Bus duct fault detection method and system based on virtual reality modeling

Abstract

The invention discloses a bus duct fault detection method and system based on virtual reality modeling, and relates to the technical field of electrical fault detection, and the method comprises the steps: carrying out the calibration of monitoring points according to a target bus duct arrangement scheme and a target application building scene, and obtaining a calibrated monitoring point array; traversing the calibrated monitoring point array, arranging a multi-source sensor group, performing communication connection with the virtual reality modeling platform, and constructing a bus duct monitoring virtual model in combination with a bus duct arrangement scheme and a building scene; constructing an initial monitoring analysis framework based on a sensor type, extracting sensor data from the virtual model, and obtaining a calibration monitoring analysis framework sequence array in combination with the analysis framework; the monitoring analysis framework sequence array is used for bus duct fault analysis, a fault detection result is obtained and visually displayed, and the fault detection result comprises a fault calibration monitoring point set and a fault type set. Therefore, the technical effects of dynamically sensing and deducing the position of a potential fault part and improving the accuracy of fault detection are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of electrical fault detection, and in particular to a bus duct fault detection method and system based on virtual reality modeling. Background Art

[0002] As an important power transmission equipment, busbar fault detection is of key significance for ensuring the stable operation of the power system. Traditional busbar fault detection methods mainly rely on single sensor data, such as temperature monitoring, current monitoring or partial discharge monitoring, which has certain limitations in practical applications. On the one hand, various types of monitoring data (such as temperature, current, partial discharge, etc.) are often relatively scattered, making it difficult to perform effective data fusion and correlation judgment, resulting in limited accuracy and timeliness of fault detection. On the other hand, traditional methods cannot achieve dynamic perception and location deduction of potential fault locations, making it difficult to meet the needs of busbar fault detection in complex environments. Summary of the Invention

[0003] The present invention provides a bus duct fault detection method and system based on virtual reality modeling to solve the technical problems of monitoring data dispersion, data fusion and difficulty in association judgment in the prior art, realize dynamic perception and location deduction of potential fault sites, and improve the accuracy of fault detection.

[0004] In a first aspect, the present invention provides a bus duct fault detection method based on virtual reality modeling, wherein the bus duct fault detection method based on virtual reality modeling includes:

[0005] The monitoring points are calibrated based on the target busbar layout scheme and the target application building scenario to obtain a calibrated monitoring point array.

[0006] The multi-source sensor group is deployed by traversing the calibration monitoring point array, and the deployed multi-source sensor group array is communicatively connected to the virtual reality modeling platform. The target bus duct arrangement scheme and the target application building scene are combined for modeling to construct a bus duct monitoring virtual model.

[0007] An initial monitoring and analysis framework is constructed based on the sensor types in the multi-source sensor group, and sensor data of the multi-source sensor group array for a preset monitoring window is extracted from the bus duct monitoring virtual model. Combined with the initial monitoring and analysis framework, a calibrated monitoring and analysis framework sequence array is obtained.

[0008] Based on the calibration monitoring and analysis framework sequence array, bus duct fault analysis is performed to obtain fault detection results, which are visualized in the bus duct monitoring virtual model, wherein the fault detection results include a fault calibration monitoring point set and a fault type set.

[0009] In a feasible implementation, performing bus duct fault analysis based on the calibration monitoring and analysis framework sequence array to obtain a fault detection result includes:

[0010] Traverse the calibration monitoring analysis framework sequence array to perform feature perception analysis to obtain a calibration monitoring feature array.

[0011] A joint fault analysis is performed based on the calibrated monitoring feature array to obtain the fault detection result.

[0012] In a feasible implementation, traversing the calibration monitoring analysis framework sequence array to perform feature perception analysis to obtain a calibration monitoring feature array includes:

[0013] Based on the feature perception branch set, multi-scale convolution analysis is performed on the calibration monitoring analysis frame sequence array to obtain a convolution monitoring feature set array, wherein the receptive fields of any two feature perception branches in the feature perception branch set are different.

[0014] Traversing any one of the convolution monitoring feature set arrays to perform iterative feature fusion to obtain the calibration monitoring feature array.

[0015] In a feasible implementation, traversing any one of the convolution monitoring feature set arrays to perform iterative feature fusion to obtain the calibration monitoring feature array includes:

[0016] A first convolution monitoring feature and a second convolution monitoring feature are randomly extracted without replacement from any convolution monitoring feature set in the convolution monitoring feature set array.

[0017] Iterative feature fusion is performed on the first convolution monitoring feature and the second convolution monitoring feature to obtain a first iterative convolution monitoring feature.

[0018] Then, a third convolution monitoring feature is randomly extracted from the corresponding convolution monitoring feature set without replacement, and the third convolution monitoring feature is iteratively fused with the first iterative convolution monitoring feature to obtain a second iterative convolution monitoring feature. After multiple iterations, until all the convolution monitoring features in the corresponding convolution monitoring feature set are iterated, the corresponding calibrated monitoring feature array is obtained.

[0019] In a feasible implementation, performing iterative feature fusion on the first convolution monitoring feature and the second convolution monitoring feature to obtain a first iterative convolution monitoring feature includes:

[0020] An inner product mapping analysis is performed on the first convolution monitoring feature and the second convolution monitoring feature, and the analysis results are normalized to obtain an iterative feature fusion matrix.

[0021] The convolutional network is used to perform a convolution operation on the iterative feature fusion matrix and the second convolution monitoring feature to obtain the first iterative convolution monitoring feature.

[0022] In one feasible implementation, monitoring point calibration is performed based on the target busbar layout scheme and the target application building scenario to obtain a calibrated monitoring point array, including:

[0023] The scenario characteristics of the target application building scenario are obtained, and the bus duct fault log is retrieved in combination with the target bus duct arrangement scheme.

[0024] The fault type and the fault point location are used as indexes respectively to extract data from the retrieved bus duct fault log to determine a set of fault point locations and a set of corresponding fault types.

[0025] The fault type set is aggregated in the same category, and the fault point location set is mapped and aggregated according to the aggregation result to obtain Q aggregated fault point location sets, where Q is a positive integer.

[0026] The arrangement lines in the target bus duct arrangement scheme are calibrated in combination with the Q aggregated fault point position sets to obtain the calibrated monitoring point array.

[0027] In a feasible implementation, the arrangement lines in the target busbar arrangement scheme are calibrated in combination with the Q aggregated fault point location sets to obtain the calibrated monitoring point array, including:

[0028] The aggregated fault points with the highest occurrence frequency are respectively selected from the Q aggregated fault point location sets as aggregation centers to obtain Q aggregation center sets.

[0029] Aggregation center neighborhoods of the Q aggregation center sets are constructed in the Q aggregated fault point location sets to obtain Q aggregation center neighborhood sets.

[0030] Aggregate fault location points other than the Q aggregation center neighborhood sets in the Q aggregate fault point location sets are randomly sampled according to a preset first sampling amount to obtain Q extracted monitoring point sets.

[0031] Random sampling of aggregated fault location points is performed in the Q aggregation center neighborhood sets according to a preset second sampling amount to obtain Q neighborhood monitoring point sets.

[0032] A union is performed on the Q extracted monitoring point sets and the Q neighborhood monitoring point sets to obtain the calibration monitoring point array.

[0033] In a feasible implementation, constructing the aggregation center neighborhoods of the Q aggregation center sets in the Q aggregated fault point location sets to obtain the Q aggregation center neighborhood sets includes:

[0034] In the Q aggregated fault point location sets, Q initial aggregation center neighborhood sets of the Q aggregation center sets are constructed according to a preset diffusion bandwidth.

[0035] The Q initial clustering center neighborhood sets are respectively subjected to neighborhood diffusion according to a preset diffusion bandwidth until the neighborhood density of two adjacent diffusions is less than or equal to a preset neighborhood density threshold, and then the diffusion is stopped to obtain the Q clustering center neighborhood sets.

[0036] In a feasible implementation, the multi-source sensor group includes a temperature sensor, a current sensor, a voltage sensor, and a partial discharge signal sensor.

[0037] In a second aspect, the present invention further provides a bus duct fault detection system based on virtual reality modeling, wherein the bus duct fault detection system based on virtual reality modeling includes:

[0038] The monitoring point calibration module is used to calibrate the monitoring points based on the target busbar layout plan and the target application building scenario to obtain a calibrated monitoring point array.

[0039] The sensor layout and modeling module is used to traverse the calibration monitoring point array to deploy a multi-source sensor group, communicate with the deployed multi-source sensor group array and the virtual reality modeling platform, and model the target bus duct layout scheme and the target application building scenario to construct a bus duct monitoring virtual model.

[0040] An analysis framework construction module is used to construct an initial monitoring and analysis framework based on the sensor types in the multi-source sensor group, extract sensor data of the multi-source sensor group array for a preset monitoring window from the bus duct monitoring virtual model, and combine the initial monitoring and analysis framework to obtain a calibrated monitoring and analysis framework sequence array.

[0041] A fault analysis and visualization module is used to perform bus duct fault analysis based on the calibration monitoring and analysis framework sequence array, obtain fault detection results, and visualize them in the bus duct monitoring virtual model, wherein the fault detection results include a fault calibration monitoring point set and a fault type set.

[0042] The present invention discloses a bus duct fault detection method and system based on virtual reality modeling, comprising: calibrating monitoring points according to a target bus duct arrangement scheme and a target application building scene to obtain a calibrated monitoring point array; traversing the calibrated monitoring point array and arranging a multi-source sensor group, and after completion, communicating and connecting the sensor group array with a virtual reality modeling platform, and modeling in combination with the bus duct arrangement scheme and the building scene to construct a bus duct monitoring virtual model; constructing an initial monitoring and analysis framework based on sensor type, extracting sensor data from the virtual model and combining it with the analysis framework to obtain a calibrated monitoring and analysis framework sequence array; using the monitoring and analysis framework sequence array to perform bus duct fault analysis, obtain fault detection results, and visualize them in the virtual model, wherein the fault detection results include a fault calibrated monitoring point set and a fault type set. The bus duct fault detection method and system based on virtual reality modeling disclosed by the present invention solve the technical problems of monitoring data dispersion, data fusion and difficulty in association judgment, realize dynamic perception and position deduction of potential fault locations, and improve the accuracy of fault detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 Schematic diagram of the process of bus duct fault detection method based on virtual reality modeling of the present invention;

[0044] Figure 2 It is a structural schematic diagram of the bus duct fault detection system based on virtual reality modeling of the present invention.

[0045] Explanation of the reference numerals: monitoring point calibration module 11 , sensor layout and modeling module 12 , analysis framework construction module 13 , fault analysis and visualization module 14 . DETAILED DESCRIPTION

[0046] The above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods of the specification to better understand the above technical solution. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments used only to explain the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. In addition, it should be noted that, for the convenience of description, only the parts related to the present invention, rather than all, are shown in the drawings.

[0047] Example 1, as Figure 1 The figure is a flow chart of a bus duct fault detection method based on virtual reality modeling according to the present invention, wherein the bus duct fault detection method based on virtual reality modeling includes:

[0048] S100: Calibrate monitoring points based on the target busbar arrangement scheme and the target application building scenario to obtain a calibrated monitoring point array.

[0049] Specifically, the target bus duct arrangement plan refers to the arrangement and layout planning of the bus duct in a specific application scenario, which includes information such as the direction and branch location of the bus duct; the target application building scenario refers to the building environment where the bus duct is actually installed and used, such as factories and commercial buildings. Different building scenarios correspond to specific building structures and usage requirements, which in turn affect the subsequent selection of monitoring points and the construction of the monitoring point array.

[0050] Specifically, based on the target busbar layout and the target building scenario, the location and number of monitoring points within the target building scenario can be reasonably determined to ensure the comprehensiveness and accuracy of the monitoring data. The calibration monitoring point array is a collection of monitoring points obtained through the monitoring point calibration process. These monitoring points will serve as the basis for the subsequent multi-source sensor group deployment.

[0051] The above steps calibrate the monitoring points by comprehensively considering the building scenario and busbar layout plan, which can ensure the targetedness and effectiveness of subsequent monitoring work. Through the reasonable layout of monitoring points, the key areas of the busbar are fully covered, providing guarantees for accurately obtaining operating data and timely discovering potential faults.

[0052] In some embodiments, monitoring point calibration is performed based on the target bus duct arrangement scheme and the target application building scenario to obtain a calibrated monitoring point array, including:

[0053] Obtain scenario features of the target application building scenario, and retrieve bus duct fault logs in combination with the target bus duct arrangement scheme; extract data from the retrieved bus duct fault logs using fault type and fault point location as indexes, respectively, to determine a set of fault point locations and a corresponding set of fault types; perform similar aggregation on the fault type sets, and map and aggregate the fault point location sets based on the aggregation results to obtain Q aggregated fault point location sets, where Q is a positive integer; calibrate the arrangement lines in the target bus duct arrangement scheme in combination with the Q aggregated fault point location sets to obtain the calibrated monitoring point array.

[0054] Specifically, first, the scene characteristics of the target application building scene are obtained: for example, if the target application building scene is a factory workshop, its scene characteristics may include the length, width, and height of the workshop, the installation position of the bus duct (such as laying along the wall, passing through the floor, etc.), and the distribution of other equipment in the workshop, etc. Then, combined with the direction of the bus duct in the workshop specified by the target bus duct layout plan (such as starting from the distribution room, extending north along the A axis to area B, and then turning west to near the C device), previous bus duct fault logs are retrieved in the building scene of the workshop to find fault records that have occurred on this layout path.

[0055] Specifically, data is extracted from the bus duct fault log using the fault type and fault point location as indexes. For example, assuming that there are records in the retrieved fault log showing that an insulation fault caused by water leakage occurred at the location where the bus duct passed through the floor, and that the bus duct near the C device had an abnormal temperature increase due to overload, then the fault type can be used as an index to extract a set of fault types such as "insulation fault" and "temperature abnormality"; and the fault point location can be used as an index to extract a set of fault point locations such as "through the floor" and "near the C device".

[0056] Specifically, the fault type set is aggregated into similar categories, and based on the aggregation results, the fault location set is mapped and aggregated. First, the extracted fault types are classified into different categories, such as "insulation fault," "temperature anomaly," and "partial discharge." The corresponding fault records are aggregated and the fault location set is mapped. This means that multiple fault location locations corresponding to the same fault type are grouped into an aggregated fault location set. For example, if a "temperature anomaly" fault occurs at multiple bends in the bus duct, these bends are grouped into an aggregated fault location set. This process ultimately yields Q aggregated fault location sets, where Q is a positive integer.

[0057] Furthermore, the Q aggregated fault point location sets obtained are associated with the routing lines in the target busbar trunking layout scheme, and the routing lines are calibrated in the target busbar trunking layout scheme. For example, monitoring points are set at certain intervals (e.g., one monitoring point every 5 meters) at locations such as floor crossings and corners corresponding to the aggregated fault point location set in the "insulation fault prone area"; the density of monitoring points is increased (e.g., one monitoring point every 3 meters) at locations such as busbar trunking bends and near high-power equipment corresponding to the aggregated fault point location set in the "temperature anomaly high area"; and monitoring points specifically for detecting partial discharges are selected at appropriate locations (e.g., at busbar trunking joints and branches) corresponding to the aggregated fault point location set in the "partial discharge concentrated area" to ultimately obtain a calibrated monitoring point array covering the entire busbar trunking layout.

[0058] Through the above steps, first, it is possible to ensure that the layout of monitoring points fully considers the historical fault characteristics and the actual needs of the building scenario, so that the monitoring points are reasonably distributed in key parts and fault-prone areas of the bus duct, thereby improving the comprehensiveness and accuracy of the monitoring data. Secondly, this monitoring point calibration method based on historical fault data and scenario characteristics can enhance the targeted nature of fault detection, allowing subsequent fault detection to focus more on fault-prone areas, improving the efficiency and accuracy of fault detection. At the same time, it also has strong adaptability and can be flexibly adjusted according to different target application building scenarios and bus duct layout plans to meet the needs of bus duct fault detection in various complex environments.

[0059] In some implementations, calibrating the busbar arrangement scheme in combination with the Q aggregated fault point location sets to obtain the calibrated monitoring point array includes:

[0060] The aggregated fault points with the highest occurrence frequency are respectively selected from the Q aggregated fault point location sets as aggregation centers to obtain Q aggregation center sets; the aggregation center neighborhoods of the Q aggregation center sets are constructed in the Q aggregated fault point location sets to obtain Q aggregation center neighborhood sets; the aggregated fault location points other than the Q aggregation center neighborhood sets in the Q aggregated fault point location sets are randomly sampled according to a preset first sampling amount to obtain Q extracted monitoring point sets; the aggregated fault location points are randomly sampled according to a preset second sampling amount in the Q aggregation center neighborhood sets to obtain Q neighborhood monitoring point sets; the Q extracted monitoring point sets and the Q neighborhood monitoring point sets are unioned to obtain the calibrated monitoring point array.

[0061] Specifically, first, the aggregated fault points with the highest frequency are extracted from each aggregated fault point location set, and the output is Q aggregation center sets. These Q aggregation center sets represent the most typical and representative fault points in the Q aggregated fault point location sets; then, for each aggregation center, its neighborhood is constructed in the corresponding aggregated fault point location set, and Q aggregation center neighborhood sets are obtained. Each set contains other fault location points within a certain range around the aggregation center. These points together reflect the fault characteristic distribution of the area. The construction of the neighborhood is realized based on preset neighborhood range parameters (for example: a certain spatial radius or topological distance).

[0062] Specifically, in each set of Q aggregated fault point locations, the remaining aggregated fault location points outside the constructed Q aggregation center neighborhood sets are excluded, and the remaining fault location points are randomly sampled according to preset parameters to obtain Q extracted monitoring point sets. These monitoring points are used to supplement the data covering areas not captured by the neighborhood to ensure the balance of the overall monitoring point distribution.

[0063] For example, in the "insulation fault prone area" set, after removing the neighborhood 2 meters before and after the floor crossing location, 5 points are randomly selected from other insulation fault point locations; in the "temperature abnormality high incidence area" set, after removing the neighborhood 1.5 meters before and after the high-power motor turning point, 5 points are randomly selected from other temperature abnormality fault point locations; in the "partial discharge concentration area" set, after removing the neighborhood 0.5 meters before and after the joint, 5 points are randomly selected from other partial discharge fault point locations. This results in Q sets of extracted monitoring points. Three points are randomly selected from the cluster center neighborhood of the "insulation fault prone area" (the interval 2 meters before and after the floor crossing location); three points are randomly selected from the cluster center neighborhood of the "temperature abnormality high incidence area" (the area 1.5 meters before and after the turning point); and three points are randomly selected from the cluster center neighborhood of the "partial discharge concentration area" (the range 0.5 meters before and after the joint). Thus, Q sets of neighborhood monitoring points are obtained.

[0064] Furthermore, the Q extracted monitoring point sets (each containing five monitoring points) and the Q neighborhood monitoring point sets (each containing three monitoring points) obtained above are merged to form a complete set of monitoring points. This set of monitoring points includes both randomly selected points from non-neighborhood areas and randomly selected points from neighborhood areas, forming a calibration monitoring point array that covers the critical fault-prone areas and adjacent areas of the busbar layout, providing a scientific and reasonable layout basis for the subsequent deployment of multi-source sensor groups.

[0065] Through the above steps, the reasonable distribution of monitoring points on the entire bus duct wiring is ensured, laying a solid foundation for the subsequent accurate acquisition of bus duct operation data and timely detection of potential faults.

[0066] In some implementations, constructing the aggregation center neighborhoods of the Q aggregation center sets in the Q aggregated fault point location sets to obtain the Q aggregation center neighborhood sets includes:

[0067] Construct Q initial aggregation center neighborhood sets of the Q aggregation center sets according to a preset diffusion bandwidth in the Q aggregated fault point location sets; perform neighborhood diffusion on the Q initial aggregation center neighborhood sets respectively according to the preset diffusion bandwidth until the neighborhood density of two adjacent diffusions is less than or equal to a preset neighborhood density threshold, stop diffusion, and obtain the Q aggregation center neighborhood sets.

[0068] Specifically, the preset diffusion bandwidth is a pre-set range parameter used to construct the initial cluster center neighborhood and perform neighborhood diffusion. It determines the size of the neighborhood and the diffusion step size. The neighborhood density is a metric used to measure the density of fault points within the neighborhood during the neighborhood diffusion process. It can be expressed as the number of fault points per unit area or unit length. The preset neighborhood density threshold is the critical value used to determine whether neighborhood diffusion should cease. When the neighborhood density is less than or equal to this value, the neighborhood diffusion is considered to have reached the appropriate range.

[0069] Specifically, assuming Q = 3, which corresponds to the three aggregate fault point location sets of "insulation fault prone area", "temperature anomaly high area" and "partial discharge concentrated area", and their aggregation centers are the location where the bus duct crosses the floor, the turning point near the high-power motor, and the joint of the bus duct. The default diffusion bandwidth is 1 meter. For the aggregation center of the "insulation fault prone area" - the location where the floor is crossed, the initial aggregation center neighborhood is constructed according to the 1-meter diffusion bandwidth, that is, the bus duct area 1 meter before and after the location where the floor is crossed as the center; for the aggregation center of the "temperature anomaly high area" - the turning point near the high-power motor, the initial aggregation center neighborhood is constructed according to the 1-meter diffusion bandwidth, that is, the area 1 meter before and after the turning point; for the aggregation center of the "partial discharge concentrated area" - the joint of the bus duct, the initial aggregation center neighborhood is constructed according to the 1-meter diffusion bandwidth, that is, the range of 0.5 meters before and after the joint (assuming that the diffusion bandwidth here is adjusted to 0.5 meters according to the actual situation). In this way, Q initial aggregation center neighborhood sets are obtained.

[0070] Specifically, neighborhood diffusion is performed on each of the Q initial cluster center neighborhood sets according to a preset diffusion bandwidth until the neighborhood density of two consecutive diffusions is less than or equal to a preset neighborhood density threshold. Diffusion is then stopped, resulting in Q cluster center neighborhood sets. For example, taking the initial cluster center neighborhood of the "insulation fault prone area" as an example, assume the preset neighborhood density threshold is 0.3 fault points per meter. First, the neighborhood density within this initial neighborhood is calculated, i.e., the number of insulation fault points contained in this interval is counted. Assuming there are 5, and the interval length is 2 meters, the neighborhood density is 5 / 2 = 2.5 (2.5 fault points per meter). Since 2.5 is greater than the preset neighborhood density threshold of 0.3, neighborhood diffusion is required. Using the preset diffusion bandwidth of 1 meter, the neighborhood is expanded 1 meter to each end. At this point, the neighborhood becomes an interval of 2 meters in front and 2 meters in back, with a length of 4 meters. The number of insulation fault points in this new neighborhood is again counted. Assuming there are 8, the neighborhood density is 8 / 4 = 2 (2 fault points per meter), which is still greater than 0.3, so diffusion continues. Expanding 1 meter in each direction, the neighborhood becomes 3 meters in front and 3 meters in length, with a total length of 6 meters. The number of fault points counted is 10, and the neighborhood density is 10 / 6 ≈ 1.67 (1.67 fault points per meter), which is still greater than 0.3. Continue expanding. Suppose another expansion is performed, and the neighborhood becomes 4 meters in front and 4 meters in length, with a total length of 8 meters. The number of fault points counted is 12, and the neighborhood density is 12 / 8 = 1.5 (1.5 fault points per meter), which is still greater than 0.3. Continue expanding. After a certain expansion, the neighborhood density is less than or equal to 0.3, and then the expansion stops. The resulting neighborhood is the cluster center neighborhood of the "insulation fault prone area." Similarly, expand the initial cluster center neighborhoods of the "temperature anomaly high incidence area" and "partial discharge concentrated area" until the neighborhood density conditions are met, ultimately obtaining a set of Q cluster center neighborhoods.

[0071] Through the above steps, the initial aggregation center neighborhood set is constructed according to the preset diffusion bandwidth. With the aggregation center as the core, a relatively concentrated fault-prone area can be preliminarily determined, providing a basic range for subsequent neighborhood diffusion. Secondly, during the neighborhood diffusion process, by continuously expanding the neighborhood range and calculating the neighborhood density until the neighborhood density is less than or equal to the preset neighborhood density threshold, the size of the neighborhood can be dynamically adjusted to ensure that the final aggregation center neighborhood set not only covers the core area with high fault incidence, but also appropriately expands to the surrounding areas where potential faults may exist, avoiding the problem of the neighborhood range being too small or too large, making the construction of the aggregation center neighborhood more scientific and reasonable, and better able to adapt to the distribution of busbar faults under different fault types and building scenarios.

[0072] S200: traverse the calibration monitoring point array to deploy a multi-source sensor group, communicate and connect the deployed multi-source sensor group array with the virtual reality modeling platform, combine the target bus duct arrangement scheme and the target application building scene for modeling, and construct a bus duct monitoring virtual model.

[0073] Specifically, first, the information of each monitoring point (such as coordinates, area, preset monitoring type, etc.) is read in sequence. Based on the monitoring requirements (temperature, vibration, current, humidity, etc.), the type and number of sensors required for each monitoring point are determined. Sensors are installed at each calibration monitoring point according to the preset plan to form a multi-source sensor group. Then, a wired network (Ethernet, RS485, etc.) or wireless communication (Wi-Fi, LoRa, etc.) is used to achieve real-time data transmission between the sensor group array and the virtual reality modeling platform. The virtual reality modeling platform is a platform that uses virtual reality technology to build models (such as BIM). It can digitally model the bus duct, its monitoring points, sensors, and other information to achieve the creation and visualization of virtual scenes. Next, the real-time sensor data is integrated with the target bus duct layout plan and building scene data. For example, a three-dimensional virtual model of the bus duct is constructed using modeling tools such as CAD, BIM, and GIS. The sensor monitoring data is superimposed on the model to establish a virtual model of the bus duct monitoring for dynamic monitoring.

[0074] Through the above process, the constructed bus duct monitoring virtual model can display the specific location, layout and real-time monitoring data (such as temperature, vibration, fault alarm, etc.) of the target bus duct in the building scene.

[0075] In some embodiments, the multi-source sensor group includes a temperature sensor, a current sensor, a voltage sensor, and a partial discharge signal sensor.

[0076] Specifically, the multi-source sensor group is a combination of various types of sensors, such as temperature sensors, current sensors, voltage sensors, and partial discharge signal sensors, which can obtain bus duct operation data from different angles.

[0077] S300: Constructing an initial monitoring and analysis framework based on the sensor types in the multi-source sensor group, extracting sensor data of the multi-source sensor group array for a preset monitoring window from the bus duct monitoring virtual model, and combining the initial monitoring and analysis framework to obtain a calibrated monitoring and analysis framework sequence array.

[0078] Specifically, the multi-source sensor group includes various sensor types, such as temperature sensors, vibration sensors, humidity sensors, pressure sensors, etc. According to the functions and data characteristics of different sensors, each sensor can be classified to form a sensor type catalog.

[0079] Specifically, the initial monitoring and analysis framework is a preliminary framework used to guide monitoring data analysis, defining how to process and correlate data from different sensors. For example, temperature sensor data undergoes smoothing filtering and trend analysis; vibration sensor data undergoes spectrum analysis and anomaly detection; and humidity and pressure data undergo normalization and statistical analysis. The calibration monitoring and analysis framework sequence array is a set of monitoring and analysis framework sequences determined for each calibration monitoring point, combining the initial monitoring and analysis framework with sensor data within a preset monitoring window.

[0080] Specifically, in the bus duct monitoring virtual model, sensor data from the multi-source sensor array at each calibrated monitoring point is extracted according to a preset monitoring window (e.g., set to 1 hour). These data are stored in the virtual model in the form of a time series to facilitate subsequent analysis and processing. The extracted sensor data is then analyzed according to the processing methods and association rules defined in the initial monitoring and analysis framework. For example, trend analysis is performed on temperature data to determine whether there is an abnormally rising trend; frequency domain analysis is performed on current data to detect whether there are abnormal harmonic components; and threshold judgment is performed on partial discharge signal data to identify whether there is a discharge signal that exceeds the set intensity. Through these analyses, a monitoring and analysis framework sequence is generated for each calibrated monitoring point. The array records the real-time monitoring characteristics of each monitoring point in the entire monitoring network, providing data support for fault prediction, status assessment, and scheduling optimization.

[0081] S400: Perform bus duct fault analysis based on the calibration monitoring and analysis framework sequence array to obtain fault detection results, and visualize them in the bus duct monitoring virtual model, wherein the fault detection results include a fault calibration monitoring point set and a fault type set.

[0082] Specifically, the monitoring feature vectors in the calibrated monitoring feature array are first comprehensively analyzed, taking into account the correlation between different monitoring points and the similarity of fault characteristics to determine the presence of a fault, as well as the type and location of the fault. For example, if the temperature feature vectors of multiple adjacent monitoring points all show an abnormally high trend, accompanied by increased current fluctuations, combined with an increase in the frequency of partial discharge signals, it can be determined that a busbar overload fault may exist near these monitoring points. This allows the set of fault calibration monitoring points and fault types to be determined. The obtained fault detection results are then applied to a virtual busbar monitoring model, displaying the fault information in an intuitive and visual manner. For example, in the virtual model, the fault calibration monitoring points are highlighted with a red marker, and the corresponding fault type is annotated next to the marker. Different fault types can be distinguished by different colors or icons, such as yellow warning symbols for insulation faults and blue warning symbols for temperature anomalies. This allows maintenance personnel to quickly and accurately understand the fault distribution of the busbar and take appropriate maintenance measures in a timely manner.

[0083] In some embodiments, performing bus duct fault analysis based on the calibration monitoring and analysis framework sequence array to obtain a fault detection result includes:

[0084] Traversing the calibration monitoring analysis framework sequence array to perform feature perception analysis to obtain a calibration monitoring feature array; performing fault joint analysis based on the calibration monitoring feature array to obtain the fault detection result.

[0085] Specifically, the system first traverses the array of monitoring analysis framework sequences generated by each calibrated monitoring point. For each sequence, key feature values are extracted, such as multi-dimensional characteristic data such as temperature, vibration, current, and humidity, along with their statistical parameters (mean, standard deviation, peak value, etc.), to form a calibrated monitoring feature vector. Furthermore, the calibrated monitoring feature vectors of all monitoring points are arranged according to a specific rule (such as by geographic location or functional area) to form a calibrated monitoring feature array. This array comprehensively reflects the operating status and abnormal conditions of each key area of the busbar trunking.

[0086] Specifically, through feature pattern matching, the monitoring feature array is compared with the pre-built fault diagnosis rule library or historical fault feature template to identify feature combinations related to specific fault types. For example, abnormal temperature increase, increased vibration, and increased current fluctuation may indicate poor contact or local overheating of the busbar. Then, combined with the spatial correlation between the monitoring points, the trend of the fault spreading from one node to other nodes is analyzed to determine the fault cascade effect. For example, if the temperature change rate and current fluctuation amplitude of multiple monitoring points exceed the normal range, and the local discharge signal intensity increases at the same time, and the feature vectors of multiple monitoring points have the highest match with the "insulation fault" pattern, these monitoring points will be included in the fault calibration monitoring point set, and the fault type will be determined as an insulation fault. Finally, the fault calibration monitoring point set and the fault type set are integrated into the fault detection results for visualization and further processing in subsequent steps.

[0087] In some implementations, traversing the calibration monitoring analysis frame sequence array to perform feature perception analysis to obtain a calibration monitoring feature array includes:

[0088] Based on the feature perception branch set, multi-scale convolution analysis is performed on the calibration monitoring analysis framework sequence array to obtain a convolution monitoring feature set array, wherein the receptive fields of any two feature perception branches in the feature perception branch set are different; any convolution monitoring feature set in the convolution monitoring feature set array is traversed to perform iterative feature fusion to obtain the calibration monitoring feature array.

[0089] Specifically, the feature perception branch set consists of multiple feature perception branches with different receptive fields. Each feature perception branch can perceive the features of monitoring data from different scales and ranges for multi-scale convolution analysis. The convolution monitoring feature set array is a feature set obtained by performing multi-scale convolution analysis on the calibration monitoring analysis framework sequence array based on the feature perception branch set. Each convolution monitoring feature set corresponds to the multi-scale feature representation of a calibration monitoring point.

[0090] Specifically, for a monitoring and analysis framework sequence at a calibrated monitoring point, using a feature perception branch with a smaller receptive field can capture local, subtle feature changes, such as short-term temperature fluctuations; while using a feature perception branch with a larger receptive field can perceive feature trends over a larger range, such as long-term current trends. By performing convolution operations on the monitoring and analysis framework sequence through feature perception branches with different receptive fields, features at different scales can be extracted. For example, for a temperature data sequence, a convolution kernel with a small receptive field (such as 3×3) can extract short-term temperature change features, such as hourly temperature fluctuations; a convolution kernel with a large receptive field (such as 7×7 or larger) can extract long-term temperature change trends, such as the temperature change curve within a day.

[0091] Furthermore, the features extracted from each calibrated monitoring point under different receptive fields are combined to form a convolutional monitoring feature set of the monitoring point, and the convolutional monitoring feature sets of all calibrated monitoring points are integrated to form a convolutional monitoring feature set array, providing basic data for subsequent feature fusion.

[0092] Specifically, any feature set in the multi-scale convolution monitoring feature set array is traversed as the initial fusion basis, and the convolution features at different scales are fused into a unified calibration monitoring feature array, so that the final output features contain both local detail information and global trend expression. Among them, the fusion adopts a layer-by-layer fusion strategy, such as fusing the features of each branch through layer-by-layer splicing, weighted summation, or self-attention mechanism, and iteratively optimizing the fused features. The feature representation is updated after each round of fusion until the fusion result converges. After multi-scale convolution and iterative feature fusion, a unified calibration monitoring feature array is finally obtained, which serves as the input for subsequent fault prediction, health assessment, or scheduling optimization modules.

[0093] Through the above steps, after multi-scale convolution and iterative feature fusion, a unified calibration monitoring feature array is finally obtained, which serves as the input of subsequent fault prediction, health assessment or scheduling optimization modules.

[0094] In some implementations, traversing any one of the convolutional monitoring feature set arrays to perform iterative feature fusion to obtain the calibrated monitoring feature array includes:

[0095] A first convolution monitoring feature and a second convolution monitoring feature are randomly extracted from any one of the convolution monitoring feature sets in the convolution monitoring feature set array without replacement; the first convolution monitoring feature and the second convolution monitoring feature are iteratively fused to obtain a first iterative convolution monitoring feature; a third convolution monitoring feature is then randomly extracted from the corresponding convolution monitoring feature set without replacement, and the third convolution monitoring feature is iteratively fused with the first iterative convolution monitoring feature to obtain a second iterative convolution monitoring feature. After multiple iterations, until all the convolution monitoring features in the corresponding convolution monitoring feature set are iterated, the corresponding calibration monitoring feature array is obtained.

[0096] Specifically, the process of traversing the convolution monitoring feature set array to perform iterative feature fusion and obtain the calibration monitoring feature array is as follows:

[0097] First, a first convolutional monitoring feature and a second convolutional monitoring feature are randomly extracted without replacement from any convolutional monitoring feature set in the convolutional monitoring feature set array. Once extracted, these two features will not be replaced in subsequent extraction processes. Then, the first and second convolutional monitoring features are iteratively fused to obtain a first iterative convolutional monitoring feature. For example, if the short-term temperature fluctuation feature is more critical in fault detection, it is given a weight of 0.6; the long-term temperature change trend feature is given a weight of 0.4. The weighted average is calculated to obtain a first iterative convolutional monitoring feature, which integrates information about short-term and long-term temperature changes. Next, a third convolutional monitoring feature is randomly extracted without replacement from the corresponding convolutional monitoring feature set. The third convolutional monitoring feature is iteratively fused with the first iterative convolutional monitoring feature to obtain a second iterative convolutional monitoring feature. This information is further integrated and repeated multiple times until all convolutional monitoring features in the corresponding convolutional monitoring feature set have been iterated, resulting in the corresponding calibration monitoring feature array.

[0098] Through the above steps, multiple iterations are performed until all features are integrated. The obtained calibrated monitoring feature array can fully reflect the comprehensive operating status of the bus duct at each monitoring point, providing high-quality data support for subsequent fault analysis, improving the accuracy and reliability of fault detection, and overcoming the problems of scattered monitoring data, data fusion and association judgment difficulties in traditional methods.

[0099] In some implementations, performing iterative feature fusion on the first convolution monitoring feature and the second convolution monitoring feature to obtain a first iterative convolution monitoring feature includes:

[0100] An inner product mapping analysis is performed on the first convolution monitoring feature and the second convolution monitoring feature, and the analysis results are normalized to obtain an iterative feature fusion matrix; a convolution network is used to perform a convolution operation on the iterative feature fusion matrix and the second convolution monitoring feature to obtain a first iterative convolution monitoring feature.

[0101] Specifically, assuming that the first convolution monitoring feature is a vector of length n and the second convolution monitoring feature is a vector of length m, their inner product is calculated to obtain an n×m matrix M, which reflects the similarity between each element in the two feature vectors; then, use a method such as the minimum-maximum normalization method to normalize the value of each element in the matrix M to the range of [0, 1] to obtain the normalized iterative feature fusion matrix.

[0102] Specifically, a convolutional network is designed with an input layer that matches the size of the iterative feature fusion matrix. For example, the input layer is n×m×1 (assuming the matrix is two-dimensional, adding a channel dimension). The convolutional layer can use a small-sized convolution kernel (such as 3×3) and an appropriate activation function (such as ReLU) to extract local feature patterns in the matrix. The normalized iterative feature fusion matrix is input into the convolutional network for forward propagation calculation. The convolutional network performs a convolution operation on each local region in the matrix to generate a new feature map. At the same time, the second convolutional monitoring feature is used as an auxiliary input and fused with the output of the convolutional network to generate a more discriminative feature representation, ultimately obtaining the first iterative convolutional monitoring feature.

[0103] Through the above steps, based on inner product mapping and normalization, combined with the feature extraction capability of the convolutional network, the iterative fusion and extraction of multi-source monitoring features are realized, thereby generating a more discriminative and representative first-iteration convolutional monitoring feature.

[0104] In summary, the bus duct fault detection method based on virtual reality modeling provided by the present invention has the following technical effects:

[0105] By calibrating the monitoring points according to the target bus duct arrangement scheme and the target application building scenario, a calibrated monitoring point array is obtained; the calibrated monitoring point array is traversed and a multi-source sensor group is deployed. After completion, the sensor group array is communicated and connected to the virtual reality modeling platform, and modeled in combination with the bus duct arrangement scheme and the building scenario to construct a bus duct monitoring virtual model; an initial monitoring and analysis framework is constructed based on the sensor type, and sensor data is extracted from the virtual model and combined with the analysis framework to obtain a calibrated monitoring and analysis framework sequence array; the bus duct fault analysis is performed using the monitoring and analysis framework sequence array to obtain fault detection results, which are visualized in the virtual model. The fault detection results include a set of fault calibrated monitoring points and a set of fault types, thereby realizing dynamic perception and location deduction of potential fault locations, and improving the accuracy of fault detection.

[0106] Example 2, as Figure 2 This is a schematic diagram of the structure of the bus duct fault detection system based on virtual reality modeling of the present invention. For example, Figure 1 The flow chart of the bus duct fault detection method based on virtual reality modeling in the present invention can be shown as follows: Figure 2 The structure shown is implemented.

[0107] Based on the same concept as the bus duct fault detection method based on virtual reality modeling in the above embodiment, the present invention also provides a bus duct fault detection system based on virtual reality modeling, including:

[0108] The monitoring point calibration module 11 is used to calibrate the monitoring points based on the target bus duct arrangement scheme and the target application building scenario to obtain a calibrated monitoring point array.

[0109] The sensor layout and modeling module 12 is used to traverse the calibration monitoring point array to layout the multi-source sensor group, communicate with the laid-out multi-source sensor group array and the virtual reality modeling platform, combine the target bus duct layout plan and the target application building scene for modeling, and construct a bus duct monitoring virtual model.

[0110] The analysis framework construction module 13 is used to construct an initial monitoring and analysis framework based on the sensor types in the multi-source sensor group, extract the sensor data of the multi-source sensor group array for the preset monitoring window from the bus duct monitoring virtual model, and combine the initial monitoring and analysis framework to obtain a calibrated monitoring and analysis framework sequence array.

[0111] The fault analysis and visualization module 14 is used to perform bus duct fault analysis based on the calibration monitoring and analysis framework sequence array, obtain fault detection results, and visualize them in the bus duct monitoring virtual model, wherein the fault detection results include a fault calibration monitoring point set and a fault type set.

[0112] In some embodiments, the fault analysis and visualization module 14 includes:

[0113] The feature perception analysis unit is used to traverse the calibration monitoring analysis frame sequence array to perform feature perception analysis and obtain a calibration monitoring feature array.

[0114] A fault joint analysis unit is used to perform fault joint analysis based on the calibration monitoring feature array to obtain the fault detection result.

[0115] In some implementations, the feature perception analysis unit in the fault analysis and visualization module 14 includes:

[0116] A multi-scale convolution analysis unit is used to perform multi-scale convolution analysis on the calibration monitoring analysis frame sequence array based on a feature perception branch set to obtain a convolution monitoring feature set array, wherein the receptive fields of any two feature perception branches in the feature perception branch set are different.

[0117] The iterative feature fusion unit is used to traverse any convolution monitoring feature set in the convolution monitoring feature set array to perform iterative feature fusion to obtain the calibration monitoring feature array.

[0118] In some implementations, the iterative feature fusion unit in the feature perception analysis unit includes:

[0119] The convolution monitoring feature extraction unit is used to randomly extract a first convolution monitoring feature and a second convolution monitoring feature from any convolution monitoring feature set in the convolution monitoring feature set array without replacement.

[0120] The iterative feature fusion execution unit is used to perform iterative feature fusion on the first convolution monitoring feature and the second convolution monitoring feature to obtain a first iterative convolution monitoring feature.

[0121] The multiple iterative feature fusion unit is used to randomly extract a third convolution monitoring feature from the corresponding convolution monitoring feature set without replacement, and iteratively fuse the third convolution monitoring feature with the first iterative convolution monitoring feature to obtain a second iterative convolution monitoring feature. After multiple iterations, until all the convolution monitoring features in the corresponding convolution monitoring feature set are iterated, the corresponding calibrated monitoring feature array is obtained.

[0122] In some implementations, the iterative feature fusion execution unit in the iterative feature fusion unit includes:

[0123] An inner product mapping analysis unit is used to perform inner product mapping analysis on the first convolution monitoring feature and the second convolution monitoring feature, and normalize the analysis results to obtain an iterative feature fusion matrix.

[0124] The convolution operation unit is used to use a convolution network to perform a convolution operation on the iterative feature fusion matrix and the second convolution monitoring feature to obtain the first iterative convolution monitoring feature.

[0125] In some embodiments, the monitoring point calibration module 11 includes:

[0126] The scene feature and fault log retrieval unit is used to obtain the scene feature of the target application building scene and retrieve the bus duct fault log in combination with the target bus duct arrangement scheme.

[0127] The fault point location and type extraction unit is used to extract data from the retrieved bus duct fault log using the fault type and the fault point location as indexes, and determine a set of fault point locations and a set of corresponding fault types.

[0128] The fault point location aggregation unit is configured to aggregate the fault type set into the same type, and map and aggregate the fault point location set according to the aggregation result to obtain Q aggregated fault point location sets, where Q is a positive integer.

[0129] The calibration monitoring point array generation unit is used to calibrate the arrangement lines in the target bus duct arrangement scheme in combination with the Q aggregated fault point position sets to obtain the calibration monitoring point array.

[0130] In some implementations, the calibration monitoring point array generation unit in the monitoring point calibration module 11 includes:

[0131] The aggregation center screening unit is used to screen the aggregation fault points with the highest occurrence frequency from the Q aggregation fault point location sets as aggregation centers to obtain Q aggregation center sets.

[0132] The aggregation center neighborhood construction unit is used to construct the aggregation center neighborhood of the Q aggregation center sets in the Q aggregation fault point location sets to obtain Q aggregation center neighborhood sets.

[0133] The sampling monitoring point set acquisition unit is used to randomly sample the aggregated fault location points in the Q aggregated fault point location sets except the Q aggregation center neighborhood sets according to a preset first sampling amount to obtain Q extracted monitoring point sets.

[0134] The neighborhood monitoring point set acquisition unit is used to randomly extract aggregated fault location points in the Q aggregation center neighborhood sets according to a preset second extraction amount to obtain Q neighborhood monitoring point sets.

[0135] The calibration monitoring point array integration unit is used to obtain the calibration monitoring point array by performing a union on the Q extracted monitoring point sets and the Q neighborhood monitoring point sets.

[0136] In some embodiments, the aggregation center neighborhood construction unit includes:

[0137] The initial aggregation center neighborhood construction unit is configured to construct Q initial aggregation center neighborhood sets of the Q aggregation center sets according to a preset diffusion bandwidth in the Q aggregation fault point location sets.

[0138] The neighborhood diffusion unit is used to perform neighborhood diffusion on the Q initial aggregation center neighborhood sets according to a preset diffusion bandwidth until the neighborhood density of two adjacent diffusions is less than or equal to a preset neighborhood density threshold, stop diffusion, and obtain the Q aggregation center neighborhood sets.

[0139] It should be understood that the embodiments mentioned in this specification focus on their differences from other embodiments. The specific embodiments in the aforementioned embodiment one are also applicable to the bus duct fault detection system based on virtual reality modeling described in embodiment two. For the sake of brevity of the specification, no further elaboration will be given here.

[0140] It should be understood that the embodiments disclosed in the present invention and the above description can enable those skilled in the art to use the present invention to implement the present invention. At the same time, the present invention is not limited to the embodiments mentioned above. It should be understood that those skilled in the art can still modify the technical solutions described in the above embodiments or replace some of the technical features therein with equivalents; and such modifications or replacements do not deviate from the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention and are all included in the scope of protection of the present invention.

Claims

1. A bus duct fault detection method based on virtual reality modeling, characterized in that: include: Calibrate monitoring points based on the target busbar layout and target application building scenario to obtain a calibrated monitoring point array; Traversing the calibration monitoring point array to deploy a multi-source sensor group, communicating and connecting the deployed multi-source sensor group array with a virtual reality modeling platform, modeling the target bus duct arrangement scheme and the target application building scenario, and constructing a bus duct monitoring virtual model; Constructing an initial monitoring and analysis framework based on the sensor types in the multi-source sensor group, extracting sensor data of the multi-source sensor group array for a preset monitoring window from the bus duct monitoring virtual model, and combining the initial monitoring and analysis framework to obtain a calibration monitoring and analysis framework sequence array; Based on the calibration monitoring and analysis framework sequence array, bus duct fault analysis is performed to obtain fault detection results, which are visualized in the bus duct monitoring virtual model, wherein the fault detection results include a fault calibration monitoring point set and a fault type set.

2. The bus duct fault detection method based on virtual reality modeling according to claim 1 is characterized in that: Performing bus duct fault analysis based on the calibration monitoring and analysis framework sequence array to obtain fault detection results includes: Traversing the calibration monitoring analysis framework sequence array to perform feature perception analysis to obtain a calibration monitoring feature array; A joint fault analysis is performed based on the calibrated monitoring feature array to obtain the fault detection result.

3. The bus duct fault detection method based on virtual reality modeling according to claim 2 is characterized in that: Traversing the calibration monitoring analysis framework sequence array to perform feature perception analysis to obtain a calibration monitoring feature array, including: Performing multi-scale convolution analysis on the calibration monitoring analysis frame sequence array based on the feature perception branch set to obtain a convolution monitoring feature set array, wherein the receptive fields of any two feature perception branches in the feature perception branch set are different; Traversing any one of the convolution monitoring feature set arrays to perform iterative feature fusion to obtain the calibration monitoring feature array.

4. The bus duct fault detection method based on virtual reality modeling according to claim 3 is characterized in that: Traversing any one of the convolution monitoring feature set arrays to perform iterative feature fusion to obtain the calibration monitoring feature array, including: Randomly extracting a first convolution monitoring feature and a second convolution monitoring feature from any one of the convolution monitoring feature sets in the convolution monitoring feature set array without replacement; Performing iterative feature fusion on the first convolution monitoring feature and the second convolution monitoring feature to obtain a first iterative convolution monitoring feature; Then, a third convolution monitoring feature is randomly extracted from the corresponding convolution monitoring feature set without replacement, and the third convolution monitoring feature is iteratively fused with the first iterative convolution monitoring feature to obtain a second iterative convolution monitoring feature. After multiple iterations, until all the convolution monitoring features in the corresponding convolution monitoring feature set are iterated, the corresponding calibrated monitoring feature array is obtained.

5. The bus duct fault detection method based on virtual reality modeling according to claim 4 is characterized in that: Performing iterative feature fusion on the first convolution monitoring feature and the second convolution monitoring feature to obtain a first iterative convolution monitoring feature, including: Performing inner product mapping analysis on the first convolution monitoring feature and the second convolution monitoring feature, and normalizing the analysis results to obtain an iterative feature fusion matrix; The convolutional network is used to perform a convolution operation on the iterative feature fusion matrix and the second convolution monitoring feature to obtain the first iterative convolution monitoring feature.

6. The bus duct fault detection method based on virtual reality modeling according to claim 1 is characterized in that: Based on the target busbar layout and target application building scenario, the monitoring points are calibrated to obtain a calibrated monitoring point array, including: Obtaining the scenario characteristics of the target application building scenario, and searching the bus duct fault log in combination with the target bus duct arrangement scheme; Using the fault type and the fault point location as indexes respectively, extracting data from the retrieved bus duct fault log, and determining a set of fault point locations and a set of corresponding fault types; Aggregate the fault type set by the same type, and map and aggregate the fault point location set according to the aggregation result to obtain Q aggregated fault point location sets, where Q is a positive integer; The arrangement lines in the target bus duct arrangement scheme are calibrated in combination with the Q aggregated fault point position sets to obtain the calibrated monitoring point array.

7. The bus duct fault detection method based on virtual reality modeling according to claim 6 is characterized in that: Calibrate the arrangement lines in the target busbar arrangement scheme in combination with the Q aggregated fault point location sets to obtain the calibrated monitoring point array, including: Selecting the most frequently occurring aggregated fault points from the Q aggregated fault point location sets as aggregation centers to obtain Q aggregation center sets; Constructing the aggregation center neighborhoods of the Q aggregation center sets in the Q aggregation fault point location sets to obtain Q aggregation center neighborhood sets; Randomly sampling the aggregated fault location points excluding the Q aggregation center neighborhood sets from the Q aggregated fault location sets according to a preset first sampling amount to obtain Q extracted monitoring point sets; Randomly extracting aggregated fault location points in the Q aggregation center neighborhood sets according to a preset second extraction amount to obtain Q neighborhood monitoring point sets; A union is performed on the Q extracted monitoring point sets and the Q neighborhood monitoring point sets to obtain the calibration monitoring point array.

8. The bus duct fault detection method based on virtual reality modeling according to claim 7 is characterized in that: Constructing the aggregation center neighborhoods of the Q aggregation center sets in the Q aggregation fault point location sets to obtain the Q aggregation center neighborhood sets, including: Constructing Q initial aggregation center neighborhood sets of the Q aggregation center sets according to a preset diffusion bandwidth in the Q aggregation fault point location sets; The Q initial clustering center neighborhood sets are respectively subjected to neighborhood diffusion according to a preset diffusion bandwidth until the neighborhood density of two adjacent diffusions is less than or equal to a preset neighborhood density threshold, and then the diffusion is stopped to obtain the Q clustering center neighborhood sets.

9. The bus duct fault detection method based on virtual reality modeling according to claim 1, characterized in that: The multi-source sensor group includes a temperature sensor, a current sensor, a voltage sensor and a partial discharge signal sensor.

10. The bus duct fault detection system based on virtual reality modeling is characterized by: A bus duct fault detection method based on virtual reality modeling for implementing any one of claims 1 to 9, comprising: The monitoring point calibration module is used to calibrate the monitoring points based on the target busbar layout plan and the target application building scenario to obtain a calibrated monitoring point array; A sensor layout and modeling module is used to traverse the calibration monitoring point array to deploy a multi-source sensor group, connect the deployed multi-source sensor group array to the virtual reality modeling platform, and perform modeling based on the target bus duct layout plan and the target application building scenario to construct a bus duct monitoring virtual model; An analysis framework construction module is configured to construct an initial monitoring and analysis framework based on the sensor types in the multi-source sensor group, extract sensor data of the multi-source sensor group array for a preset monitoring window from the bus duct monitoring virtual model, and combine the initial monitoring and analysis framework to obtain a calibrated monitoring and analysis framework sequence array; A fault analysis and visualization module is used to perform bus duct fault analysis based on the calibration monitoring and analysis framework sequence array, obtain fault detection results, and visualize them in the bus duct monitoring virtual model, wherein the fault detection results include a fault calibration monitoring point set and a fault type set.

Images