Big data-based transformer substation fault intelligent analysis and diagnosis method and system

Abstract

The invention provides a transformer substation fault intelligent analysis and diagnosis method and system based on big data, and the method comprises the steps: obtaining the surface and internal temperature data flows of all power equipment in a transformer substation, and the environment temperature and humidity data flow of the transformer substation in the operation environment of the transformer substation; performing cross-time-scale dynamic coupling on the temperature data stream and the environment temperature and humidity data stream to generate a heterogeneous temperature characteristic stream; based on the heterogeneous temperature characteristic flow, dynamically adjusting an association rule of the temperature anomaly of the power equipment and the operation state of the transformer substation; performing cross-device coupling analysis on the heterogeneous temperature characteristic flow through a deep belief network to generate a dynamic conduction topological structure; and generating a transformer substation fault diagnosis result in combination with the adjusted fault association rule and the dynamic conduction topological structure. According to the invention, the early warning accuracy and diagnosis efficiency of the substation fault are improved.

Description

Technical Field

[0001] This application relates to the technical field of intelligent operation and maintenance of power systems, and particularly to an intelligent analysis and diagnosis method and system for substation faults based on big data. Background Art

[0002] With the advancement of the construction of smart grids, the scale of substation equipment has expanded and the operation complexity has increased. Traditional fault diagnosis methods are difficult to meet the requirements of real-time and accuracy. Modern substations require an intelligent analysis technology that can integrate multi-source monitoring data, automatically identify the abnormal correlation relationship between equipment, and achieve early fault warning to meet the accurate control requirements of equipment status under high-load operating conditions.

[0003] Currently, there is a solution that uses a temperature anomaly diagnosis model based on support vector machines. By collecting the surface temperature data of equipment, combining with preset temperature threshold rules for anomaly detection, and using the historical fault data of equipment to train a classification model, the identification of common fault types is achieved. This solution improves the diagnosis efficiency through a data-driven method and reduces manual intervention.

[0004] This solution only relies on a single data dimension of the surface temperature of equipment, and the static threshold rules are difficult to adapt to the temperature change characteristics under different operating conditions, resulting in a high false alarm rate. At the same time, the model lacks the ability to analyze the thermal conduction correlation between equipment. Summary of the Invention

[0005] This application provides an intelligent analysis and diagnosis method and system for substation faults based on big data to solve the problems of low accuracy of early fault warning and poor diagnosis efficiency in the prior art for substation faults.

[0006] In the first aspect, this application provides an intelligent analysis and diagnosis method for substation faults based on big data, including:

[0007] In the substation operation environment, obtain the temperature data streams on the surface and inside of each power equipment in the substation, and the environmental temperature and humidity data streams of the substation;

[0008] Perform dynamic coupling of the temperature data stream and the environmental temperature and humidity data stream across time scales to generate a heterogeneous temperature feature stream;

[0009] Based on the heterogeneous temperature feature stream, dynamically adjust the association rules between power equipment temperature anomalies and substation operation states;

[0010] Perform cross-device coupling analysis on the heterogeneous temperature feature stream through a deep belief network to generate a dynamic conduction topology;

[0011] Combine the adjusted fault association rules with the dynamic conduction topology to generate a substation fault diagnosis result.

[0012] Optionally, the cross-device coupling analysis of the heterogeneous temperature feature stream by the deep belief network to generate a dynamic conduction topology structure includes:

[0013] Performing multi-level feature processing on the heterogeneous temperature feature stream through a deep belief network to construct an initial topology network;

[0014] Based on the time persistence of the heterogeneous temperature feature stream, iteratively correcting the initial topology network to generate a target topology network;

[0015] Performing spatial matching between the target topology network and the physical location coordinate data of each power device in the substation to generate a dynamic conduction topology structure.

[0016] Optionally, the performing spatial matching between the target topology network and the physical location coordinate data of each power device in the substation to generate a dynamic conduction topology structure includes:

[0017] Obtaining the physical location coordinate data of each power device in the substation;

[0018] Comparing the connection paths between nodes in the target topology network with the physical location coordinate data to filter out all initial valid conduction connection paths with lengths less than a preset spatial distance threshold;

[0019] Performing weight correction on each of the initial valid conduction connection paths according to the electrical circuit relationship and the heat dissipation area orientation of each power device in the substation. After the correction is completed, deleting redundant paths that do not meet the preset conduction conditions to generate target valid conduction connection paths;

[0020] Mapping the target valid conduction connection paths to the actual layout orientation in the physical location coordinate data to generate a dynamic conduction topology structure.

[0021] Optionally, the performing weight correction on each of the initial valid conduction connection paths according to the electrical circuit relationship and the heat dissipation area orientation of each power device in the substation includes:

[0022] Identifying device node pairs belonging to the same electrical connection group among all the initial valid conduction connection paths, and performing in-circuit enhancement processing on the conduction weight values of the device node pairs according to a preset ratio. The electrical connection group is a set of device groups divided according to the electrical circuit relationship;

[0023] Calculating the conduction attenuation factor of the cross-interval device node pairs based on the azimuth angle data of adjacent heat dissipation intervals in the heat dissipation area orientation;

[0024] The conduction weight value after the same-loop enhancement process is multiplied by the conduction attenuation factor to generate a comprehensive correction weight value;

[0025] According to the comprehensive correction weight value, weight correction is performed on each of the initial effective conductive connection paths.

[0026] Optionally, performing multi-level feature processing on the heterogeneous temperature feature stream through a deep belief network to construct an initial topological network includes:

[0027] By using the first level processing module of the deep belief network, the temperature fluctuation synchronization characteristics corresponding to the internal equipment of the group unit are extracted from the heterogeneous temperature feature stream based on the group units divided by the physical position relationship of each power equipment in the substation;

[0028] The temperature fluctuation synchronization feature is superimposed on the substation environment temperature and humidity data stream through the second-level processing module to generate an intermediate feature set;

[0029] The temperature correlation features exceeding the preset conduction threshold are screened out from the intermediate feature set through the third-level processing module, and an initial topology network is constructed according to the conduction strength relationship between the power equipment in the substation corresponding to the temperature correlation features. The temperature correlation features refer to the temperature correlation features between the power equipment in the substation.

[0030] Optionally, dynamically coupling the temperature data stream and the ambient temperature and humidity data stream across time scales to generate a heterogeneous temperature feature stream includes:

[0031] Divide the second-level time window based on the temperature sampling period of each power device, synchronously match the ambient temperature and humidity data with the temperature data stream within the second-level time window, and generate an instantaneous temperature data unit;

[0032] Divide the hourly time window according to the load change cycle of the substation, and perform accumulation processing on the instantaneous temperature data unit to extract the temperature trend change characteristics of each power equipment within the hourly time window;

[0033] Combine the historical operation data of the substation to divide the monthly time window, associate and map the temperature trend change characteristics with the historical temperature fluctuation range under the same seasonal conditions, and generate a temperature feature set;

[0034] The instantaneous temperature data unit, the temperature trend change feature and the temperature feature set are hierarchically integrated according to timestamps to form a heterogeneous temperature feature stream.

[0035] Optionally, combining the adjusted fault association rule with the dynamic conduction topology structure to generate a substation fault diagnosis result includes:

[0036] Mark the device nodes in the dynamic conduction topology structure whose conduction intensity exceeds the preset alarm threshold as suspicious fault source nodes, and extract the temperature anomaly characteristic patterns corresponding to the suspicious fault source nodes;

[0037] Match and verify the temperature anomaly characteristic patterns with the adjusted fault association rules. When the match verification result indicates that the suspicious fault source nodes simultaneously meet the device aging gradient change law and the environmental temperature and humidity coupling condition in the adjusted fault association rules, confirm the suspicious fault source nodes as diagnosed fault source nodes;

[0038] Generate associated device risk warning information according to the conduction path intensity between the diagnosed fault source nodes and other device nodes in the dynamic conduction topology structure;

[0039] Combine the location information of the diagnosed fault source nodes with the associated device risk warning information to generate a substation fault diagnosis result.

[0040] In a second aspect, the present application provides a substation fault intelligent analysis and diagnosis system based on big data, including:

[0041] An acquisition module, configured to acquire the temperature data stream on the surface and inside of each power device in the substation and the environmental temperature and humidity data stream of the substation in the substation operation environment;

[0042] A coupling module, configured to perform dynamic coupling of the temperature data stream and the environmental temperature and humidity data stream across time scales to generate a heterogeneous temperature characteristic stream;

[0043] An adjustment module, configured to dynamically adjust the association rules between the temperature anomaly of the power device and the operation state of the substation based on the heterogeneous temperature characteristic stream;

[0044] An analysis module, configured to perform cross-device coupling analysis on the heterogeneous temperature characteristic stream through a deep belief network to generate a dynamic conduction topology structure;

[0045] A generation module, configured to generate a substation fault diagnosis result by combining the adjusted fault association rules and the dynamic conduction topology structure.

[0046] In a third aspect, the present application provides a computing device, including a processor and a memory, where a computer program is stored in the memory, and the processor is configured to run the computer program to execute any one of the methods for intelligent analysis and diagnosis of substation faults based on big data in the first aspect.

[0047] In a fourth aspect, the present application provides a computer storage medium, on which computer program instructions are stored. When the computer program instructions are executed by a processor, a method for intelligent analysis and diagnosis of substation faults based on big data according to any one of the first aspects is implemented.

[0048] In the present application, a method for intelligent analysis and diagnosis of substation faults based on big data is provided. The method includes: in the substation operation environment, acquiring the temperature data streams on the surface and inside of each power device in the substation and the ambient temperature and humidity data streams of the substation; performing dynamic coupling of the temperature data streams and the ambient temperature and humidity data streams across time scales to generate heterogeneous temperature feature streams; based on the heterogeneous temperature feature streams, dynamically adjusting the association rules between power device temperature anomalies and substation operation states; performing cross-device coupling analysis on the heterogeneous temperature feature streams through a deep belief network to generate a dynamic conduction topology structure; and combining the adjusted fault association rules with the dynamic conduction topology structure to generate a substation fault diagnosis result.

[0049] The technical solution provided by the present application has the following beneficial effects:

[0050] In the present application, by synchronously collecting the surface and internal temperature data of the device and the ambient temperature and humidity parameters, a multi-dimensional monitoring system covering the full operating state of the device is established, significantly improving data integrity; realizing dynamic association modeling of environmental parameters and device temperature data, overcoming the neglect of the influence of environmental factors in traditional methods, and enhancing the reliability of temperature anomaly determination; adaptively optimizing the fault determination rules based on real-time data, enabling the diagnostic criteria to automatically adjust with the operating load and environment, greatly reducing misjudgments; revealing the hidden heat conduction relationship between devices through a deep belief network, and for the first time realizing visual analysis of the topological conduction path of substation device faults; combining the dual verification mechanisms of dynamic rules and conduction topology, significantly improving the diagnostic accuracy and early warning ability of complex faults.

[0051] Furthermore, in the present application, an initial network is constructed through multi-level feature processing, and then the network structure is iteratively optimized based on time persistence. Finally, a spatially matched dynamic conduction topology is generated in combination with the physical location of the device.

[0052] Moreover, accurate modeling of the heat conduction relationship between devices is achieved, effectively identifying chain faults caused by heat conduction, and making the diagnostic accuracy and early warning timeliness of conduction-type faults reach the leading level in the industry.

[0053] These aspects or other aspects of the present application will be more clearly understood in the following description of the embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on these drawings.

[0055] Figure 1 It is a flowchart of a method for intelligent analysis and diagnosis of substation faults based on big data provided by an embodiment of the present application;

[0056] Figure 2 It is a schematic structural diagram of a system for intelligent analysis and diagnosis of substation faults based on big data provided by an embodiment of the present application;

[0057] Figure 3 It is a schematic structural diagram of a computing device provided by an embodiment of the present application. Detailed implementation manners

[0058] In order to enable those skilled in the art to better understand the solutions of the present application, the following will clearly and completely describe the technical solutions in the embodiments of the present application in conjunction with the accompanying drawings in the embodiments of the present application.

[0059] In some processes described in the specification, claims and the above accompanying drawings of the present application, there are multiple operations that appear in a specific order. However, it should be clearly understood that these operations can be executed not in the order in which they appear in this article or in parallel. The operation numbers such as 101, 102, etc. are only used to distinguish different operations, and the numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations can be executed in sequence or in parallel. It should be noted that the descriptions such as "first" and "second" in this article are used to distinguish different messages, devices, modules, etc., and do not represent a sequence, nor do they limit that "first" and "second" are of different types.

[0060] The existing substation fault diagnosis solutions mainly rely on the support vector machine model to perform static threshold analysis on the surface temperature data of equipment. Its core defects are as follows: on the one hand, the single temperature dimension monitoring cannot reflect the influence of environmental temperature and humidity on the operation state of the equipment, resulting in the disconnection between the abnormal judgment result and the actual working condition; on the other hand, the fixed threshold rule is difficult to adapt to the temperature change characteristics under different load conditions, and the model lacks the modeling ability of the heat conduction correlation between equipment, resulting in a high misdiagnosis rate of conduction-type faults. The essence of these problems lies in the insufficient analysis of the dynamic association of multi-source data and the implicit equipment coupling relationship in the existing technology, which is difficult to meet the requirements of intelligent substations for accurate fault early warning.

[0061] In view of the above deficiencies, the present invention proposes an intelligent analysis and diagnosis method for substation faults based on big data. The core lies in generating heterogeneous temperature feature streams by dynamically coupling the surface / inner temperature of equipment with the ambient temperature and humidity data, realizing multi-dimensional real-time perception of the operating state; adaptively adjusting fault correlation rules based on the feature streams to dynamically optimize the diagnostic criteria according to the working conditions; and further using a deep belief network to mine the conduction topology relationship between equipment and establish a fault propagation path model. This method breaks through the limitations of traditional single data analysis and static rules. Through the dual technical paths of multi-source data fusion and implicit feature extraction, it not only solves the problem of misjudgment caused by environmental factor interference but also realizes the accurate positioning of conduction faults, significantly improving the reliability and timeliness of substation fault diagnosis.

[0062] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0063] Figure 1 The flowchart of an intelligent analysis and diagnosis method for substation faults based on big data provided by an embodiment of the present application is as Figure 1 shown, and the method includes:

[0064] Step 101: In the substation operating environment, obtain the temperature data streams on the surface and inside of each power equipment in the substation and the ambient temperature and humidity data streams of the substation.

[0065] In step 101, the temperature data stream represents the temperature time-series data continuously collected on the surface and inside of the power equipment, including the temperature change information of the key parts of the equipment. The ambient temperature and humidity data stream represents the continuous monitoring data of the air temperature and humidity collected at the monitoring points in the substation.

[0066] In the embodiment of the present application, in the substation operating environment, distributed temperature sensors installed at key parts such as transformer bushings and circuit breaker contacts are used to collect the surface and inner temperature data of the equipment at a fixed sampling frequency to form a temperature data stream; at the same time, the temperature and humidity data of areas such as the main control room and switch room are collected through environmental monitoring devices, and both types of data are synchronously cached based on the time stamp to provide the original input for subsequent processing.

[0067] For example, a 500kV substation deploys temperature sensors at 12 key points, such as transformer oil temperature probes and GIS (Gas Insulated Switchgear, GIS) equipment casings, to collect temperature data once a second. At the same time, four temperature and humidity monitoring nodes are arranged in the equipment area to collect environmental data every 5 seconds. Both types of data are marked with collection timestamps and stored in a real-time database.

[0068] Step 102: Dynamically couple the temperature data stream and the ambient temperature and humidity data stream across time scales to generate a heterogeneous temperature feature stream.

[0069] In step 102, dynamic coupling represents an environmental parameter compensation mechanism that is automatically adjusted according to the operating status of the equipment. The heterogeneous temperature feature stream represents a composite data stream that integrates multi-time dimension features, including second-level instantaneous fluctuations, hour-level trend changes, and monthly-level historical comparison features.

[0070] In an embodiment of the present application, first, the second-level time window is divided, the equipment temperature data is aligned with the ambient temperature and humidity according to the timestamp, and a data unit with an environmental compensation mark is generated; then the hourly window is divided according to the substation load change cycle, and the average temperature and fluctuation amplitude of each device are calculated as trend characteristics; finally, the historical data of the same period is associated, and the current trend characteristics are compared with the data range of the same season and load period in previous years to generate a heterogeneous temperature feature stream containing three time dimensions.

[0071] For example, the above-mentioned substation matches the oil temperature data of the last 5 minutes with the ambient temperature and humidity, and marks it with the "high temperature and high humidity compensation" logo; counts the characteristics of the oil temperature rise rate in the last 2 hours exceeding the historical same period; combines the oil temperature fluctuation range during the same period last summer, and generates the current composite feature data that includes environmental compensation, trend warning, and historical reference.

[0072] Step 103: Based on the heterogeneous temperature characteristic flow, dynamically adjust the association rule between the temperature anomaly of the power equipment and the operating status of the substation.

[0073] In step 103, dynamic adjustment means automatically updating the rule weight parameter according to the new feature data. The association rule means describing the mapping relationship between the abnormal temperature of the equipment and the operating status parameters such as the load rate and the aging degree.

[0074] In an embodiment of the present application, temperature trend anomalies in heterogeneous feature flows are extracted, and operating parameters such as the current substation load rate and equipment operating years are matched to calculate the similarity between the anomaly pattern and historical fault cases. When a new anomaly pattern appears, the environmental temperature and humidity influencing factors are added to the original rule base, and the rule triggering threshold is updated. After each adjustment, the validity of the rule is verified through recent fault data.

[0075] For example, when it is found that the hourly rising trend of the main transformer oil temperature does not match the load, combined with the aging coefficient of the main transformer that has been in operation for 15 years, an aging compensation term is added to the original "oil temperature - load" rule, and the warning threshold is changed from a fixed value to a dynamic value that increases with the operation years.

[0076] Step 104: Perform cross-device coupling analysis on the heterogeneous temperature feature stream through a deep belief network to generate a dynamic conduction topology.

[0077] In step 104, the deep belief network refers to a hierarchical structure composed of multiple feature processing layers, where the bottom layer processes the temperature synchronization characteristics of physically adjacent nodes of the device, the middle layer analyzes the conduction delay characteristics of cross-regional devices, and the top layer integrates the conduction intensity and spatial constraint conditions; this network extracts the implicit heat conduction mode between device groups through layer-by-layer non-linear transformation, and its inter-layer connection weights are pre-trained according to the temperature conduction characteristics of substation devices, and the final output layer generates a conduction topology connection relationship that conforms to the actual spatial layout. Cross-device coupling analysis represents a method for identifying abnormal heat conduction paths between device groups. The dynamic conduction topology represents a directed graph with devices as nodes and heat conduction intensity as edge weights.

[0078] In the embodiment of the present application, the temperature characteristics of the device group are extracted layer by layer through a deep belief network. First, the temperature synchronization of physically adjacent devices is identified, and then the conduction delay characteristics of cross-regional devices are analyzed; device pairs with conduction intensity exceeding the threshold are constructed as topological edges, and unreasonable connections in space are removed in combination with the actual positions of the devices, and finally a dynamic network reflecting the real heat conduction path is generated.

[0079] For example, it is analyzed that there is a 3-minute delay in the temperature fluctuations between the No. 1 main transformer and the adjacent 201 switch, and the conduction intensity reaches the warning value, and the actual distance between the two is within the allowable range, so a conduction edge is established in the topology; however, the equally strongly correlated No. 1 main transformer and the remote 202 switch are excluded due to the excessive distance.

[0080] Step 105: Combine the adjusted fault association rule with the dynamic conduction topology to generate a substation fault diagnosis result.

[0081] In step 105, the fault diagnosis result includes warning information on the location of the faulty device and the associated influence range.

[0082] In the embodiment of the present application, device nodes with sudden changes in conduction intensity are located in the dynamic topology, and it is checked whether their temperature characteristics conform to the aging or environmental coupling mode in the association rule; after confirming the fault source, trace the conduction path along the topological edge and calculate the risk probability of the associated devices; integrate the fault source characteristics and the conduction path analysis results to generate a diagnostic report.

[0083] For example, after diagnosing the heating fault of the bushing joint of the No. 1 main transformer, it is found along the topology that the conduction risks of the 201 switch and the 101 disconnecting switch decrease in turn, and the hierarchical diagnosis conclusion of "Abnormality (high risk) of the bushing joint of the No. 1 main transformer, and it is necessary to synchronously check the 201 switch (medium risk) and the 101 disconnecting switch (low risk)" is output.

[0084] This method constructs composite features including environmental compensation, operation trend, and historical comparison through multi-source data fusion, and uses the collaborative analysis of a dynamic rule base and a conduction topology to achieve accurate positioning of substation equipment faults and prediction of associated impacts. Compared with traditional threshold alarms, it can identify hidden conduction faults and distinguish risk levels, providing a complete evidence chain for operation and maintenance decisions. The implementation example shows its diagnostic effectiveness for the heating of the main transformer bushing and associated equipment, verifying the engineering practicability of the method.

[0085] To solve the problem that it is difficult to identify hidden thermal conduction faults between substation equipment, in some embodiments, step 104: performing cross-device coupling analysis on the heterogeneous temperature feature stream through a deep belief network to generate a dynamic conduction topology structure, including:

[0086] Step 201: Performing multi-level feature processing on the heterogeneous temperature feature stream through a deep belief network to construct an initial topology network.

[0087] In step 201, the initial topology network refers to a directed graph structure with power equipment as nodes and preliminary conduction relationships as edges, where the nodes contain device temperature feature vectors and the edge weights represent the conduction intensity without spatial verification.

[0088] In the embodiments of the present application, the deep belief network first extracts the second-level temperature synchronization features in the heterogeneous feature stream to construct the basic connections of the devices, and then combines the hourly trend similarity to calculate the preliminary conduction weights to form an initial network including all potential conduction paths.

[0089] Step 202: Based on the time persistence of the heterogeneous temperature feature stream, iteratively correct the initial topology network to generate a target topology network.

[0090] In step 202, the time persistence of the heterogeneous temperature feature stream is reflected by the change rules of temperature data in different time dimensions in the feature stream, specifically manifested as: the continuity of temperature instantaneous fluctuations within a second-level time window, the coherence of temperature trend changes within an hourly time window, and the periodic recurrence of historical temperature features in a monthly time window. These three data evolution features in different time dimensions together constitute the quantitative basis for time persistence. The target topology network refers to a conduction relationship network verified by time persistence, where its nodes represent power equipment, the edges represent stable conduction paths confirmed through multiple rounds of iterative correction, and the edge weights comprehensively reflect the conduction intensity and time stability, which is the optimized result of the initial network after weeding out the false and retaining the true.

[0091] In the embodiment of the present application, the initial network is corrected in multiple rounds: first, the conduction paths that repeatedly appear in three consecutive load cycles are screened out, then the weights are adjusted according to the change trend of the path strength, and finally the temporary connections generated by single abnormal fluctuations are removed to generate a target network reflecting the stable conduction mode.

[0092] Step 203: Perform spatial matching on the target topological network and the physical position coordinate data of each power device in the substation to generate a dynamic conduction topological structure.

[0093] In step 203, the physical position coordinate data refers to the precise installation position information of each power device in the substation in three-dimensional space, which is derived from the equipment layout diagram in the substation design drawings or the actual coordinate data obtained through laser mapping, and includes the X / Y / Z axis coordinates and azimuth angle parameters of the equipment center point. Spatial matching refers to the process of physically verifying the rationality of the network connection path and the actual installation position of the equipment.

[0094] In the embodiment of the present application, the actual distance between the devices corresponding to each edge in the target network is calculated, the connections that meet the maximum effective heat transfer distance between the substation devices are retained, and the conduction direction weight is adjusted according to the equipment arrangement orientation, and finally a dynamic conduction topology that meets both data characteristics and spatial constraints is generated.

[0095] The following is a specific example:

[0096] In a 500 kV substation implementation case, the system first collects temperature data every second through temperature sensors deployed at 12 key points such as the transformer oil temperature probe and the GIS equipment shell, and at the same time, 4 environmental monitoring nodes collect temperature and humidity data every 5 seconds. After both types of data are marked with time stamps, they are stored in the real-time database. Subsequently, the main transformer oil temperature data in the most recent 5 minutes is matched with the environmental temperature and humidity data, and a high-temperature and high-humidity compensation flag is automatically marked when the environmental temperature exceeds 35 degrees and the humidity is greater than 80%. The system statistics show that the oil temperature rise rate of the No. 1 main transformer in the most recent 2 hours reaches 2.5 degrees per hour, which is obtained by calculating the difference between the current oil temperature and the oil temperature 2 hours ago and dividing by the time interval, exceeding the average rise rate of 1.8 degrees per hour in the same load period in the same period last summer by 39%. Considering the aging coefficient of the main transformer that has been in operation for 15 years, the warning threshold of the oil temperature - load correlation rule is adjusted from the fixed value of 85 degrees to a dynamic value of 85 degrees calculated according to the formula "threshold = 80 + operating years × 0.33". The deep belief network analysis shows that there is a 3-minute delay in the temperature fluctuations between the No. 1 main transformer and the 201 switch, and the conduction intensity value is obtained by calculating the peak value of the cross-correlation function of the temperature sequences of the two devices as 0.78, exceeding the preset threshold of 0.7, and the actual distance between the two is 3 meters within the allowable range of 5 meters, so a conduction edge is established in the topology.

[0097] In the embodiments of the present application, the dynamic conduction topology constructed through multi-level feature extraction and spatio-temporal double verification can accurately identify the true heat conduction path between devices, avoid misjudgment caused by environmental interference or measurement errors, and provide a reliable conduction relationship evidence chain for fault tracing.

[0098] To improve the accuracy of conduction path analysis in substation fault diagnosis, in some embodiments, step 203: the spatial matching of the target topology network with the physical position coordinate data of each power device in the substation to generate a dynamic conduction topology structure includes:

[0099] Step 301: Obtain the physical position coordinate data of each power device in the substation.

[0100] In the embodiments of the present application, the plane layout coordinates of main devices such as transformers and circuit breakers are extracted from the substation design drawings, and combined with the height data collected by the on-site laser rangefinder to construct a three-dimensional coordinate data set including the spatial position and orientation angle of the devices.

[0101] Step 302: Compare the connection paths between the nodes in the target topology network with the physical position coordinate data to filter out all initial effective conduction connection paths with lengths less than the preset spatial distance threshold.

[0102] In step 302, the comparison process is specifically as follows: calculate the theoretical conduction path length between each pair of device nodes in the topology network, and compare the difference with the actual installation distance measured in the physical position coordinates. When the absolute value of the difference is less than the preset tolerance threshold, retain the path. Exemplarily, for a circuit breaker A and a disconnector B in a 220 kV substation, a strong conduction connection is shown in the topology network, but the measured installation distance in the actual coordinates is 8 meters (exceeding the preset threshold of 5 meters), then it is determined that this conduction path is invalid. The initial effective conduction connection path refers to the connection path between devices in the target topology network that simultaneously satisfies the following two conditions: one is that the physical distance between nodes is less than the preset spatial distance threshold, and the other is that the conduction weight value is still higher than the preset conduction intensity threshold after azimuth correction. "Effective" means a conduction path that simultaneously meets three requirements: one is that the conduction intensity of the topology network meets the standard, the second is that the physical space distance conforms to the actual installation position between devices, and the third is that the weight value after correction by the electrical circuit relationship and heat dissipation orientation exceeds the preset threshold. The three together constitute the effectiveness determination criterion.

[0103] In the embodiments of the present application, calculate the theoretical conduction distance between each device node in the target topology network, compare it with the actual installation distance in the physical coordinate data, retain the connection paths with actual distances less than the preset maximum effective heat transfer distance, and mark the remaining paths as to-be-verified status.

[0104] Step 303: Modify the weights of each of the initial effective conduction connection paths according to the electrical circuit relationship and the orientation of the heat dissipation areas of the electrical equipment in the substation. After the modification is completed, delete the redundant paths that do not meet the preset conduction conditions to generate the target effective conduction connection paths.

[0105] In step 303, the electrical circuit relationship refers to the combination relationship of equipment that forms a current path through electrical connections such as cables and busbars in the substation, and this relationship is derived from the electrical main wiring diagram and equipment connection method clearly marked in the substation design drawings. The orientation of the heat dissipation area refers to the directional position relationship of the temperature influence areas divided according to the heat dissipation characteristics and spatial layout of the equipment in the substation, and it is derived from the equipment heat dissipation partition data divided in the substation thermal design specifications. The preset conduction condition refers to a composite standard for determining the effectiveness of the conduction path, including three dimensions: the conduction intensity threshold, the lower limit of the electrical circuit enhancement coefficient, and the upper limit of the heat dissipation area attenuation coefficient. Only the paths that still have a conduction intensity higher than the set threshold after being corrected by electrical enhancement and heat dissipation attenuation, and whose physical distance and orientation conform to the engineering reality, will be retained. The target effective conduction connection path refers to the set of true and effective heat conduction relationships between equipment that have been finally confirmed after spatial distance screening, electrical circuit weight enhancement, heat dissipation area attenuation correction, and orientation verification. Each path of it meets the preset conduction conditions and can accurately reflect the actual heat conduction characteristics between the substation equipment.

[0106] In the embodiment of the present application, first, identify the equipment pairs belonging to the same electrical circuit in the initial paths, and enhance their conduction weights according to the tightness of the circuit connection; then, according to the position of the equipment in the heat dissipation area, attenuate the weights of the cross-region paths according to the thermal resistance characteristics between regions; finally, delete the redundant paths with weights lower than the threshold after comprehensive modification.

[0107] Step 304: Map the target effective conduction connection paths to the actual layout orientation in the physical position coordinate data to generate a dynamic conduction topology.

[0108] In step 304, the actual layout orientation is the specific installation position coordinates of the equipment in the substation, and the heat dissipation area orientation is the functional area divided based on the actual layout orientation by superimposing the heat dissipation characteristics of the equipment. The two are the relationship between the basic physical position and the derived functional partition.

[0109] In the embodiment of the present application, calibrate the orientation of the corrected effective paths with the actual coordinates of the equipment to ensure that the conduction direction is consistent with the physical orientation of the equipment, and eliminate the connections with too large direction deviations, and finally generate a conduction topology that conforms to the real spatial layout.

[0110] The following is a specific example:

[0111] In the implementation case of a certain 500 kV substation, the system obtains the physical location coordinate data of equipment such as the No. 1 main transformer, the 201 switch, and the 202 switch. The measured distance between the No. 1 main transformer and the 201 switch is 3 meters, and the distance from the 202 switch is 8 meters. Compare the initial conduction connection path between the No. 1 main transformer and the 201 switch in the target topological network with the coordinate data. The path length of 3 meters is less than the preset spatial distance threshold of 5 meters and is retained as the initial effective conduction connection path, while the path between the No. 1 main transformer and the 202 switch is screened out because the 8 meters exceeds the limit. According to the electrical main wiring diagram, it is confirmed that the No. 1 main transformer and the 201 switch belong to the 220 kV Bus I circuit, and the electrical circuit enhancement coefficient 1.2 is given. At the same time, since they are located in different heat dissipation areas respectively, the heat dissipation attenuation coefficient is calculated as cos60° = 0.5 according to the azimuth angle of 60 degrees between the areas. Multiply the initial conduction intensity of 0.78 by the enhancement coefficient of 1.2 and then by the attenuation coefficient of 0.5 to obtain the comprehensive correction value of 0.468, which is lower than the preset conduction condition threshold of 0.7. However, considering that this connection has persisted in the historical data, the system starts the exception handling mechanism, temporarily lowers the threshold to 0.45, and then retains this path. Finally, match the corrected effective path with the actual orientation of the equipment. After confirming that the orientation of the bushing of the No. 1 main transformer is consistent with the orientation of the connection part of the 201 switch, establish this conduction edge in the dynamic conduction topology for subsequent diagnostic analysis.

[0112] In the embodiment of the present application, the dynamic conduction topology generated by spatial matching not only retains the conduction relationship driven by data but also conforms to the physical layout constraints of the substation, effectively solving the problem of false conduction paths in traditional methods and providing a more reliable topological basis for fault location.

[0113] In order to further improve the accuracy of the conduction path weight correction, in some embodiments, step 303: the weight correction of each of the initial effective conduction connection paths according to the electrical circuit relationship and the heat dissipation area orientation of each power equipment in the substation includes:

[0114] Step 401: Identify the equipment node pairs belonging to the same electrical connection group among all the initial effective conduction connection paths, and perform the same-circuit enhancement processing on the conduction weight values of the equipment node pairs according to a preset ratio. The electrical connection group is a set of equipment groups divided according to the electrical circuit relationship.

[0115] In step 401, the electrical connection group is a subset of the power equipment in the substation that are directly connected by cables or busbars. It is a functional combination of the power equipment divided according to the electrical connection relationship. The two are in a part-whole relationship. The same-loop enhancement process refers to the operation of amplifying the weight of the conduction path between the equipment within the same electrical loop. The equipment node pair refers to the combination of two power equipment nodes with potential heat conduction relationship in the dynamic conduction topology structure. Each node corresponds to a specific power equipment, and the node pair indicates that there may be a heat transfer path between these two equipment. The conduction weight value represents a quantitative index of the heat conduction intensity between the two equipment in the node pair, which is derived from the analysis result of the deep belief network for the heterogeneous temperature feature flow, and the initial conduction intensity estimate is obtained by calculating the cross-correlation and trend synchronization of the temperature change sequences of the two equipment.

[0116] In the embodiment of the present application, first, the connection relationship of the equipment at each voltage level is extracted from the electrical main wiring diagram of the substation, and the directly electrically connected equipment is divided into connection groups. Then, the weight of the conduction path between the equipment within the group is multiplied by a preset enhancement coefficient to reflect the promotion effect of the electrical loop on heat conduction.

[0117] Step 402: Calculate the conduction attenuation factor of the cross-interval equipment node pair based on the azimuth angle data of the adjacent heat dissipation intervals in the heat dissipation area orientation.

[0118] In step 402, the adjacent heat dissipation intervals refer to the sub-regions in the heat dissipation area that are physically adjacent and have the possibility of heat conduction. The heat dissipation area is divided into several heat dissipation intervals according to the heat dissipation characteristics and spatial positions of the equipment. The adjacent intervals have a common boundary or overlapping area. The azimuth angle data refers to the angle value between the line connecting the center points of two adjacent heat dissipation intervals and the normal direction of the main heat dissipation surface of the equipment, which is derived from the azimuth coordinates of the equipment layout in the substation design drawings, and the specific angle degree is obtained through spatial geometric calculation. Cross-interval means that the two equipment in the equipment node pair are located in different heat dissipation intervals, and these two heat dissipation intervals satisfy the adjacent relationship, and their heat conduction needs to cross the interval boundary. The cross-interval equipment node pair and the equipment node pair of the same electrical connection group are two classification concepts of independent dimensions: the former is based on the thermodynamic division of the heat dissipation area, and the latter is based on the topological structure of the electrical connection. The same pair of equipment nodes may have both attributes at the same time (such as being in different heat dissipation intervals but belonging to the same electrical loop), or may only have one of the attributes (such as in the same interval but different loops). The two are cross-correlated through the physical position and electrical connection relationship of the equipment but are not the same concept. The conduction attenuation factor refers to a parameter that reflects the degree of attenuation of the heat conduction path intensity when crossing different heat dissipation areas, and is inversely proportional to the azimuth angle between the regions.

[0119] In the embodiments of the present application, the heat dissipation area is divided according to the equipment layout diagram, and the included angle between the connecting line of the regional center points and the equipment orientation is calculated. The larger the included angle, the smaller the attenuation factor, which reflects the hindering effect of the difference in heat dissipation conditions on conduction.

[0120] Step 403: Multiply the conduction weight value after the same-loop enhancement process by the conduction attenuation factor to generate a comprehensive corrected weight value.

[0121] In step 403, the comprehensive corrected weight value refers to the final conduction intensity evaluation value considering both the electrical loop enhancement and the heat dissipation attenuation.

[0122] In the embodiments of the present application, multiply the path weight after the same-loop enhancement process by the corresponding attenuation factor to obtain a corrected weight that reflects both the electrical connection characteristics and conforms to the laws of thermodynamics.

[0123] Step 404: Perform weight correction on each of the initial effective conduction connection paths according to the comprehensive corrected weight value.

[0124] In step 404, the weight correction is a process of adjusting the conduction path strength according to the comprehensive evaluation result.

[0125] In the embodiments of the present application, compare the corrected weight value with a preset threshold, retain the effective paths and update their conduction strength values, and delete the non-compliant paths to complete the network optimization.

[0126] The following is a specific example:

[0127] In the implementation case of a certain 500 kV substation, the system first identifies that the No. 1 main transformer and the 201 switch in the initial effective conduction connection paths belong to the same electrical connection group, which is determined according to the electrical main wiring diagram of the 220 kV I busbar. The system performs the same-loop enhancement process on the initial conduction intensity of 0.78 according to the standard enhancement coefficient of 1.2 for the equipment of this voltage level, and obtains an intermediate value of 0.936. Then, based on the heat dissipation area layout diagram, it is determined that the No. 1 main transformer is located in the main transformer heat dissipation area, and the 201 switch is located in the switch heat dissipation area. The included angle between the connecting line of the centers of the two areas and the normal line of the equipment heat dissipation surface is 60 degrees, and the conduction attenuation factor is calculated according to the formula cos60° = 0.5. Multiply the enhanced conduction intensity of 0.936 by the attenuation factor of 0.5 to obtain a comprehensive corrected weight value of 0.468.

[0128] In the embodiments of the present application, the conduction weight is corrected by the dual factors of the electrical loop and the heat dissipation area, which not only reflects the unique electrical connection characteristics of the substation but also considers the influence of the actual heat dissipation conditions, making the generated conduction topology more conform to the true heat conduction law between the equipment and providing a more reliable path basis for fault diagnosis.

[0129] In some embodiments, in order to more accurately construct the initial conduction topology network between substation devices, step 201: performing multi-level feature processing on the heterogeneous temperature feature stream through a deep belief network to construct an initial topology network, including:

[0130] Step 501: Through the first-level processing module of the deep belief network, based on the group units divided according to the physical position relationship of each power device in the substation, extract the temperature fluctuation synchronization features corresponding to the devices inside the group units from the heterogeneous temperature feature stream.

[0131] In step 501, the physical position relationship refers to the relative position and connection relationship between devices. The physical position coordinate data refers to the specific coordinate position data of the devices in the substation. The former is used to establish group units, and the latter is used to verify the physical rationality of the conduction path. A group unit is an analysis unit composed of multiple power devices with adjacent physical positions and potential heat conduction. Its form is a device combination divided according to the device installation position and electrical connection relationship, and is used to analyze local temperature conduction characteristics. The internal devices specifically refer to the subset of devices to be analyzed within the group unit, which are part of the power devices in the substation with position relevance. The two are in an inclusion relationship. The temperature fluctuation synchronization feature refers to the correlation feature of the temperature changes of the devices within the group in terms of time and amplitude.

[0132] In the embodiments of the present application, first, the devices with adjacent installation positions are divided into groups according to the spatial layout of substation devices, and then the time-domain correlation of the temperature data of the devices in each group is analyzed to extract the device combinations with synchronous rising / falling trends and their quantization values of synchronization degree.

[0133] Step 502: Through the second-level processing module, perform superposition processing on the temperature fluctuation synchronization feature and the substation environment temperature and humidity data stream to generate an intermediate feature set.

[0134] In step 502, the superposition processing refers to the process of fusing and calculating the device temperature features and environmental parameters. The intermediate feature set is a transition data set containing the temperature conduction features after environmental compensation.

[0135] In the embodiments of the present application, align the extracted temperature synchronization features with the environmental temperature and humidity data in time, compensate and correct the temperature fluctuation amplitude according to the environmental parameters, and generate a temperature conduction feature set containing environmental influencing factors.

[0136] Step 503: Through the third-level processing module, screen out the temperature correlation features exceeding the preset conduction threshold from the intermediate feature set, and construct an initial topology network according to the conduction intensity relationship between each power device in the substation corresponding to the temperature correlation features. The temperature correlation feature refers to the temperature correlation feature between each power device in the substation.

[0137] In step 503, the conduction threshold is the lowest characteristic intensity standard for determining whether there is a significant heat conduction relationship between devices. The temperature correlation characteristic refers to the characteristic index reflecting the possibility of heat transfer between devices. The devices include electrical devices both between the internal devices within the group unit and between different group units, including both internal device conduction and cross-group device conduction. The conduction intensity relationship is a quantitative index of the heat transfer ability between devices, which is derived from the analysis result of the deep belief network on the temperature characteristic flow and is comprehensively obtained by calculating the time-domain correlation, trend synchronization, and amplitude matching degree of the device temperature sequence, reflecting the strength of the actual heat conduction between devices.

[0138] In the embodiment of the present application, strong correlation device pairs exceeding the threshold are screened out from the intermediate feature set, and network edge weights are constructed according to their correlation intensity values. Taking the devices as nodes and the conduction relationship as edges, an initial topological network is formed.

[0139] The following is a specific example:

[0140] In a 500 kV substation implementation case, the system first divides the devices within 3 meters of each other, such as the No. 1 main transformer, 201 switch, and 101 disconnecting switch, into a group unit according to the device layout diagram. By analyzing the temperature data of each device in the group in the recent 24 hours through a deep belief network, the cross-correlation coefficient between the temperature sequences of the No. 1 main transformer and the 201 switch is calculated to reach 0.75. This value is obtained by calculating the correlation coefficients of the two device temperature sequences at different time lags through a formula and taking the maximum value. Subsequently, the synchronization characteristic of 0.75 is superimposed and processed with the ambient temperature and humidity data. When the ambient temperature is 35 degrees and the humidity is 85%, the corrected characteristic value of 0.82 is calculated according to the compensation formula: synchronization characteristic value = original value × [1 + 0.1 × (ambient temperature - 30) / 5] × [1 + 0.05 × (humidity - 80) / 5]. The system screens out the temperature correlation characteristics exceeding the preset threshold of 0.7 from all group characteristics. Among them, the 0.82 of the No. 1 main transformer - 201 switch meets the standard, while the 0.68 of the No. 1 main transformer - 101 disconnecting switch is screened out. According to the qualified characteristic value, an initial topological network with the No. 1 main transformer and the 201 switch as nodes and an edge weight of 0.82 is constructed, and this weight directly uses the compensated characteristic value.

[0141] In the embodiment of the present application, the initial topological network constructed through multi-level feature processing not only considers the group characteristics of the physical locations of the devices but also integrates the environmental influence factors, and can effectively identify the potential heat conduction relationships between substation devices, laying a foundation for subsequent accurate diagnosis.

[0142] In order to more accurately construct a temperature characteristic model of substation equipment, in some embodiments, step 102: dynamically coupling the temperature data stream and the ambient temperature and humidity data stream across time scales to generate a heterogeneous temperature characteristic stream includes:

[0143] Step 601: Divide the time window into seconds based on the temperature sampling period of each power device, synchronously match the ambient temperature and humidity data with the temperature data stream within the time window, and generate an instantaneous temperature data unit.

[0144] In step 601, the second-level time window refers to the minimum analysis period synchronized with the temperature sensor sampling period. The instantaneous temperature data unit is a data structure containing the instantaneous value of the device temperature after environmental compensation and its rate of change.

[0145] In an embodiment of the present application, the time window is divided based on the sampling interval of the temperature sensor, and within each window, the device temperature reading is accurately matched with the ambient temperature and humidity monitoring value according to the collection time, and the temperature value is compensated in real time according to the environmental parameters to generate a temperature data unit with an environmental mark.

[0146] Step 602: Divide the hourly time window according to the load change cycle of the substation, and perform accumulation processing on the instantaneous temperature data units to extract the temperature trend change characteristics of each power equipment within the hourly time window.

[0147] In step 602, the hourly time window refers to the analysis period that reflects the typical load fluctuation cycle of the substation. The temperature trend change characteristics include indicators such as the equipment temperature mean, extreme value and change rate.

[0148] In an embodiment of the present application, the windows are divided according to the daily load curve of the substation, the distribution characteristics of the equipment temperature data in each window are counted, the temperature change trend slope is calculated, the abnormal fluctuation period is identified, and a trend feature set reflecting the change of the equipment operating status is formed.

[0149] Step 603: Divide the monthly time window in combination with the historical operation data of the substation, associate and map the temperature trend change characteristics with the historical temperature fluctuation range in the same period under the same seasonal conditions, and generate a temperature feature set.

[0150] In step 603, the monthly time window refers to the analysis period that covers the seasonal operation cycle of the device. The same season means that the current analysis period has the same seasonal characteristics as the historical data (such as the high-temperature period in summer / low-temperature period in winter), and the historical same period specifically refers to the historical time period that is close to the current analysis period in terms of calendar date (such as the same month last year ± 15 days). The two jointly define the time range for comparing historical data. The temperature feature set refers to a composite feature data set that integrates the comparative analysis results of the current device temperature trend features and historical same-period data, including dimensional information such as the relative position, deviation degree, and seasonal change law of the current temperature feature in the historical same period, and is used to reflect the abnormal degree and seasonal adaptability characteristics of the device temperature change during long-term operation.

[0151] In the embodiment of the present application, historical data is organized on a monthly basis, the current trend features are compared with historical same-period data, the percentile of the current value in the historical distribution is calculated, and a temperature feature set including seasonal adaptability is generated.

[0152] Step 604: Hierarchically integrate the instantaneous temperature data unit, the temperature trend change feature, and the temperature feature set according to the time stamp to form a heterogeneous temperature feature stream.

[0153] In step 604, hierarchical integration refers to the process of organically combining features in different time dimensions along a unified time axis. The specific hierarchical integration process is as follows: The second-level instantaneous temperature data unit corresponding to the same time stamp is used as the basic layer, the temperature trend change feature of the corresponding hourly window is superimposed as the intermediate layer, and then the historical temperature feature set of the monthly window is associated as the reference layer to form a three-dimensional nested data structure. Exemplarily, the monitoring data of a transformer at 10:30:00 on a certain day in a certain month and year is integrated as follows: basic layer (the temperature value of 42.3 °C after environmental correction at this second), intermediate layer (the temperature rise rate of 0.5 °C / min in the 10:00 - 11:00 period), reference layer (the temperature benchmark range of 38 - 45 °C in the same period of the same month last year under the same working conditions). The three are associated through a unified time stamp to form a complete feature unit.

[0154] In the embodiment of the present application, based on the device operation time axis, the second-level instantaneous value, the hourly trend feature, and the monthly historical comparison feature are vertically associated to construct a composite feature stream including complete time dimension information.

[0155] The following is a specific example:

[0156] In the implementation case of a certain 500 kV substation, the system first divides the second-level time window at intervals of 1 second, and matches the 85.3 degrees collected by the oil temperature probe of the No. 1 main transformer with the temperature of 35 degrees and humidity data of 80% recorded by the environmental monitoring node at the same moment. According to the compensation formula "compensated temperature = original temperature + 0.2×(ambient temperature - 30) - 0.1×(humidity - 75)", the compensated temperature value of 86.1 degrees is calculated, and an instantaneous temperature data unit with an environmental compensation mark is generated. Subsequently, the system divides the time window according to a 2-hour load cycle, and counts that the oil temperature of the No. 1 main transformer rises from 82.5 degrees to 87.3 degrees within this window. The calculated rising rate is 2.4 degrees per hour, which is obtained by subtracting the temperature at the start time from the temperature at the end time of the window and dividing by the time interval. The system retrieves the historical data of the same load period in the same period last summer, and calculates that the average rising rate of the oil temperature during this period is 1.9 degrees per hour, and the standard deviation is 0.3 degrees. It is determined that the current rate deviates from the historical mean by 1.67 standard deviations, and a risk feature including the degree of trend deviation is generated. Finally, the features of the three dimensions of the second-level compensation value of 86.1 degrees, the hourly rising rate of 2.4 degrees / hour, and the historical deviation degree of 1.67σ are integrated according to the unified timestamp to form a complete heterogeneous temperature feature stream.

[0157] In the embodiment of the present application, the heterogeneous temperature feature stream constructed by feature fusion at multiple time scales captures both the instantaneous fluctuations of the equipment temperature and reflects the long-term operation rules, providing a comprehensive and three-dimensional temperature feature representation for fault diagnosis, and significantly improving the coverage of the spatio-temporal dimension of the analysis.

[0158] In order to more accurately locate the substation fault source and evaluate the associated risks, in some embodiments, step 105: generating a substation fault diagnosis result by combining the adjusted fault association rule with the dynamic conduction topology structure, includes:

[0159] Step 701: Mark the device nodes in the dynamic conduction topology structure whose conduction intensity exceeds the preset warning threshold as suspicious fault source nodes, and extract the temperature anomaly feature patterns corresponding to the suspicious fault source nodes.

[0160] In step 701, the suspicious fault source node refers to a device node in the conduction topology with abnormal conduction intensity but not yet verified. The temperature anomaly feature pattern includes the temperature deviation features of this node at the second-level, hourly, and monthly time scales.

[0161] In the embodiment of the present application, first scan the weight values of all edges in the dynamic conduction topology, screen out the device nodes that exceed the preset warning value, and then extract the temperature anomaly features of this node in each time dimension from the heterogeneous temperature feature stream to form a set of fault patterns to be verified.

[0162] Step 702: Match and verify the temperature anomaly feature pattern with the adjusted fault association rules. When the match verification result indicates that the suspicious fault source node simultaneously conforms to the equipment aging gradient change law and the environmental temperature and humidity coupling condition in the adjusted fault association rules, confirm the suspicious fault source node as the diagnosed fault source node.

[0163] In step 702, the match verification process is specifically as follows: double judgment is carried out by comparing the key indicators in the temperature anomaly feature pattern with the preset thresholds of the fault association rules. Among them, the compliance of the equipment aging gradient change law is manifested as the temperature anomaly value showing an exponential growth trend with the operation duration (for example, the temperature rise rate of a certain circuit breaker exceeds 0.8 °C / month after 3 years of operation), and the compliance of the environmental temperature and humidity coupling condition is manifested as a lagged association between the temperature fluctuation and the change of environmental parameters (for example, when the humidity > 70%, the equipment temperature continuously exceeds the reference value by 15% for 2 hours). When the temperature anomaly feature pattern presents both of these features at the same time, it is determined to meet the conditions. The equipment aging gradient change law refers to the typical change trend of temperature anomaly with the increase of equipment operation years. The environmental temperature and humidity coupling condition refers to the correlation characteristics between temperature anomaly and environmental parameters. The diagnosed fault source node refers to the suspicious node that has passed the double verification of the fault association rules, and its temperature anomaly characteristics simultaneously meet the equipment aging gradient change law and the environmental temperature and humidity coupling condition, and is confirmed as the power equipment node with a real fault, which is the core positioning target in the fault diagnosis result.

[0164] In the embodiment of the present application, the temperature characteristics of the suspicious node are compared with the aging mode in the fault rule library to check whether its change trend conforms to the equipment aging curve, and at the same time verify whether the temperature anomaly is synchronized with the change of environmental parameters. Only the nodes that pass the double verification are confirmed as the real fault sources.

[0165] Step 703: Generate associated equipment risk warning information according to the conduction path strength between the diagnosed fault source node and other equipment nodes in the dynamic conduction topology structure.

[0166] In step 703, other devices refer to all power device nodes in the dynamic conduction topology structure except the diagnosed fault source node, specifically the power devices in the substation that have a thermal conduction correlation relationship with the fault source. The conduction path strength is a quantitative index of the thermal conduction relationship between the diagnosed fault source node and other nodes, which is derived from the weight value of the edge in the dynamic conduction topology structure and obtained through the analysis of the temperature correlation characteristics between devices by the deep belief network, reflecting the attenuation degree of the fault heat conduction along the path. The conduction path strength specifically refers to the quantitative value of the conduction relationship between two specific device nodes in the dynamic conduction topology structure. The conduction intensity is the specific manifestation of the conduction path strength on a single path. The two are the relationship between the overall topological attribute and the single-path attribute, not the same concept. The associated device risk warning information refers to the devices that may be affected by the fault source along the conduction path and the assessment of their risk levels.

[0167] In the embodiment of the present application, starting from the diagnosed fault source, traverse the adjacent nodes along the conduction topology edge, divide the risk levels according to the attenuation degree of the conduction intensity, and generate a multi-level warning list including directly associated devices and indirectly associated devices.

[0168] Step 704: Combine the positioning information of the diagnosed fault source node with the associated device risk warning information to generate a substation fault diagnosis result.

[0169] In the embodiment of the present application, combine the location information and fault type of the diagnosed fault source with the associated device risk list to form a complete diagnosis report including suggestions on the processing priority.

[0170] The following is a specific example:

[0171] In a 500kV substation implementation case, the system first found that the conduction strength of the No. 1 main transformer node in the dynamic conduction topology reached 0.85, exceeding the preset alarm threshold of 0.8, and marked it as a suspected fault source node. The temperature characteristics of the node were extracted and showed that the oil temperature of the second window was 87 degrees, which was 88.2 degrees after environmental compensation. The calculation method was the original temperature plus the environmental temperature compensation value 0.2×(35-30) minus the humidity compensation value 0.1×(80-75); the hourly window oil temperature rise rate was 2.8 degrees / hour, which was calculated by subtracting the temperature two hours ago from the current temperature and dividing it by the time difference; it deviated from the historical average of 1.9 degrees / hour by 1.67 standard deviations, and the standard deviation was calculated using the historical data statistical method. The system matched the above characteristics with the adjusted fault association rules and confirmed that it met the gradient change curve formula y=0.1x+1.2 for the aging of the main transformer bushing after 15 years of operation, where x is the operating years and changes synchronously with the high temperature and high humidity conditions of the environment. Therefore, the No. 1 main transformer node is confirmed as the confirmed fault source node. According to the conduction topology analysis, the conduction path strength from the No. 1 main transformer to the 201 switch is 0.62, and to the 101 knife switch is 0.45. These values ​​are directly taken from the topological network edge weights. The final system combination generates the diagnosis results: the No. 1 main transformer casing joint abnormality needs to be handled immediately, the 201 switch connector needs to be checked within 24 hours, and the 101 knife switch needs to be observed within 72 hours.

[0172] In the embodiment of the present application, through the dual verification mechanism of conduction topology and fault rules, the accuracy of fault location is guaranteed and the quantitative assessment of associated risks is realized, providing a complete decision-making basis for substation operation and maintenance from the source of the fault to the scope of impact, and significantly improving the systematicness and predictability of fault handling.

[0173] Figure 2 A schematic diagram of the structure of a substation fault intelligent analysis and diagnosis system based on big data provided in an embodiment of the present application, such as Figure 2 As shown, the system includes:

[0174] The acquisition module 21 is used to acquire the temperature data stream of the surface and interior of each power equipment in the substation and the ambient temperature and humidity data stream of the substation in the operating environment of the substation.

[0175] The coupling module 22 is used to dynamically couple the temperature data stream and the ambient temperature and humidity data stream across time scales to generate a heterogeneous temperature feature stream.

[0176] The adjustment module 23 is used to dynamically adjust the association rule between the abnormal temperature of the power equipment and the operating status of the substation based on the heterogeneous temperature characteristic flow.

[0177] The analysis module 24 is used to perform cross-device coupling analysis on the heterogeneous temperature characteristic flow through a deep belief network to generate a dynamic conduction topology structure.

[0178] A generation module 25, configured to combine the adjusted fault association rules with the dynamic conduction topology to generate a substation fault diagnosis result.

[0179] Figure 2 The described intelligent analysis and diagnosis system for substation faults based on big data can execute Figure 1 The described intelligent analysis and diagnosis method for substation faults based on big data in the illustrated embodiment, and its implementation principle and technical effects will not be elaborated further. For the intelligent analysis and diagnosis system for substation faults based on big data in the above embodiments, the specific manners in which each module and unit perform operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0180] In a possible design, Figure 2 The intelligent analysis and diagnosis system for substation faults based on big data in the illustrated embodiment can be implemented as a computing device, such as Figure 3 shown, the computing device may include a storage component 31 and a processing component 32;

[0181] 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.

[0182] The processing component 32 above Figure 1 The intelligent analysis and diagnosis method for substation faults based on big data in the illustrated embodiment.

[0183] Among them, the processing component 32 may 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 may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components for executing the above method.

[0184] The storage component 31 is configured to store various types of data to support the operations of the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device 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 memory, flash memory, magnetic disks or optical discs.

[0185] Of course, the computing device may necessarily further include other components, such as an input / output interface, a display component, a communication component, etc.

[0186] The input / output interface provides an interface between the processing component and the peripheral interface module, and the above-mentioned peripheral interface module can be an output device, an input device, etc.

[0187] The communication component is configured to facilitate communication between the computing device and other devices in a wired or wireless manner, etc.

[0188] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. At this time, the computing device can refer to a cloud server, and the above-mentioned processing component, storage component, etc. can be basic server resources leased or purchased from a cloud computing platform.

[0189] The embodiment of the present application further provides a computer storage medium storing a computer program, and when the computer program is executed by a computer, it can implement the above-mentioned Figure 1 intelligent analysis and diagnosis method for substation faults based on big data shown in the embodiment.

[0190] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0191] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without creative effort.

[0192] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0193] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An intelligent analysis and diagnosis method for substation faults based on big data, characterized in that, Including: In the substation operation environment, obtaining the temperature data streams on the surface and inside of each power equipment in the substation and the ambient temperature and humidity data stream of the substation; Performing dynamic coupling of the temperature data stream and the ambient temperature and humidity data stream across time scales to generate a heterogeneous temperature feature stream; Based on the heterogeneous temperature feature stream, dynamically adjusting the association rules between power equipment temperature anomalies and substation operation states; Performing cross-device coupling analysis on the heterogeneous temperature feature stream through a deep belief network to generate a dynamic conduction topology; Combining the adjusted fault association rules and the dynamic conduction topology to generate a substation fault diagnosis result.

2. The method according to claim 1, wherein The performing cross-device coupling analysis on the heterogeneous temperature feature stream through a deep belief network to generate a dynamic conduction topology includes: Performing multi-level feature processing on the heterogeneous temperature feature stream through a deep belief network to construct an initial topology network; Based on the time persistence of the heterogeneous temperature feature stream, iteratively correcting the initial topology network to generate a target topology network; Performing spatial matching between the target topology network and the physical position coordinate data of each power equipment in the substation to generate a dynamic conduction topology.

3. The method according to claim 2, characterized in that, The performing spatial matching between the target topology network and the physical position coordinate data of each power equipment in the substation to generate a dynamic conduction topology includes: Obtaining the physical position coordinate data of each power equipment in the substation; Comparing the connection paths between nodes in the target topology network with the physical position coordinate data to filter out all initial effective conduction connection paths with lengths less than a preset spatial distance threshold; Performing weight correction on each of the initial effective conduction connection paths according to the electrical circuit relationship and the heat dissipation area orientation of each power equipment in the substation. After the correction is completed, deleting redundant paths that do not meet the preset conduction conditions to generate target effective conduction connection paths; Mapping the target effective conduction connection paths to the actual layout orientation in the physical position coordinate data to generate a dynamic conduction topology.

4. The method according to claim 3, characterized in that The performing weight correction on each of the initial effective conduction connection paths according to the electrical circuit relationship and the heat dissipation area orientation of each power equipment in the substation includes: Identifying device node pairs belonging to the same electrical connection group among all the initial effective conduction connection paths, and performing same-circuit enhancement processing on the conduction weight values of the device node pairs according to a preset ratio. The electrical connection group is a set of device groups divided according to the electrical circuit relationship; Based on the azimuth angle data between adjacent heat dissipation intervals in the heat dissipation area orientation, calculating the conduction attenuation factor of cross-interval device node pairs; Performing a multiplication operation on the conduction weight value after the same-circuit enhancement processing and the conduction attenuation factor to generate a comprehensive correction weight value; According to the comprehensive correction weight value, performing weight correction on each of the initial effective conduction connection paths.

5. The method according to claim 2, characterized in that, The performing multi-level feature processing on the heterogeneous temperature feature stream through a deep belief network to construct an initial topology network includes: By using the first level processing module of the deep belief network, the temperature fluctuation synchronization characteristics corresponding to the internal equipment of the group unit are extracted from the heterogeneous temperature feature stream based on the group units divided by the physical position relationship of each power equipment in the substation; The temperature fluctuation synchronization feature is superimposed on the substation environment temperature and humidity data stream through the second-level processing module to generate an intermediate feature set; The temperature correlation features exceeding the preset conduction threshold are screened out from the intermediate feature set through the third-level processing module, and an initial topology network is constructed according to the conduction strength relationship between the power equipment in the substation corresponding to the temperature correlation features. The temperature correlation features refer to the temperature correlation features between the power equipment in the substation.

6. The method according to claim 1, wherein The step of dynamically coupling the temperature data stream and the ambient temperature and humidity data stream across time scales to generate a heterogeneous temperature feature stream includes: Divide the second-level time window based on the temperature sampling period of each power device, synchronously match the ambient temperature and humidity data with the temperature data stream within the second-level time window, and generate an instantaneous temperature data unit; Divide the hourly time window according to the load change cycle of the substation, and perform accumulation processing on the instantaneous temperature data unit to extract the temperature trend change characteristics of each power equipment within the hourly time window; Combine the historical operation data of the substation to divide the monthly time window, associate and map the temperature trend change characteristics with the historical temperature fluctuation range under the same seasonal conditions, and generate a temperature feature set; The instantaneous temperature data unit, the temperature trend change feature and the temperature feature set are hierarchically integrated according to timestamps to form a heterogeneous temperature feature stream.

7. The method according to claim 1, characterized in that, The combining the adjusted fault association rules with the dynamic conduction topology structure to generate a substation fault diagnosis result includes: Marking the device nodes whose conduction strength in the dynamic conduction topology exceeds the preset alarm threshold as suspected fault source nodes, and extracting the temperature anomaly characteristic pattern corresponding to the suspected fault source nodes; The temperature anomaly characteristic pattern is matched and verified with the adjusted fault association rule. When the matching verification result indicates that the suspected fault source node meets both the equipment aging gradient change law and the environmental temperature and humidity coupling condition in the adjusted fault association rule, the suspected fault source node is confirmed as a confirmed fault source node; Generate risk warning information of associated equipment according to the strength of the conduction path between the confirmed fault source node and other equipment nodes in the dynamic conduction topology structure; The location information of the confirmed fault source node is combined with the risk warning information of the associated equipment to generate a substation fault diagnosis result.

8. An intelligent analysis and diagnosis system for substation faults based on big data, characterized in that, include: An acquisition module is used to acquire the surface and internal temperature data streams of each power equipment in the substation and the ambient temperature and humidity data streams of the substation in the substation operating environment; A coupling module, used for dynamically coupling the temperature data stream and the ambient temperature and humidity data stream across time scales to generate a heterogeneous temperature feature stream; An adjustment module, used for dynamically adjusting the association rules between power equipment temperature anomalies and substation operation status based on the heterogeneous temperature characteristic stream; An analysis module for performing cross-device coupling analysis on the heterogeneous temperature feature stream through a deep belief network to generate a dynamic conduction topology; A generation module for combining the adjusted fault association rules with the dynamic conduction topology to generate a substation fault diagnosis result.

9. A computing device, characterized in that, It includes 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 implement a big data-based intelligent analysis and diagnosis method for substation faults as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, A computer program is stored, and when the computer program is executed by a computer, it implements a big data-based intelligent analysis and diagnosis method for substation faults as described in any one of claims 1 to 7.

Images