A method and system for locating power equipment faults based on dynamic knowledge graph

Abstract

The present invention discloses a method and system for locating power equipment faults based on a dynamic knowledge graph, which relates to the technical field of power equipment fault analysis. The present invention determines the monitoring target of the power equipment and collects operating data of the target power equipment; constructs a large parameter model and a small model adversarial learning framework, and processes the output data of the large parameter model and the small model data; performs feature analysis on the target power equipment data through the small model, and performs parameter prediction on the target power equipment data in combination with the large parameter model to determine the fault analysis data of the target power equipment; constructs an equipment diagnosis knowledge graph, performs node analysis on the fault analysis data of the target equipment, determines the target power equipment fault data inference path, and outputs the target power equipment fault data; the present invention makes up for the manual dependence of existing equipment diagnosis; improves the insufficient utilization of multi-source and multi-modal data, and improves the detection performance of equipment faults.

Description

Technical Field

[0001] The present invention relates to the field of power equipment fault analysis, and specifically to a power equipment fault location method and system based on a dynamic knowledge graph. Background Art

[0002] Power equipment health status diagnosis is the core of power grid asset management and is crucial to the safe and stable operation of the power grid. According to statistics, power equipment failure itself is the main cause of power grid accidents. Power equipment health status diagnosis can be divided into single-state diagnosis and multi-state root cause analysis. Among them, single-state diagnosis is mainly based on regulations and uses threshold analysis, trend analysis, phase comparison, etc. for judgment; multi-state root cause analysis mainly relies on expert experience combined with regulations to make judgments and find the root causes of equipment defects in a timely and accurate manner. However, its manual analysis is highly complex and difficult, resulting in the inability to guarantee the accuracy of the diagnostic results. In addition, high-level expert experience often requires a long time to accumulate and is difficult to pass on. Therefore, there is an urgent need to solve the technical defects of existing equipment diagnosis relying on manual experience, insufficient use of multi-source and multi-modal data, and the inability to guarantee diagnostic accuracy. Summary of the Invention

[0003] The purpose of the present invention is to provide a method and system for locating power equipment faults based on a dynamic knowledge graph to solve the problems raised in the prior art.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] A method for locating power equipment faults based on a dynamic knowledge graph, the method comprising the following steps:

[0006] Determine the monitoring targets of power equipment and collect operating data of target power equipment;

[0007] Construct a large-parameter model and a small-model adversarial learning framework, and process the output data of the large-parameter model and the small-model data;

[0008] Performing feature analysis on the target power equipment data using a small model, and performing parameter prediction on the target power equipment data using a large parameter model to determine fault analysis data of the target power equipment;

[0009] Based on the large parameter model and the small model, the fault analysis data of the target power equipment is analyzed, and an equipment diagnosis knowledge graph is constructed. Node analysis is performed on the fault analysis data of the target equipment, the inference path of the target power equipment fault data is determined, and the target power equipment fault data is output.

[0010] Furthermore, the grid equipment operation node network is obtained according to the grid control port to determine the number and location data of the target power equipment; the target power equipment includes transformers, high-voltage circuit breakers and GIS equipment;

[0011] The operation data packet of the target power equipment is retrieved through the power grid control port to construct a periodic target power equipment operation data table; the corresponding target power equipment operation data includes dissolved gas data in oil, transformer frequency response data, transformer mechanical status data, circuit breaker winding current status data and GIS partial discharge data, etc.

[0012] Furthermore, the large parameter model and small model adversarial learning framework is constructed, and the data output format is planned for each target power equipment type and the corresponding operation data analysis scenario, and data conversion rules are formulated for the large parameter model and the output data of the small model corresponding to each target power equipment data analysis, so as to achieve alignment processing of the output data of the large parameter model and the small model corresponding to each target power equipment data analysis; wherein the output data formats corresponding to different power equipment and operation data scenarios are quite different, and the format is unified through conversion rules, which is conducive to improving the efficiency and accuracy of data processing;

[0013] Based on the aligned large parameter model and the output data of the small model for analyzing the data of each target power equipment, the output data are compared and judged manually to determine the difference between the large parameter model and the output data of the small model for analyzing the data of each target power equipment, and the model network data training structure or parameters of the large parameter model are adjusted based on the difference, and the adversarial learning data of the large parameter model is updated; if there is a deviation between the large parameter model and the output data of the small model for analyzing a certain operating data of a certain power equipment, the model network structure or parameters can be adjusted, such as by adjusting the neuron connection weight value in the neural network to optimize the output, the model parameters can be continuously updated to enhance the learning ability of the model.

[0014] Furthermore, by retrieving the periodic target power equipment operation data table of each target power equipment, various types of data corresponding to the periodic operation of the target power equipment in the corresponding period are obtained, and the obtained data are respectively input into the corresponding target power equipment data analysis model for feature processing; the operating data of the corresponding type of period at each time point of the target power equipment are used to construct the feature vector of the corresponding type of periodic operation data;

[0015] By integrating the characteristic vectors corresponding to various types of operating data in the cycle of the target power equipment, a unified characteristic vector space is constructed with the corresponding target power equipment as the central node, and the characteristic vector set of the cycle operating data of each target power equipment is determined; the various types of operating data in the adjacent next cycle of each target power equipment are predicted through a large parameter model, and a characteristic vector set of the adjacent next cycle predicted operating data of each target power equipment is constructed; combining the characteristic vector set of the cycle operating data of each target power equipment in the current cycle and the characteristic vector set of the adjacent next cycle predicted operating data, a fault analysis of each target power equipment is performed, and the faulty power equipment is determined based on the analysis results; wherein, the fault analysis calculation is

[0016] ;

[0017] Where Fd(H) is the fault index of the target power equipment with the corresponding number H; V(H) m and V x (H) m The characteristic vectors of the periodic operation data feature vector set of the current cycle of the target power equipment numbered H and the characteristic vectors of the operation data corresponding to type number m in the characteristic vector set of the predicted operation data of the adjacent next cycle respectively; m is the number of the operation data type; according to the fault analysis result, the fault index of each of the target power equipment is judged, and by setting a judgment threshold, a fault prompt is given to the power equipment corresponding to the fault index greater than the threshold.

[0018] Furthermore, based on the large parameter model combined with the small model corresponding to each type of operating data analysis of the target power equipment, the fault analysis results of the periodic operating data of each target power equipment are analyzed to determine the faulty power equipment; by constructing an equipment diagnosis knowledge graph, the faulty equipment is taken as the main node, the various types of operating data within the period of the faulty equipment are taken as the A-layer child nodes, the historical fault type data of the faulty equipment are taken as the B-layer child nodes, and the historical fault cause data corresponding to the historical fault type of the faulty equipment are taken as the C-layer child nodes; by analyzing the fault index of each type of operating data within the period of the faulty power equipment, the periodic fault operating data of the faulty power equipment is determined; wherein, for the fault index analysis of the operating data, the corresponding type of operating data in the current period and the corresponding type of operating data in the adjacent next period are taken, vectors are constructed for the data at each time point, and a fault index analysis is performed. The calculation method and judgment method thereof refer to the fault index calculation method of the above-mentioned power equipment;

[0019] Based on the determined fault operation data, the corresponding fault operation data node is connected to the equipment main node in the equipment diagnosis knowledge graph to construct a fault data chain of the faulty power equipment; by performing a corresponding fault type probability distribution analysis on the determined fault operation data of the faulty power equipment, the fault type data of the current fault operation data is determined based on the analysis results; the historical fault cause data corresponding to the fault type data of the corresponding fault operation fault data is matched, and the fault operation data inference path corresponding to the fault operation data of the current faulty power equipment is constructed according to the matching results; the fault proportion of the fault data inference path corresponding to each fault data is analyzed respectively, and the fault data table of the current faulty power equipment is constructed based on the analysis results, and each fault data is output with priority; wherein the corresponding fault type probability distribution analysis of the fault operation data is performed, and the calculation is:

[0020] ;

[0021] Among them, L(n) is the fault type distribution function value of the fault operation data corresponding to type number n; n(k), n x (k) and n(k) j They correspond to the data values at each time point k in the current cycle of the fault operation data numbered n, the data values at each time point k in the next cycle of the fault operation data numbered n, and the average data values at each time point k in the historical cycle of the fault operation data numbered n; by querying the fault type corresponding to the interval in which the fault type distribution function value of the corresponding fault operation data is located, the fault type corresponding to the fault data of the faulty power equipment in the current cycle is determined; by determining the historical influence weight of the corresponding fault type and the fault cause, the fault type of the current fault operation data is matched with each fault cause; according to the fault type of the corresponding fault operation data and each matched fault cause data, the corresponding nodes are connected in the equipment diagnosis knowledge graph respectively to obtain the inference path of the current fault operation data; based on the inference path of the fault operation data of the current faulty power equipment, the proportion of each inference path of each fault operation data is analyzed respectively, and the analysis calculation is as follows:

[0022] ;

[0023] Among them, Y (g q (n)) corresponds to the fault operation data inference path number g numbered n q Fd(n) is the failure index of the fault operation data corresponding to number n; L(n) i is the distribution function value of the fault type corresponding to number i of the fault operation data corresponding to number n; W q,iis the weight value of the historical fault cause data corresponding to the fault type number i and the fault cause number q; since the fault type corresponding to a single fault data may have multiple fault causes, when constructing the fault operation data inference path, there will be multiple inference paths corresponding to a single fault data; the fault proportions of the inference paths corresponding to the fault operation data of the current faulty power equipment are sorted, and a fault data table of the current faulty power equipment is constructed. The inference paths are sorted from large to small according to the fault proportion values of the inference paths, and the fault data table is output. The corresponding fault operation data, fault type and fault cause of the faulty power equipment in the fault data table are warned.

[0024] A power equipment fault location system based on a dynamic knowledge graph, the system comprising a power equipment monitoring module, a model building and learning module, a fault determination module, and a fault analysis module;

[0025] The power equipment monitoring module determines the power equipment monitoring target and collects operating data of the target power equipment; the model construction and learning module constructs a large parameter model and a small model adversarial learning framework, and processes the output data of the large parameter model and the small model data; the fault determination module performs feature analysis on the target power equipment data through the small model, and performs parameter prediction on the target power equipment data in combination with the large parameter model to determine the fault analysis data of the target power equipment; the fault analysis module constructs an equipment diagnosis knowledge graph based on the fault analysis data of the target power equipment based on the large parameter model and the small model, performs node analysis on the fault analysis data of the target equipment, determines the inference path of the target power equipment fault data, and outputs the target power equipment fault data.

[0026] Furthermore, the power equipment monitoring module includes a target equipment determination unit and a data acquisition unit;

[0027] The target device determination unit obtains the grid device operation node network according to the grid control port and determines the number and location data of the target power device;

[0028] The data acquisition unit retrieves the operation data packet of the target power equipment through the power grid control port and constructs a periodic target power equipment operation data table.

[0029] Furthermore, the model building learning module includes a model data processing unit and a model parameter adjustment unit;

[0030] The model data processing unit constructs a large parameter model and a small model adversarial learning framework, plans the data output format for each target power equipment type and the corresponding operation data analysis scenario, and formulates data conversion rules for the large parameter model and the output data of the corresponding small model for the data analysis of each target power equipment, thereby achieving alignment processing of the output data of the large parameter model and the small model for the data analysis of each target power equipment;

[0031] The model parameter adjustment unit compares and judges the output data of the large parameter model after alignment and the small model corresponding to each target power equipment data analysis manually, determines the difference between the large parameter model and the output data of the small model corresponding to each target power equipment data analysis, and adjusts the model network data training structure or parameters of the large parameter model based on the difference, and updates the adversarial learning data of the large parameter model.

[0032] Furthermore, the fault determination module includes an operation data feature processing unit and a faulty power equipment determination unit;

[0033] The operation data feature processing unit retrieves the periodic target power equipment operation data table of each target power equipment, obtains various types of data corresponding to the periodic operation of the target power equipment in the corresponding period, and inputs the obtained data into the corresponding target power equipment data analysis model for feature processing; constructs the feature vector of the corresponding type of periodic operation data for the operation data of each time point of the corresponding type of period of the target power equipment;

[0034] The faulty power equipment determination unit integrates the characteristic vectors corresponding to various types of operating data within the cycle of the target power equipment, constructs a unified characteristic vector space with the corresponding target power equipment as the central node, and determines the characteristic vector set of the cycle operating data of each target power equipment; predicts various types of operating data in the adjacent next cycle of each target power equipment through a large parameter model, and constructs a characteristic vector set of the adjacent next cycle predicted operating data of each target power equipment; combines the characteristic vector set of the cycle operating data of each target power equipment in the current cycle and the characteristic vector set of the predicted operating data of the adjacent next cycle, performs fault analysis on each target power equipment, and determines the faulty power equipment based on the analysis results.

[0035] Furthermore, the fault analysis module includes a knowledge graph construction unit and a fault data feedback unit;

[0036] The knowledge graph construction unit is based on a large parameter model combined with a small model corresponding to each type of operating data analysis of the target power equipment, and determines the faulty power equipment based on the fault analysis results of the periodic operating data of each target power equipment; by constructing an equipment diagnosis knowledge graph, the faulty equipment is the main node, the various types of operating data within the period of the faulty equipment are the A-layer child nodes, the historical fault type data of the faulty equipment is the B-layer child node, and the historical fault cause data corresponding to the historical fault type of the faulty equipment is the C-layer child node; by analyzing the fault index of each type of operating data within the period of the faulty power equipment, the periodic fault operating data of the faulty power equipment is determined;

[0037] The fault data feedback unit connects the corresponding fault operation data node with the equipment main node in the equipment diagnosis knowledge graph based on the determined fault operation data to construct a fault data chain of the faulty power equipment; performs a corresponding fault type probability distribution analysis on the determined fault operation data of the faulty power equipment, and determines the fault type data of the current fault operation data based on the analysis result; matches the historical fault cause data corresponding to the fault type data of the corresponding fault operation fault data, and constructs a fault operation data inference path corresponding to the fault operation data of the current faulty power equipment based on the matching result; analyzes the fault proportion of the fault data inference path corresponding to each fault data, and constructs a fault data table of the current faulty power equipment based on the analysis result, and outputs each fault data with priority.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] The present invention integrates multi-source data of monitoring equipment by constructing a large parameter model and a distributed small model, and determines the faulty power equipment by performing feature processing on the power equipment operation data; analyzes the fault data based on the faulty power equipment, and constructs a device diagnosis knowledge graph to perform fault path analysis on the fault data; the present invention performs adversarial learning on the large parameter model and the small model, performs feature processing on the equipment operation data and performs fault analysis, and then after determining the fault data of the faulty equipment, analyzes the fault type and fault cause in a hierarchical manner, performs root cause analysis based on the knowledge graph to construct a fault reasoning path, and outputs the faulty equipment, fault data, fault type and fault cause; the present invention makes up for the manual dependence of existing equipment diagnosis; improves the insufficient utilization of multi-source and multi-modal data, and improves the detection performance of equipment faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 This is a structural diagram of a power equipment fault location system based on a dynamic knowledge graph according to the present invention;

[0041] Figure 2This is a flow chart of a method for locating faults in power equipment based on a dynamic knowledge graph according to the present invention. DETAILED DESCRIPTION

[0042] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0043] Example: Figure 1 As shown, the present invention provides a technical solution:

[0044] A power equipment fault location system based on dynamic knowledge graph, including a power equipment monitoring module, a model building and learning module, a fault determination module and a fault analysis module;

[0045] Among them, the power equipment monitoring module determines the power equipment monitoring target and collects the operating data of the target power equipment; the model construction and learning module constructs a large-parameter model and a small-model adversarial learning framework, and processes the output data of the large-parameter model and the small-model data; the fault determination module performs feature analysis on the target power equipment data through the small model, and predicts the parameters of the target power equipment data in combination with the large-parameter model to determine the fault analysis data of the target power equipment; the fault analysis module constructs an equipment diagnosis knowledge graph based on the fault analysis data of the target power equipment based on the large-parameter model and the small model, performs node analysis on the fault analysis data of the target equipment, determines the inference path of the target power equipment fault data, and outputs the target power equipment fault data.

[0046] Furthermore, the power equipment monitoring module includes a target equipment determination unit and a data acquisition unit;

[0047] The target device determination unit obtains the grid equipment operation node network according to the grid control port and determines the number and location data of the target power equipment;

[0048] The data acquisition unit retrieves the operating data packet of the target power equipment through the power grid control port and constructs a periodic target power equipment operating data table.

[0049] Furthermore, the model building learning module includes a model data processing unit and a model parameter adjustment unit;

[0050] The model data processing unit constructs a large-parameter model and a small-model adversarial learning framework. By planning the data output format for each target power equipment type and the corresponding operation data analysis scenario, and formulating data conversion rules for the large-parameter model and the output data of the small-model corresponding to each target power equipment data analysis, the alignment processing of the large-parameter model and the output data of the small-model corresponding to each target power equipment data analysis is achieved.

[0051] The model parameter adjustment unit compares and judges the output data of the large parameter model after alignment and the small model corresponding to each target power equipment data analysis manually, determines the difference between the large parameter model and the output data of the small model corresponding to each target power equipment data analysis, and adjusts the model network data training structure or parameters of the large parameter model based on the difference, and updates the adversarial learning data of the large parameter model.

[0052] Furthermore, the fault determination module includes an operation data feature processing unit and a faulty power equipment determination unit;

[0053] The operation data feature processing unit retrieves the periodic target power equipment operation data table of each target power equipment, obtains various types of periodic operation data of the corresponding target power equipment in the corresponding period, and inputs the obtained data into the corresponding target power equipment data analysis model for feature processing; the operation data of the target power equipment at each time point of the corresponding type of period is used to construct the feature vector of the corresponding type of periodic operation data;

[0054] The faulty power equipment determination unit integrates the characteristic vectors corresponding to various types of operating data within the cycle of the corresponding target power equipment, constructs a unified characteristic vector space with the corresponding target power equipment as the central node, and determines the characteristic vector set of the cycle operating data of each target power equipment; predicts various types of operating data in the adjacent next cycle of each target power equipment through a large parameter model, and constructs a characteristic vector set of the adjacent next cycle predicted operating data of each target power equipment; combines the characteristic vector set of the cycle operating data of each target power equipment in the current cycle and the characteristic vector set of the predicted operating data of the adjacent next cycle, performs fault analysis on the corresponding target power equipment, and determines the faulty power equipment based on the analysis results.

[0055] Furthermore, the fault analysis module includes a knowledge graph construction unit and a fault data feedback unit;

[0056] The knowledge graph construction unit is based on a large parameter model combined with a small model for analyzing the various types of operating data of the corresponding target power equipment, and determines the faulty power equipment based on the fault analysis results of the periodic operating data of each target power equipment. By constructing an equipment diagnosis knowledge graph, the faulty equipment is the main node, the various types of operating data within the faulty equipment cycle are the A-layer child nodes, the historical fault type data of the faulty equipment is the B-layer child node, and the historical fault cause data corresponding to the historical fault type of the faulty equipment is the C-layer child node. By analyzing the fault index of the various types of operating data within the faulty power equipment cycle, the periodic fault operating data of the faulty power equipment is determined.

[0057] Based on the determined fault operation data, the fault data feedback unit connects the corresponding fault operation data node with the device main node in the equipment diagnosis knowledge graph to build a fault data chain for the faulty power equipment; by performing a corresponding fault type probability distribution analysis on the determined fault operation data of the faulty power equipment, the fault type data of the current fault operation data is determined based on the analysis results; the fault type data corresponding to the historical fault cause data of the corresponding fault operation data are matched, and a fault operation data inference path corresponding to the fault operation data of the current faulty power equipment is constructed based on the matching results; the fault proportion of the fault data inference path corresponding to each fault data is analyzed respectively, and based on the analysis results, a fault data table of the current faulty power equipment is constructed respectively, and each fault data is output with priority;

[0058] like Figure 2 As shown, the present invention provides another technical solution:

[0059] A method for locating power equipment faults based on a dynamic knowledge graph, the method comprising the following steps:

[0060] Determine the monitoring targets of power equipment and collect operating data of target power equipment;

[0061] Construct a large-parameter model and a small-model adversarial learning framework, and process the output data of the large-parameter model and the small-model data;

[0062] Perform feature analysis on target power equipment data through a small model, and perform parameter prediction on target power equipment data by combining with a large parameter model to determine the fault analysis data of the target power equipment;

[0063] Based on the fault analysis data of the target power equipment using the large parameter model and the small model, an equipment diagnosis knowledge graph is constructed, node analysis is performed on the fault analysis data of the target equipment, the inference path of the target power equipment fault data is determined, and the target power equipment fault data is output.

[0064] Furthermore, the network of power grid equipment operation nodes is obtained based on the power grid control port to determine the number and location data of the target power equipment; the target power equipment includes transformers, high-voltage circuit breakers and GIS equipment;

[0065] The operation data packet of the target power equipment is retrieved through the power grid control port to construct a periodic target power equipment operation data table; the corresponding target power equipment operation data includes dissolved gas data in oil, transformer frequency response data, transformer mechanical status data, circuit breaker winding current status data and GIS partial discharge data, etc.

[0066] Furthermore, a large-parameter model and a small-model adversarial learning framework are constructed. By planning the data output format for each target power equipment type and the corresponding operation data analysis scenario, and formulating data conversion rules for the output data of the large-parameter model and the small-model corresponding to each target power equipment data analysis, the alignment processing of the output data of the large-parameter model and the small-model corresponding to each target power equipment data analysis is achieved. The output data formats corresponding to different power equipment and operation data scenarios vary greatly. Unifying the formats through conversion rules is conducive to improving the efficiency and accuracy of data processing.

[0067] Based on the aligned large parameter model and the output data of the small model for analyzing the data of each target power equipment, the output data are compared and judged manually to determine the difference between the large parameter model and the output data of the small model for analyzing the data of each target power equipment, and based on the difference, the model network data training structure or parameters of the large parameter model are adjusted, and the adversarial learning data of the large parameter model is updated; if there is a deviation between the large parameter model and the output data of the small model for analyzing a certain operating data of a certain power equipment, the model network structure or parameters can be adjusted, such as optimizing the output by adjusting the neuron connection weight value in the neural network, and the model parameters can be continuously updated to enhance the learning ability of the model.

[0068] Furthermore, by retrieving the periodic target power equipment operation data table of each target power equipment, various types of periodic operation data of the corresponding target power equipment in the corresponding period are obtained, and the obtained data are respectively input into the corresponding target power equipment data analysis model for feature processing; the operating data of the target power equipment at each time point of the corresponding type of period is used to construct the feature vector of the corresponding type of periodic operation data;

[0069] By integrating the characteristic vectors corresponding to various types of operating data within the cycle of the corresponding target power equipment, a unified characteristic vector space is constructed with the corresponding target power equipment as the central node, and the characteristic vector set of the periodic operating data of each target power equipment is determined; the various types of operating data within the adjacent next cycle of each target power equipment are predicted through a large parameter model, and a characteristic vector set of the adjacent next cycle predicted operating data of each target power equipment is constructed; the characteristic vector set of the periodic operating data of each target power equipment in the current cycle and the characteristic vector set of the adjacent next cycle predicted operating data are combined to perform fault analysis on each target power equipment, and the faulty power equipment is determined based on the analysis results; wherein, the fault analysis calculation is

[0070] ;

[0071] Where Fd(H) is the fault index of the target power equipment with the corresponding number H; V(H) m and V x (H) m The characteristic vectors of the periodic operation data feature vector set of the current cycle of the target power equipment numbered H and the characteristic vectors of the operation data of type numbered m in the characteristic vector set of the predicted operation data of the adjacent next cycle respectively; m is the number of the operation data type; according to the fault analysis results, the fault index of each target power equipment is judged, and by setting a judgment threshold, a fault prompt is given to the power equipment corresponding to the fault index greater than the threshold.

[0072] Furthermore, based on the large parameter model combined with the corresponding small model for analyzing the various types of operating data of the target power equipment, the fault analysis results of the periodic operating data of each target power equipment are analyzed to determine the faulty power equipment; by constructing an equipment diagnosis knowledge graph, the faulty equipment is taken as the main node, the various types of operating data within the period of the faulty equipment are taken as the A-layer child nodes, the historical fault type data of the faulty equipment are taken as the B-layer child nodes, and the historical fault cause data corresponding to the historical fault type of the faulty equipment are taken as the C-layer child nodes; by analyzing the fault index of various types of operating data within the period of the faulty power equipment, the periodic fault operating data of the faulty power equipment is determined; wherein, for the fault index analysis of the operating data, the corresponding type of operating data in the current period and the corresponding type of operating data in the adjacent next period are taken, vectors are constructed for the data at each time point, and a fault index analysis is performed. The calculation method and judgment method thereof refer to the fault index calculation method of the above-mentioned power equipment.

[0073] Based on the determined fault operation data, the corresponding fault operation data node is connected to the equipment main node in the equipment diagnosis knowledge graph to construct a fault data chain of the faulty power equipment; by performing a corresponding fault type probability distribution analysis on the determined fault operation data of the faulty power equipment, the fault type data of the current fault operation data is determined based on the analysis results; the historical fault cause data corresponding to the fault type data of the corresponding fault operation fault data is matched, and the fault operation data inference path corresponding to the fault operation data of the current faulty power equipment is constructed according to the matching results; the fault proportion of the fault data inference path corresponding to each fault data is analyzed respectively, and the fault data table of the current faulty power equipment is constructed based on the analysis results, and each fault data is output with priority; wherein the corresponding fault type probability distribution analysis of the fault operation data is performed, and the calculation is:

[0074] ;

[0075] Among them, L(n) is the fault type distribution function value of the fault operation data corresponding to type number n; n(k), n x (k) and n(k) j They correspond to the data values at each time point k in the current cycle of the fault operation data numbered n, the data values at each time point k in the next cycle of the fault operation data numbered n, and the average data values at each time point k in the historical cycle of the fault operation data numbered n; by querying the fault type corresponding to the interval in which the fault type distribution function value of the corresponding fault operation data is located, the fault type corresponding to the fault data of the faulty power equipment in the current cycle is determined; by determining the historical influence weight of the corresponding fault type and the fault cause, the fault type of the current fault operation data is matched with each fault cause; according to the fault type of the corresponding fault operation data and each matched fault cause data, the corresponding nodes are connected in the equipment diagnosis knowledge graph respectively to obtain the inference path of the current fault operation data; based on the inference path of the fault operation data of the current faulty power equipment, the proportion of each inference path of each fault operation data is analyzed respectively, and the analysis calculation is as follows:

[0076] ;

[0077] Among them, Y (g q (n)) corresponds to the fault operation data inference path number g numbered n q Fd(n) is the failure index of the fault operation data corresponding to number n; L(n) i is the distribution function value of the fault type corresponding to number i of the fault operation data corresponding to number n; W q,iis the weight value of the historical fault cause data corresponding to the fault type number i and the fault cause number q; since the fault type corresponding to a single fault data may have multiple fault causes, when constructing the fault operation data inference path, there will be multiple inference paths corresponding to a single fault data; the fault proportions of the inference paths corresponding to the fault operation data of the current faulty power equipment are sorted, and a fault data table of the current faulty power equipment is constructed. The inference paths are sorted from large to small according to the fault proportion values of the inference paths, and the fault data table is output. The corresponding fault operation data, fault type and fault cause of the faulty power equipment in the fault data table are warned.

[0078] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. A method for locating power equipment faults based on a dynamic knowledge graph, characterized by: The method comprises the following steps: Determine the monitoring targets of power equipment and collect operating data of target power equipment; Construct a large-parameter model and a small-model adversarial learning framework, and process the output data of the large-parameter model and the small-model data; Performing feature analysis on the target power equipment data using a small model, and performing parameter prediction on the target power equipment data using a large parameter model to determine fault analysis data of the target power equipment; Based on the large parameter model combined with the small model corresponding to the various types of operating data analysis of the target power equipment, the faulty power equipment is determined by analyzing the fault analysis results of the periodic operating data of each target power equipment; by constructing an equipment diagnosis knowledge graph, the faulty equipment is taken as the main node, the various types of operating data within the period of the faulty equipment are taken as the A-layer child nodes, the historical fault type data of the faulty equipment is taken as the B-layer child nodes, and the historical fault cause data corresponding to the historical fault type of the faulty equipment is taken as the C-layer child nodes; by analyzing the fault index of the various types of operating data within the period of the faulty power equipment, the periodic fault operating data of the faulty power equipment is determined; Based on the determined fault operation data, the corresponding fault operation data node is connected to the device master node in the equipment diagnosis knowledge graph to construct a fault data chain for the faulty power equipment. By performing a probability distribution analysis on the corresponding fault type of the determined fault operation data of the faulty power equipment and querying the fault type corresponding to the interval in which the fault type distribution function value of the corresponding fault operation data lies, the fault type corresponding to the fault data of the faulty power equipment in the current cycle is determined. By determining the historical impact weights of the corresponding fault type and fault cause, the fault type of the current fault operation data is matched to each fault cause. Based on the fault type of the corresponding fault operation data and each matched fault cause data, the corresponding nodes are connected in the equipment diagnosis knowledge graph to obtain the inference path of the current fault operation data. Based on the inference path of the fault operation data of the current faulty power equipment, the proportion of each inference path of each fault operation data is analyzed. The fault proportion of each inference path corresponding to each fault operation data of the current faulty power equipment is ranked, and a fault data table of the current faulty power equipment is constructed. The inference path fault proportion values are ranked from large to small, and the fault data table is output. A warning is issued for the corresponding fault operation data, fault type, and fault cause of the faulty power equipment in the fault data table.

2. The method for locating power equipment faults based on a dynamic knowledge graph according to claim 1, characterized in that: Obtain the grid equipment operation node network according to the grid control port, and determine the number and location data of the target power equipment; The operation data packet of the target power equipment is retrieved through the power grid control port to construct a periodic target power equipment operation data table.

3. The method for locating power equipment faults based on a dynamic knowledge graph according to claim 2, characterized in that: The large parameter model and the small model adversarial learning framework are constructed, and the data output format is planned for each target power equipment type and the corresponding operation data analysis scenario, and data conversion rules are formulated for the large parameter model and the output data of the small model corresponding to each target power equipment data analysis, so as to achieve alignment processing of the output data of the large parameter model and the small model corresponding to each target power equipment data analysis; Based on the aligned large parameter model and the output data of the small model corresponding to each of the target power equipment data analysis, the output data is compared and judged manually to determine the difference between the large parameter model and the output data of the small model corresponding to each of the target power equipment data analysis, and based on the difference, the model network data training structure or parameter adjustment of the large parameter model is performed, and the adversarial learning data of the large parameter model is updated.

4. The method for locating power equipment faults based on a dynamic knowledge graph according to claim 3, characterized in that: By retrieving the periodic target power equipment operation data table of each target power equipment, various types of data corresponding to the periodic operation of the target power equipment in the corresponding period are obtained, and the obtained data are respectively input into the corresponding target power equipment data analysis model for feature processing; For the operation data of the target power equipment at each time point of the corresponding type period, construct a feature vector of the corresponding type period operation data; By integrating the characteristic vectors corresponding to various types of operating data within the cycle of the target power equipment, a unified characteristic vector space is constructed with the corresponding target power equipment as the central node, and the characteristic vector set of the periodic operating data of each target power equipment is determined; the various types of operating data in the adjacent next cycle of each target power equipment are predicted through a large parameter model, and a characteristic vector set of the adjacent next cycle predicted operating data of each target power equipment is constructed; combined with the characteristic vector set of the periodic operating data of each target power equipment in the current cycle and the characteristic vector set of the predicted operating data of the adjacent next cycle, a fault analysis of each target power equipment is performed, and the faulty power equipment is determined based on the analysis results.

5. A power equipment fault location system based on a dynamic knowledge graph, using a power equipment fault location method based on a dynamic knowledge graph according to any one of claims 1 to 4, characterized in that: The system includes an electric equipment monitoring module, a model building and learning module, a fault determination module, and a fault analysis module; The power equipment monitoring module determines the power equipment monitoring target and collects the operating data of the target power equipment; the model building and learning module builds a large parameter model and a small model adversarial learning framework and processes the output data of the large parameter model and the small model; The fault determination module performs feature analysis on the target power equipment data using a small model, performs parameter prediction on the target power equipment data in combination with a large parameter model, and determines the fault analysis data of the target power equipment; The fault analysis module constructs an equipment diagnosis knowledge graph based on the fault analysis data of the target power equipment based on the large parameter model and the small model, performs node analysis on the fault analysis data of the target power equipment, determines the inference path of the target power equipment fault data, and outputs the target power equipment fault data.

6. The power equipment fault location system based on dynamic knowledge graph according to claim 5 is characterized by: The power equipment monitoring module includes a target equipment determination unit and a data acquisition unit; The target device determination unit obtains the grid device operation node network according to the grid control port and determines the number and location data of the target power device; The data acquisition unit retrieves the operation data packet of the target power equipment through the power grid control port and constructs a periodic target power equipment operation data table.

7. The power equipment fault location system based on dynamic knowledge graph according to claim 6 is characterized by: The model building learning module includes a model data processing unit and a model parameter adjustment unit; The model data processing unit constructs a large parameter model and a small model adversarial learning framework, plans the data output format for each target power equipment type and the corresponding operation data analysis scenario, and formulates data conversion rules for the large parameter model and the output data of the corresponding small model for the data analysis of each target power equipment, thereby achieving alignment processing of the output data of the large parameter model and the small model for the data analysis of each target power equipment; The model parameter adjustment unit compares and judges the output data of the large parameter model after alignment and the small model corresponding to each target power equipment data analysis manually, determines the difference between the large parameter model and the output data of the small model corresponding to each target power equipment data analysis, and adjusts the model network data training structure or parameters of the large parameter model based on the difference, and updates the adversarial learning data of the large parameter model.

8. The power equipment fault location system based on dynamic knowledge graph according to claim 7 is characterized by: The fault determination module includes an operation data feature processing unit and a faulty power equipment determination unit; The operation data feature processing unit retrieves the periodic target power equipment operation data table of each target power equipment, obtains various types of data corresponding to the periodic operation of the target power equipment in the corresponding period, and inputs the obtained data into the corresponding target power equipment data analysis model for feature processing; For the operation data of the target power equipment at each time point of the corresponding type period, construct a feature vector of the corresponding type period operation data; The faulty power equipment determination unit integrates the characteristic vectors corresponding to various types of operating data within the cycle of the target power equipment, constructs a unified characteristic vector space with the corresponding target power equipment as the central node, and determines the characteristic vector set of the cycle operating data of each target power equipment; predicts various types of operating data in the adjacent next cycle of each target power equipment through a large parameter model, and constructs a characteristic vector set of the adjacent next cycle predicted operating data of each target power equipment; combines the characteristic vector set of the cycle operating data of each target power equipment in the current cycle and the characteristic vector set of the predicted operating data of the adjacent next cycle, performs fault analysis on each target power equipment, and determines the faulty power equipment based on the analysis results.

9. The power equipment fault location system based on dynamic knowledge graph according to claim 8, characterized in that: The fault analysis module includes a knowledge graph construction unit and a fault data feedback unit; The knowledge graph construction unit analyzes the fault results of the periodic operation data of each target power device based on the large parameter model combined with the small model corresponding to each type of operation data of the target power device to determine the faulty power device; By constructing an equipment diagnosis knowledge graph, the faulty equipment is the main node, the various types of operating data within the faulty equipment cycle are the A-layer child nodes, the historical fault type data of the faulty equipment is the B-layer child node, and the historical fault cause data corresponding to the historical fault type of the faulty equipment is the C-layer child node; By analyzing the fault index of various types of operating data within the period of the faulty power equipment, the periodic fault operating data of the faulty power equipment is determined; The fault data feedback unit connects the corresponding fault operation data node with the device master node in the device diagnosis knowledge graph based on the determined fault operation data to build a fault data chain of the faulty power equipment; By performing a probability distribution analysis of the corresponding fault type on the determined fault operation data of the faulty power equipment, the fault type data of the current fault operation data is determined based on the analysis result; Match the historical fault cause data corresponding to the fault type data of the corresponding fault operation fault data, and construct a fault operation data inference path corresponding to the fault operation data of the current faulty power equipment based on the matching result; The fault proportion of each fault data corresponding to the fault data inference path is analyzed respectively, and the fault data table of the current faulty power equipment is constructed based on the analysis results, and the priority of each fault data is output.

Images