state, but also meets the requirements of the operation and maintenance procedure, thereby improving the decision reliability of the power question and answer system.

[0155] In order to further improve the accuracy of the knowledge representation of the power equipment, in some embodiments, step 102: the rule constraint item and the event association item are mapped into a unified semantic space, the cross interference of repeated semantic units in the semantic space is eliminated, and a topology structure is constructed, including:

[0156] Step 601: converting the text description unit in the rule constraint item and the historical event identification unit in the event association item into a semantic vector unit.

[0157] In step 601, the text description unit in the rule constraint item is derived from the semantic analysis of the power system regulation document. By extracting the content of the clauses in the document that define the device operation specification, each clause is converted into a structured text segment containing three elements: the operation subject, the action type, and the constraint condition, forming the text description unit. The historical event identification unit in the event association item is derived from the structured processing of the maintenance records. By identifying the device number, fault type, and treatment measure key fields in the records, each maintenance event is encoded as a three-tuple identification of "device number-fault code-measure code", forming the historical event identification unit. The semantic vector unit is a feature vector representation in a multi-dimensional space.

[0158] In the embodiments of the present application, the system decomposes the regulation clauses into three-tuple segments according to the subject-predicate-object syntax structure, encodes the historical events into feature labels according to the device-fault-measure structure, and converts the two types of data into vector representations of the same dimension through a semantic model, ensuring that the data from different sources can be compared.

[0159] Step 602: Identify the repetitive semantic unit group in the semantic vector that describes the same power device entity or the same operation behavior, and perform semantic superposition on all repetitive semantic units in each group of repetitive semantic unit groups to generate a semantic vector expression unit.

[0160] In step 602, the repetitive semantic unit group is a collection of multi-source data describing the same device attribute or operation behavior. Semantic superposition is a weighted fusion process of similar vectors. The semantic vector expression unit is the representative vector after fusion.

[0161] In the embodiments of the present application, the cosine similarity between all semantic vectors is calculated, and the weighted average processing is performed on the vector groups with similarity exceeding the threshold value. The weight is determined according to the authority and timeliness of the data source, and the unified vector expression representing the semantic group is generated.

[0162] Step 603: Determine the subordinate relationship between the semantic vector expression units, and establish a one-way dependent link between the power device entities in the unified semantic space according to the subordinate relationship.

[0163] In step 603, the subordinate relationship is the logical relationship such as control and protection between device entities. The one-way dependent link is an arrowed edge in the directed graph representing the master-slave relationship.

[0164] In the embodiments of the present application, the correlation of the semantic vector expression units in the device type, functional attribute, and other dimensions is analyzed, and a directed connection is established from the dominant device vector to the controlled device vector according to the device hierarchical relationship defined by the regulation. The connection types include control, protection, linkage, etc.

[0165] Step 604: Form a topology structure by hierarchical expansion of the unidirectional dependency link.

[0166] In step 604, the hierarchical expansion is a process of recursively traversing to establish a complete device relationship network.

[0167] In the embodiment of the present application, starting from the core device node, the secondary devices and the associated relationships are added along the established unidirectional dependency link, until all device nodes are included in the network, forming a complete device topology structure.

[0168] The following is a specific example:

[0169] In a certain 220kV substation system, the processing module receives the rule clause R01 "The oil temperature of the main transformer shall not exceed 85℃" and the event record E01 "B01 main transformer oil temperature anomaly causes light gas action". First, R01 is decomposed into three dimensions of device type "transformer", monitoring parameter "oil temperature" and constraint condition "≤85℃", converted into semantic vector [0.8, 0.9, 0.7], wherein the weight of each dimension is calculated by TF-IDF algorithm, the formula is w=term frequency × inverse document frequency. At the same time, E01 is converted into vector [0.9, 0.8, 0.6], and device number B01 is mapped to 0.9 through hash algorithm. The cosine similarity of the two vectors in the device type dimension is 0.95, which exceeds the preset threshold 0.85, and it is determined that it describes the same device property. The two groups of vectors are weighted and fused, the weight coefficient is taken as rule 0.6 and event 0.4, and a new vector [0.84, 0.86, 0.66] is generated, which represents the oil temperature monitoring abnormal feature of B01 main transformer. According to the "oil temperature anomaly triggers protection" clause in the regulation, the vector is established with the circuit breaker vector [0.7, 0.5, 0.8] in the station "protection action" relationship edge, and the edge weight is set to 0.75, which is obtained by multiplying the protection action emergency degree coefficient 0.5 and the device association degree coefficient 1.5. Continue to add the cooler node [0.6, 0.7, 0.5] according to the "cooler controls oil temperature" rule and establish the "control" relationship edge, the weight is 0.9, and finally form a topology structure containing B01 main transformer, circuit breaker, cooler three nodes and their relationship, which provides a basic framework for subsequent state evolution analysis.

[0170] In the embodiment of the present application, the semantic fusion of multi-source data eliminates the ambiguity of knowledge representation, and the accurate topology network is constructed based on the device relationship defined by the regulation, which provides a complete structure and clear semantic device relationship basis for subsequent state evolution analysis, and improves the accuracy and availability of power knowledge representation.

[0171] To further improve the dynamic characterization capability of power equipment state evolution, in some embodiments, step 103: the cross-modal dynamic coupling of the topology structure and the time sequence state signal generates a power equipment state evolution graph, including:

[0172] Step 701: Extract the power equipment state parameter set of each collection time point in the time sequence state signal.

[0173] In step 701, the power equipment state parameter set is the multi-dimensional operation data collected by the sensor at a single sampling time, including temperature, current, voltage and other key parameters.

[0174] In the embodiments of the present application, the system obtains equipment operation data from each sensor node according to a preset sampling period, time aligns and formats the multi-source monitoring data of each time point, and forms a set of device state snapshots with timestamps.

[0175] Step 702: Bind the power equipment state parameter set to the corresponding power equipment entity node in the topology structure.

[0176] In step 702, binding refers to the process of establishing a mapping relationship between real-time monitoring data and topology nodes.

[0177] In the embodiments of the present application, each state parameter set is associated with the corresponding equipment entity node in the topology structure through device location coding and type matching, as the instantaneous state attribute of the node.

[0178] Step 703: At consecutive collection time points, calculate the state transfer amount between adjacent nodes along the connection line direction of the topology structure according to the device state parameter set.

[0179] In step 703, continuity judgment refers to the time series data points generated by the sensor at a fixed sampling interval (such as 1 minute), which is determined to be continuous when there is no data missing between two time points and the interval is equal to the preset sampling period. For example, if the sampling period is 5 minutes, "08:00, 08:05, 08:10" are consecutive time points, and "08:00, 08:07" are not consecutive. Adjacent only refers to nodes directly connected by connection lines in the topology structure, and is independent of the physical location of the sensor. For example, the topology has a connection line "transformer to circuit breaker", which is determined to be adjacent. The state transfer amount is a quantitative value of the state change between adjacent equipment nodes, reflecting the propagation strength of faults or abnormalities.

[0180] In the embodiments of the present application, the difference value of the state parameters of the adjacent nodes at the current time point is calculated, the difference value is transmitted along the topology connection line direction, and the transfer amount is weighted according to the connection type.

[0181] Step 704: When the state transition amount exceeds a preset transition threshold, a state transition label from the current collection time point to the next collection time point is added in the topology structure.

[0182] In step 704, the transition threshold is a state change sensitivity preset according to the device type. The collection time point represents the time when the sensor actually records data (e.g., 08:00). The current time point represents the collection time point being processed (e.g., 08:00). The next time point represents the subsequent collection time point next to the current time point (e.g., 08:05). The state transition label is a space-time label recording the abnormal propagation characteristics.

[0183] In the embodiments of the present application, when the transition amount exceeds the device bearing capacity threshold, a triple label containing the time stamp, the transition direction and the amount value is added on the corresponding connection line.

[0184] Step 705: By aggregating the state transition labels of all collection time points, a state transition path network with time stamps is formed in the topology structure and the basic correlation strength value between adjacent nodes is generated.

[0185] In the embodiments of the present application, the transition labels of all time points are aggregated to form a propagation network with time dimension, and the mean value of the transition amount on each connection line is counted as the basic correlation strength.

[0186] Step 706: According to the state transition path network and the basic correlation strength value, a power device state evolution map is constituted.

[0187] In the embodiments of the present application, the basic topology relationship and the time sequence transition network are superimposed to form a multi-layer graph structure with nodes representing device entities and edges containing static relationships and dynamic propagation characteristics.

[0188] The following is a specific example:

[0189] In a 220 kV substation system, the oil temperature monitoring data of the main transformer B01 node during 08:00-08:30 is collected at 5-minute intervals. At 08:00, the 65℃ is collected and bound to the B01 node of the topology structure. At 08:05, the temperature rise of 2℃ is calculated, and the control edge is transmitted from B01 to the cooler C01. The transmission amount calculation formula is ΔT×Kc, where ΔT is the temperature difference of 2℃, Kc is the cooling system adjustment coefficient of 0.8, and the transmission amount is 1.6, which is lower than the preset threshold value of 3. When 78℃ is collected at 08:25, the temperature rise difference of 7℃ is calculated, and the transmission amount of 7×0.8=5.6 exceeds the threshold value of 3. The state transition mark of the time stamp 08:25, the transmission direction B01 to C01, and the transmission amount 5.6 is added to the edge of B01 to C01. When 82℃ is collected at 08:30, the transmission amount of 6.4 also exceeds the threshold value, and the corresponding mark is added. The system aggregates all the transition marks of the time points, calculates the average transmission amount of the B01 to C01 edge as 5.6+6.4 / 2=6.0, and takes it as the basic correlation strength value between the two nodes. In the finally generated state evolution graph, the B01 node contains time series temperature data, the B01 to C01 edge contains the control relationship type and the average transmission strength of 6.0, and the dynamic propagation process of the oil temperature abnormality from the main transformer to the cooling system is completely recorded.

[0190] In the embodiments of the present application, through the dynamic coupling of time series data and topology structure, the complete characterization of the state of the power equipment from the static relationship to the dynamic evolution is realized, so that the system can intuitively present the abnormal propagation path and the key influencing link, and provide a visual analysis basis for accurate diagnosis and disposal decision.

[0191] Figure 2 The structure schematic diagram of the dynamic power question and answer system optimization system based on the knowledge graph provided in the embodiments of the present application is as shown in Figure 2 The system comprises:

[0192] The acquisition module 21 is used for collecting the heterogeneous operation data at different positions of the power equipment in real time. The heterogeneous operation data comprises a rule constraint item of a procedure document, a time series state signal continuously generated by a sensor, and an event association item in a historical maintenance record.

[0193] The construction module 22 is used for mapping the rule constraint item and the event association item into a unified semantic space, eliminating the cross interference of repetitive semantic units in the semantic space, and constructing a topology structure.

[0194] The generation module 23 is used for performing cross-modal dynamic coupling between the topology structure and the time series state signal, and generating a power equipment state evolution graph.

[0195] The calibration module 24 is configured to calibrate the basic correlation strength values between nodes in the power equipment state evolution graph based on the current equipment state verification parameters uploaded on each node in the distributed sensor network deployed in the power equipment, and apply boundary condition constraints on the calibrated correlation strength values through the rule constraint term.

[0196] The matching module 25 is configured to match the event correlation term with the correlation strength values subjected to the boundary condition constraints, and generate an abnormal reasoning path optimization strategy for the question and answer system request.

[0197] Figure 2 The dynamic power question and answer system optimization system based on the knowledge graph can perform Figure 1 The implementation principle and technical effects of the dynamic power question and answer system optimization method based on the knowledge graph are not described again. The specific operation manner of each module and unit of the dynamic power question and answer system optimization system based on the knowledge graph in the above embodiment has been described in detail in the embodiment related to the method, and will not be described in detail here.

[0198] In one possible design, Figure 2 The dynamic power question and answer system optimization system based on the knowledge graph can be implemented as a computing device, such as a server. Figure 3 As shown in the figure, the computing device can include a storage component 31 and a processing component 32.

[0199] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32.

[0200] The processing component 32 performs the above Figure 1 The dynamic power question and answer system optimization method based on the knowledge graph of the embodiment.

[0201] The processing component 32 can include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component can also be one or more Application-Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Process Device (DSPD), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components, for executing the above method.

[0202] The storage component 31 is configured to store various types of data to support the operation of the terminal. The storage component can be implemented by any type of volatile or non-volatile storage devices or a combination thereof, such as Static Random-Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0203] Of course, the computing device can also include other components, such as an input / output interface, a display component, a communication component, etc.

[0204] The input / output interface provides an interface between the processing component and peripheral interface modules, which can be output devices, input devices, etc.

[0205] The communication component is configured to facilitate wired or wireless communication between the computing device and other devices, etc.

[0206] The computing device can be a physical device or an elastic computing host provided by a cloud computing platform, and the computing device can refer to a cloud server, and the processing component, the storage component, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0207] The embodiment of the application further provides a computer storage medium, which stores a computer program, and the computer program can realize the above method when being executed by a computer. Figure 1 The embodiment of the application further provides a knowledge graph-based dynamic power question and answer system optimization method.

[0208] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, and will not be described here.

[0209] The device embodiments described above are only schematic, and the units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiment scheme. Those skilled in the art can understand and implement without creative labor.

[0210] From the above description of the embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software and necessary general hardware platforms, and of course, can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of software products, which can be stored in a computer readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and include a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0211] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the application, and not to limit them; although the application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the application.

Claims

1. A method for optimizing a dynamic power question-answering system based on a knowledge graph, characterized in that, The method comprises the following steps: Real-time acquisition of heterogeneous operation data at different positions of the power equipment, wherein the heterogeneous operation data includes rule constraint items of procedure documents, time-series state signals continuously generated by sensors, and event association items in historical maintenance records; Mapping the rule constraint items and the event association items into a unified semantic space, eliminating cross interference of repetitive semantic units in the semantic space, and constructing a topology structure; Cross-modal dynamic coupling of the topology structure and the time-series state signals to generate a power equipment state evolution graph; Dynamic calibration of basic association strength values between nodes in the topology structure based on the power equipment state evolution graph and current equipment state verification parameters uploaded on each node in the distributed sensor network deployed in the power equipment, and imposing boundary condition constraints on the calibrated association strength values through the rule constraint items; Matching the event association items with the association strength values subjected to the boundary condition constraints to generate an abnormal reasoning path optimization strategy for a question and answer system request; The current equipment state verification parameters include spatial position identifiers and real-time monitoring values of the distributed sensor nodes; The dynamic calibration of the basic association strength values between nodes in the topology structure based on the power equipment state evolution graph and the current equipment state verification parameters uploaded on each node in the distributed sensor network deployed in the power equipment, and the imposition of boundary condition constraints on the calibrated association strength values through the rule constraint items, comprise: According to the matching result of the spatial position identifier and the corresponding entity node in the power equipment state evolution graph, calculating the deviation amount of the real-time monitoring value and the historical state transition path of the corresponding entity node; When the deviation result exceeds the preset association strength allowable fluctuation range, backtracking to the deviation starting node along the state transition path network in the topology structure to calibrate the basic association strength values between nodes on the backtracking path; Taking the operation critical value defined in the rule constraint item as the boundary condition constraint, and taking the operation critical value as the association strength value subjected to the boundary condition constraint when the calibrated association strength value is greater than the operation critical value; The backtracking to the deviation starting node along the state transition path network in the topology structure to calibrate the basic association strength values between nodes on the backtracking path, comprises: Detecting the difference direction between the deviation result and the preset association strength allowable fluctuation range, and determining the backtracking mode along the forward or reverse direction of the state transition path network according to the difference direction; According to the backtracking mode, backtracking to the deviation starting node along the direction of the connection line in the state transition path network from the current entity node, and recording the time stamp interval length and state transfer amount cumulative value of the current entity node on the backtracking path; According to the time stamp interval length and the state transfer amount cumulative value, calculating the target association strength decay coefficient between every two adjacent entity nodes on the backtracking path; On the backtracking path, the target correlation strength attenuation coefficient is multiplied with the basic correlation strength value between corresponding entity nodes in the direction from the deviation starting node to the current entity node to generate a calibrated correlation strength value.

2. The method of claim 1, wherein, The calculation of the target correlation strength attenuation coefficient between every two adjacent entity nodes in the backtracking path according to the timestamp interval length and the state transfer amount cumulative value comprises: span comparison between the timestamp interval length and a preset reference time window to generate a time span influence factor; amplitude comparison between the state transfer amount cumulative value and a preset reference transfer amount to generate a transfer amount influence factor; generation of an initial correlation strength attenuation coefficient according to the inverse ratio relationship between the time span influence factor and the transfer amount influence factor; when the numerical value of the initial correlation strength attenuation coefficient exceeds a preset attenuation allowed range, the upper and lower limit values of the attenuation allowed range are taken as forced constraint boundaries to perform boundary truncation processing on the initial correlation strength attenuation coefficient to generate a target correlation strength attenuation coefficient.

3. The method of claim 1, wherein, The matching of the event correlation item with the correlation strength value subjected to boundary condition constraint to generate an abnormal reasoning path optimization strategy for a question and answer system request comprises: extracting a feature parameter set of a historical abnormal event from the event correlation item, the feature parameter set including an event trigger node identifier, a correlation strength mutation threshold value and an influence path length; searching for a key entity node matching the event trigger node identifier in the power equipment state evolution graph; when the corresponding correlation strength value subjected to boundary condition constraint between the key entity node and an adjacent node reaches the correlation strength mutation threshold value, marking the key entity node as an abnormal trigger source; reversely tracing a dependence path from the abnormal trigger source in the correlation strength attenuation direction in the power equipment state evolution graph; sorting all nodes in the traced dependence path according to a spatial attenuation mode to generate an ordered path chain from the abnormal trigger source to a set threshold value node, and generating an abnormal reasoning path optimization strategy according to the ordered path chain and a question and answer system request.

4. The method of claim 1, wherein, The mapping of the rule constraint item and the event correlation item into a unified semantic space, elimination of cross interference of repetitive semantic units in the semantic space, construction of a topological structure, comprises: converting a text description unit in the rule constraint item and a historical event identifier unit in the event correlation item into a semantic vector unit; identifying a repetitive semantic unit group for describing a same power equipment entity or a same operation behavior in the semantic vector, performing semantic superposition on all repetitive semantic units in each repetitive semantic unit group to generate a semantic vector expression unit; determining a subordination relationship between the semantic vector expression units, and establishing a unidirectional dependence link between power equipment entities in the unified semantic space according to the subordination relationship; forming a topological structure through hierarchical expansion of the unidirectional dependence link.

5. The method of claim 1, wherein, The cross-modal dynamic coupling of the topological structure and the time sequence state signal to generate a power equipment state evolution graph comprises: extracting a set of power equipment state parameters at each collection time point in the time sequence state signal; binding the set of power equipment state parameters with the corresponding power equipment entity node in the topology structure; calculating the state transfer amount between adjacent nodes along the connection line direction of the topology structure according to the set of device state parameters at consecutive collection time points; when the state transfer amount exceeds a preset transfer threshold, adding a state transition mark from the current collection time point to the next collection time point in the topology structure; forming a state transition path network with a time stamp in the topology structure by aggregating the state transition marks of all collection time points, and generating a basic correlation strength value between adjacent nodes; constructing a power equipment state evolution graph according to the state transition path network and the basic correlation strength value.

6. A knowledge graph based dynamic power question answering system optimization system, characterized in that, It comprises: a collection module for collecting heterogeneous operation data at different positions of the power equipment in real time, wherein the heterogeneous operation data includes rule constraint items of procedure documents, time sequence state signals continuously generated by sensors, and event correlation items in historical maintenance records; a construction module for mapping the rule constraint items and the event correlation items into a unified semantic space, eliminating the cross interference of repetitive semantic units in the semantic space, and constructing a topology structure; a generation module for cross-modal dynamic coupling of the topology structure and the time sequence state signal to generate a power equipment state evolution graph; a calibration module for dynamically calibrating the basic correlation strength value between nodes in the power equipment state evolution graph based on the current device state verification parameters uploaded on each node of the distributed sensor network deployed in the power equipment, and imposing boundary condition constraints on the calibrated correlation strength value through the rule constraint items; a matching module for matching the event correlation items with the correlation strength value subjected to boundary condition constraints to generate an abnormal reasoning path optimization strategy for the question and answer system request; the current device state verification parameters include spatial position identifiers and real-time monitoring values of the distributed sensor nodes; based on the power equipment state evolution graph and the current device state verification parameters uploaded on each node of the distributed sensor network deployed in the power equipment, dynamically calibrating the basic correlation strength value between nodes in the topology structure, and imposing boundary condition constraints on the calibrated correlation strength value, comprising: calculating the deviation amount of the real-time monitoring value from the historical state transition path of the corresponding entity node according to the matching result of the spatial position identifier and the corresponding entity node in the power equipment state evolution graph; when the deviation result exceeds the preset correlation strength allowable fluctuation range, backtrack to the deviation starting node along the state transition path network in the topology structure to calibrate the basic correlation strength value between nodes on the backtracking path; taking the operation critical value defined in the rule constraint item as the boundary condition constraint, and taking the operation critical value as the correlation strength value subjected to boundary condition constraints when the calibrated correlation strength value is greater than the operation critical value; The state transition path network in the topology is traced back to the deviation starting node, and the basic correlation strength values between nodes on the backtracking path are calibrated, including: detecting the difference direction between the deviation result and the preset correlation strength allowable fluctuation range, and determining the forward or reverse backtracking mode along the state transition path network according to the difference direction; according to the backtracking mode, starting from the current entity node, tracing back to the deviation starting node along the direction of the connection line in the state transition path network, and recording the time stamp interval length and state transfer amount cumulative value of the current entity node in the backtracking path; according to the time stamp interval length and the state transfer amount cumulative value, calculating the target correlation strength attenuation coefficient between each two adjacent entity nodes in the backtracking path; on the backtracking path, multiplying the target correlation strength attenuation coefficient and the basic correlation strength value between the corresponding entity nodes in the direction from the deviation starting node to the current entity node to generate the calibrated correlation strength value.

7. A computing device, comprising: comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to realize the optimization method of the dynamic power question and answer system based on the knowledge graph according to any one of claims 1-5.

8. A computer storage medium, characterized in that, a computer program is stored, and the computer program is executed by a computer to realize the optimization method of the dynamic power question and answer system based on the knowledge graph according to any one of claims 1-5.

Images