Low-voltage transformer area risk monitoring method, system and equipment based on multi-source data fusion and medium

Abstract

The invention discloses a low-voltage transformer area risk monitoring method, system and device based on multi-source data fusion and a medium. The method comprises the steps of collecting multi-source data of a low-voltage transformer area; obtaining a rest environment working condition set and a non-rest environment working condition set of each client, and generating fitting trend data; performing user end clustering to obtain a user end rest clustering set and a user end non-rest clustering set; collecting real-time monitoring data of the selected client; performing matching and comparison to obtain a real-time operation label; and generating prediction and early warning data according to the real-time operation label. According to the invention, through distinguishing the rest time period and the non-rest time period, establishing a multi-mode environment working condition set and fitting a trend reference, and performing clustering in combination with the distance between the user terminals, the accuracy and scene adaptability of abnormity monitoring are improved, the false alarm rate and the missing report rate are reduced, and through sampling monitoring and intelligent early warning, the disposal time is bought for risk prevention and control, and the risk is prevented and controlled. And three-dimensional and refined monitoring of the operation risk of the low-voltage transformer area is realized.

Description

Technical Field

[0001] This invention relates to the field of power monitoring technology, and in particular to a method, system, equipment and medium for low-voltage distribution area risk monitoring based on multi-source data fusion. Background Technology

[0002] The stable operation of the power system is crucial to the national economy and people's livelihood. As the last mile of power supply from the power grid to end users, the operation status of low-voltage distribution areas directly affects the reliability of power supply and user experience. Traditional low-voltage distribution area monitoring technology mainly relies on manual periodic inspections and simple instrument measurements. It depends on maintenance personnel to record basic parameters such as voltage and current on-site and to troubleshoot faults by judging based on experience or fixed thresholds. This is inefficient and makes it difficult to conduct systematic data analysis and early warning. With the development of the Internet of Things and smart sensing technology, smart meters, distribution area monitoring terminals and other equipment have been gradually introduced, realizing the automatic collection and remote transmission of some electricity consumption data, laying the foundation for the automation of monitoring.

[0003] However, current monitoring methods suffer from fragmented multi-source data. The operational status of low-voltage distribution areas is influenced by multiple factors, including spatial topology, building layout, environmental conditions, and user behavior. Existing technologies often focus only on electrical quantity collection, or, although they collect diverse data, they lack effective fusion and analysis methods. This results in the data value not being fully explored, and an inability to build a comprehensive and systematic understanding of the operational risks of distribution areas. Furthermore, in complex distribution areas such as urban villages and old residential communities, different users exhibit significant differences in population structure, electricity usage habits, and building conditions. Their electricity usage behavior also shows distinct patterns at different times, such as weekdays and weekends, and day and night. Using only a uniform fixed threshold or simple model cannot establish differentiated normal behavior models and judgment benchmarks for different time periods and characteristics. This leads to high false alarm and false negative rates in complex scenarios, making it difficult to achieve accurate early warning and efficient monitoring. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a low-voltage distribution area risk monitoring method, system, equipment, and medium based on multi-source data fusion to solve the problems of fragmented multi-source data in existing monitoring methods, which rely heavily on electrical quantity acquisition or lack effective data fusion, making it difficult to form a systematic understanding. Furthermore, these methods do not consider differences at the user end and changes in time periods, and the use of a uniform threshold leads to high false alarm and false alarm rates in complex scenarios, making it impossible to achieve accurate early warning.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a low-voltage transformer area risk monitoring method based on multi-source data fusion, comprising: Collect multi-source data from low-voltage distribution areas, and obtain spatial attribute information and historical data of each user terminal based on the multi-source data; Analyze the historical data of each client terminal to obtain the set of rest environment working conditions and the set of non-rest environment working conditions for each client terminal, and generate fitting trend data based on the set of rest environment working conditions and the set of non-rest environment working conditions. Based on the spatial attribute information, the client's historical data, and the fitted trend data, client clustering is performed to obtain a client rest cluster set and a client non-rest cluster set; Real-time monitoring data of selected users is collected based on the user rest cluster set and user non-rest cluster set; the real-time monitoring data is matched and compared with the corresponding fitted trend data to obtain real-time operation labels; Based on the real-time running labels, an abnormal client set is determined, and combined with the client rest cluster set or client non-rest cluster set corresponding to the abnormal client set and real-time monitoring data, predictive warning data is generated.

[0007] As a preferred embodiment of the low-voltage transformer area risk monitoring method based on multi-source data fusion described in this invention, the steps of extracting spatial attribute information and historical data of each user terminal include: Collect multi-source data of the low-voltage distribution area and construct a three-dimensional model of the distribution area based on the multi-source data; Based on the three-dimensional model of the transformer area, extract the spatial attribute information of each household terminal; Historical client data for each client is obtained from the multi-source data, wherein the historical client data includes historical operation tags, historical period tags, and historical monitoring data.

[0008] The beneficial effects of this preferred technical solution are as follows: by constructing a three-dimensional model of the transformer area, the spatial attributes of the user terminal, such as user terminal coordinates, user terminal floor, and obstruction relationship, are digitally extracted, and physical spatial information is integrated into the monitoring and analysis system, providing a multi-dimensional geographic information foundation for subsequent clustering, improving the dimension and accuracy of data analysis, and making risk monitoring more in line with the actual physical layout.

[0009] As a preferred embodiment of the low-voltage transformer area risk monitoring method based on multi-source data fusion described in this invention, the step of obtaining the set of rest environment operating conditions and the set of non-rest environment operating conditions for each user terminal includes: From the historical data of each client, filter out the historical operation records that are labeled as normal; The operation records are classified according to their corresponding historical period labels as rest or non-rest, resulting in rest period data and non-rest period data; Cluster analysis is performed on the environmental operating condition information in the rest period data and the non-rest period data to obtain the rest environmental operating condition set and the non-rest environmental operating condition set.

[0010] The beneficial effects of this preferred technical solution are as follows: by distinguishing between rest periods and non-rest periods and clustering the environmental conditions of each period, the power consumption patterns in different periods and environments can be characterized, avoiding the limitations of traditional single threshold models, improving the accuracy of subsequent real-time matching and comparison, and reducing the risk of false alarms and missed alarms in complex power consumption scenarios.

[0011] As a preferred embodiment of the low-voltage transformer area risk monitoring method based on multi-source data fusion described in this invention, the step of generating fitting trend data based on the set of rest environment conditions and the set of non-rest environment conditions includes: Trend fitting is performed on multiple historical monitoring data contained in each set of rest environment working conditions and non-rest environment working conditions to obtain the fitting trend map of rest environment working conditions and the fitting trend map of non-rest environment working conditions, as well as the corresponding working condition benchmark information. By combining the fitting trend charts of all rest environment working conditions and non-rest environment working conditions of each client terminal with the corresponding working condition baseline information, the fitting trend data of the client terminal is obtained.

[0012] The beneficial effects of this preferred technical solution are as follows: by performing trend fitting on historical monitoring data of rest environment working conditions set and non-rest environment working conditions set, structured working condition benchmark information and fitting trend map are extracted, which provides a reliable and quantitative basis for environmental matching and trend comparison in subsequent real-time monitoring, and improves the accuracy of anomaly detection and scene adaptability.

[0013] As a preferred embodiment of the low-voltage transformer area risk monitoring method based on multi-source data fusion described in this invention, the steps of obtaining the user-end rest cluster set and the user-end non-rest cluster set include: Based on the spatial attribute information, the historical data of the client terminals, and the fitted trend data, calculate the client rest distance and client non-rest distance between clients; Clustering is performed based on the client's rest distance and non-rest distance to obtain a client rest cluster set and a client non-rest cluster set.

[0014] The beneficial effects of this preferred technical solution are as follows: by comprehensively considering spatial attribute information, historical data of user terminals, and fitted trend data, the distance between user terminals that distinguishes between rest periods and non-rest periods is calculated. Based on the distance between user terminals, clustering is performed, which can identify user terminal groups with similar electricity consumption patterns at different times. This provides the most valuable historical experience for subsequent efficient sampling monitoring and abnormal user terminal location, thereby improving monitoring efficiency and the pertinence of early warning.

[0015] As a preferred embodiment of the low-voltage transformer area risk monitoring method based on multi-source data fusion described in this invention, the step of matching and comparing the real-time monitoring data with the corresponding fitted trend data to obtain the real-time running label includes: The environmental operating condition information in the real-time monitoring data of the selected client is matched with the operating condition baseline information in the fitted trend data to obtain real-time set labels; The trend information in the real-time monitoring data is compared with the fitted trend data corresponding to the obtained real-time set labels to obtain the comparison results; Based on the comparison results, the real-time running tag of the selected client is determined.

[0016] The beneficial effects of this preferred technical solution are as follows: by first matching the environmental conditions to determine the most similar historical pattern, and then comparing the real-time and historical trends to achieve a two-layer judgment, misjudgment caused by environmental or time period mismatch is avoided, and the degree of deviation is quantified by comparison, so that anomalies can still be reliably identified under complex working conditions, thereby improving the accuracy and reliability of real-time monitoring.

[0017] As a preferred embodiment of the low-voltage transformer area risk monitoring method based on multi-source data fusion described in this invention, the step of generating predictive and early warning data includes: Selected clients that are running in real time and are labeled as abnormal are grouped into an abnormal client set. The set of client rest clusters or the set of client non-rest clusters to which each abnormal client belongs in the abnormal client set is taken as the real-time client cluster set of the abnormal client. Based on the real-time monitoring data of the abnormal client, the historical monitoring data with the historical operation tag of abnormal in the client's historical data, and the client's historical data of other clients in the real-time client cluster set, predictive warning data containing warning level, predicted anomaly type and suggested handling measures is generated.

[0018] The beneficial effects of this preferred technical solution are as follows: by comprehensively analyzing the real-time monitoring data of abnormal client terminals, the historical monitoring data of client terminals with the historical operation tag of abnormality in the historical data of the client terminals, and the client terminal historical data of other client terminals in the real-time client terminal cluster set, the solution can be upgraded from simple alarm to intelligent diagnosis, thereby improving the operability of alarms. At the same time, by learning from the processing experience of similar client terminals, the reliability of early warning and the efficiency of handling can be enhanced.

[0019] Secondly, the present invention provides a low-voltage transformer area risk monitoring system based on multi-source data fusion, comprising: The data acquisition module is used to collect multi-source data from the low-voltage distribution area and obtain spatial attribute information and historical data of each user terminal based on the multi-source data. The working condition analysis and fitting module is used to analyze the historical data of each user terminal to obtain the set of rest environment working conditions and the set of non-rest environment working conditions for each user terminal, and generate fitting trend data based on the set of rest environment working conditions and the set of non-rest environment working conditions. The client-side clustering calculation module is used to perform client-side clustering based on the spatial attribute information, the client-side historical data, and the fitted trend data to obtain a client-side rest cluster set and a client-side non-rest cluster set. The real-time monitoring and diagnosis module is used to collect real-time monitoring data of selected users based on the user rest cluster set and the user non-rest cluster set, and to match and compare the real-time monitoring data with the corresponding fitting trend data to obtain real-time operation labels; The early warning and handling module is used to determine the abnormal client set based on the real-time operation tag, and generate predictive early warning data by combining the client rest cluster set or client non-rest cluster set corresponding to the abnormal client set and real-time monitoring data.

[0020] Thirdly, the present invention provides an electronic device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the low-voltage distribution area risk monitoring method based on multi-source data fusion.

[0021] Fourthly, the present invention provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the low-voltage distribution area risk monitoring method based on multi-source data fusion.

[0022] Compared with existing technologies, the beneficial effects of this invention are as follows: by distinguishing between rest periods and non-rest periods, a multi-mode environmental condition set and fitting trend benchmark are established, and clustering is performed based on the distance between user terminals, thereby improving the accuracy and scenario adaptability of anomaly monitoring, reducing false alarm rate and missed alarm rate. This is particularly effective in complex scenarios with large differences between user terminals. At the same time, through sampling monitoring and intelligent early warning, both monitoring efficiency and handling guidance are taken into account, which buys time for proactive handling of risk prevention and control, and realizes three-dimensional and refined monitoring of the operational risks of low-voltage distribution areas. Attached Figure Description

[0023] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a schematic diagram of the overall process of a low-voltage distribution area risk monitoring method based on multi-source data fusion according to an embodiment of the present invention. Detailed Implementation

[0025] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0026] Example 1, referring to Figure 1 As an embodiment of the present invention, a low-voltage transformer area risk monitoring method based on multi-source data fusion is provided, comprising: S100. Collect multi-source data from the low-voltage distribution area, and obtain spatial attribute information and historical data of each user terminal based on the multi-source data.

[0027] S200. Analyze the historical data of each client terminal to obtain the set of rest environment working conditions and the set of non-rest environment working conditions for each client terminal, and generate fitting trend data based on the set of rest environment working conditions and the set of non-rest environment working conditions.

[0028] S300. Based on the spatial attribute information, the client's historical data, and the fitted trend data, perform client clustering to obtain a client rest cluster set and a client non-rest cluster set.

[0029] S400. Collect real-time monitoring data of selected users based on the user rest cluster set and user non-rest cluster set; match and compare the real-time monitoring data with the corresponding fitting trend data to obtain real-time operation labels.

[0030] S500. Determine the abnormal client set based on the real-time operation label, and generate prediction and early warning data by combining the client rest cluster set or client non-rest cluster set corresponding to the abnormal client set and the real-time monitoring data.

[0031] It should be noted that current low-voltage distribution area monitoring often faces challenges such as fragmented multi-source data, large differences in user behavior, and poor adaptability of fixed threshold models, resulting in high false alarm and false alarm rates in complex scenarios such as urban villages, making it difficult to achieve efficient risk early warning.

[0032] Therefore, to address the aforementioned issues of data fusion difficulties and poor scenario adaptation, the S100-S500 steps are used to first collect and fuse multi-source data to construct a three-dimensional model of the transformer area. Then, by differentiating time periods and environmental conditions, cluster analysis is performed to form a set of resting and non-resting environmental conditions for each user terminal, generating fitting trend data. Based on spatial attribute information, user terminal historical data, and fitting trend data, user terminal clustering is performed to obtain user terminal resting cluster sets and user terminal non-resting cluster sets. Through matching comparison and intelligent judgment, predictive and early warning data is generated, achieving a synergistic improvement in monitoring accuracy and efficiency, and providing reliable technical support for proactive risk prevention and control in complex low-voltage transformer areas.

[0033] Example 2, refer to Figure 1 As an embodiment of the present invention, based on the above embodiment, a low-voltage distribution area risk monitoring method based on multi-source data fusion is provided.

[0034] In this embodiment, S100 involves collecting multi-source data from the low-voltage distribution area and obtaining spatial attribute information and historical data for each customer terminal based on the multi-source data. Taking an urban village distribution area as an application scenario, the specific implementation of the final load point parameter acquisition and load parameter model construction steps for A1~A3 in S100 is as follows: A1. Collect multi-source data of the low-voltage distribution area and construct a three-dimensional model of the distribution area based on the multi-source data.

[0035] Specifically, the data is obtained by integrating Geographic Information System (GIS), Building Information Modeling (BIM) data, collecting data on transformer substation buildings in urban villages, data on alleyways such as the direction and width of alleyway networks, data on public spaces such as the location of public activity areas and power distribution facilities, and data on transformer substation topology as multi-source data. Based on multi-source data, a 3D modeling engine is used for data fusion and rendering to construct a 3D visualized model of the transformer area.

[0036] A2. Extract the spatial attribute information of each household terminal based on the three-dimensional model of the transformer area.

[0037] Specifically, in the 3D model of the transformer substation, each household terminal corresponds to an actual power consumption unit. The coordinates of each household terminal in three-dimensional space are analyzed and obtained through the coordinate system built into the 3D model of the transformer substation. , as the client coordinates; Based on the building data of the unit's location, the floor level of the unit is determined. ; By analyzing the line-of-sight path from the customer's coordinates to key associated distribution nodes, such as the connected meter box, the degree of obstruction by obstacles such as buildings, billboards, and trees along the line-of-sight path is assessed, and an obstruction coefficient between 0 and 1 is calculated. , The smaller the value, the more severe the occlusion. The spatial attribute information of a user terminal consists of its coordinates, floor level, and occlusion coefficient.

[0038] A3. Obtain the client-side historical data of each client from the multi-source data, wherein the client-side historical data includes historical operation tags, historical cycle tags, and historical monitoring data.

[0039] Specifically, historical data of each household is collected from data sources such as smart meters, transformer area monitoring terminals and environmental sensors, for each household over the past year, i.e., a specified time period. Each piece of client historical data includes historical operation tags for the overall operation status of the client, with values ​​of normal or abnormal; historical period tags for marking the time period to which a specified time period belongs, which are divided into rest periods such as weekends, statutory holidays or non-rest periods such as weekdays according to date type; and historical monitoring data that differs based on the different contents of the historical operation tags. If the historical operation label is normal, the monitoring record of the historical monitoring data includes: multiple sampling time points collected according to the first sampling interval, such as 15 minutes, and electrical characteristic values ​​constituting the historical first electrical vector. The electrical characteristics include current. ,Voltage ,power The device feature values ​​that constitute the first historical device vector include switching temperature. And so on, as well as the historical environmental condition vector obtained from comprehensive monitoring; When the historical operation label is abnormal, the historical monitoring data will also include: abnormality type such as overload, leakage, and abnormality start time. Abnormal End Time Anomaly handling measures and results; Electrical characteristic values ​​and equipment characteristics: Electrical characteristic value sampling strategies differ: during the anomaly duration... Data is collected at a more frequent second sampling interval, such as 1 minute, to form the historical second electrical vector and the historical second device vector; during other normal periods, data is collected at the first sampling interval to form the historical first electrical vector and the historical first device vector.

[0040] In an optional implementation, UAV oblique photogrammetry may be introduced in step S100. The steps are as follows: high-resolution real-scene images and point cloud data of the surface and buildings of the substation area are obtained by UAV oblique photogrammetry, and a real-scene 3D model reflecting the real appearance is generated; the real-scene 3D model is registered and fused with the substation 3D model constructed based on GIS, BIM and other technologies in step A1, which has both semantic information and real texture details, and optimizes the calculation of spatial attribute information such as occlusion coefficient in step A2.

[0041] In another optional implementation, step S100 may also introduce basic household data, which involves: synchronously acquiring basic household data for each household from the property management system or census data, including household location such as address and household population. User area And population tags for each resident, such as {infants, students, seniors, working professionals, and homeowners}, and generate household tag vectors based on the actual number of tags for each household. ,like This indicates that there are 2 students, who are part of the client and will be used for subsequent cluster analysis.

[0042] In this embodiment of the application, step S200 involves analyzing the historical data of each client terminal to obtain a set of rest environment working conditions and a set of non-rest environment working conditions for each client terminal, and generating fitting trend data based on the set of rest environment working conditions and the set of non-rest environment working conditions. Step S200 includes B1~B2: B1. Perform trend fitting on multiple historical monitoring data contained in each set of rest environment working conditions and non-rest environment working conditions to obtain the fitting trend map of rest environment working conditions and the fitting trend map of non-rest environment working conditions, as well as the corresponding working condition benchmark information.

[0043] Specifically, iterate through each set of rest environment conditions. The monitoring records include all historical monitoring data, and for each monitoring record, the electrical characteristics such as current in the historical first electrical vector are considered. Plot the change curve of the electrical characteristic in chronological order within the specified time period to obtain the historical electrical trend chart. ,in For record indexing; For each set of rest environment working conditions Historical electrical trend graph of the same electrical characteristic from all monitoring records. The moving average method was used to perform point-by-point averaging, and the set of working conditions for each rest environment was obtained by fitting. Electrical fitting trend of current Repeat this process for all electrical features and all equipment features to obtain a fitting trend chart for all electrical features. Fitting trend chart of equipment characteristic states , as a trend chart of the working conditions of the resting environment; Similarly, a fitting trend chart of non-rest environment operating conditions is generated for the set of non-rest environment operating conditions.

[0044] The steps for obtaining the set of rest environment conditions and the set of non-rest environment conditions for each client include B1.1 to B1.3: B1.1. Select the historical operation records with normal operation tags from the historical data of each client.

[0045] Specifically, for the i-th client, iterate through all its historical data for a specified time period, such as the past 365 days; for each specified time period, check its historical operation tag, and only retain the operation records with the historical operation tag as normal, thus forming a set of normal operation records for the client. Each operation record in the normal operation record set contains a historical first electrical vector, a historical first equipment vector, and a historical environmental condition vector based on the first sampling interval within a complete specified time period.

[0046] B1.2. The operation records are classified according to their corresponding historical period labels as rest or non-rest, resulting in rest period data and non-rest period data.

[0047] Specifically, for the normal operation record set For each run record, check the historical cycle label and categorize run records with the historical cycle label as rest periods into the rest period dataset. Run records labeled as not being in rest periods will be categorized into the non-rest period dataset. .

[0048] B1.3. Perform cluster analysis on the environmental operating condition information in the rest period data and the non-rest period data to obtain the rest environmental operating condition set and the non-rest environmental operating condition set.

[0049] Specifically, for the rest period dataset Extract the historical environmental condition vector for each running record. Use K-means clustering algorithm to Clustering is performed, and after clustering, all historical monitoring data records belonging to the same cluster are grouped into a rest environment condition set. ,in Let r be the set index, then we get A collection of rest environment working conditions Each set of rest environment conditions is assigned a rest set label, such as a low temperature and high humidity rest set label, and the environmental condition fitting vector is calculated. This serves as the work condition information corresponding to the rest period; Non-rest period dataset Perform the same operation to obtain A collection of non-rest environment working conditions and the corresponding non-rest set labels and environmental condition fitting vectors This serves as the baseline information for operating conditions during non-rest periods.

[0050] B2. By combining the fitting trend charts of all rest environment working conditions and non-rest environment working conditions of each client terminal with the corresponding working condition benchmark information, the fitting trend data of the client terminal is obtained.

[0051] Specifically, the fitting trend charts of all rest environment working conditions and non-rest environment working conditions of the i-th client are integrated with the corresponding working condition baseline information into structured fitting trend data. It contains two sub-dictionaries: Corresponding rest periods, For non-rest periods, each sub-dictionary contains an environmental condition fitting vector. Set of fitting trend graphs for all operating condition baseline information .

[0052] In an optional implementation, step S200 can also use an adaptive clustering algorithm instead of K-means. The steps are as follows: use a density-based spatial clustering algorithm to cluster the environmental condition vectors in the rest period or non-rest period dataset, identify noise points and determine the number of clusters, and set the neighborhood radius. and minimum sample size The environmental condition vector is divided into several core clusters and boundary points. After removing noise points, an environmental condition set is formed. This method is suitable for scenarios where the environmental conditions are unevenly distributed or where there are discrete anomalies.

[0053] In another optional implementation, step S200 may further compress and encode the fitted trend data. The steps are as follows: for each electrical fitted trend map or equipment status fitted trend map, use discrete cosine transform to convert it from the time domain to the frequency domain, retain the most important frequency coefficients, discard high-frequency details, normalize and quantize the environmental condition fitted vector, and package the encoded coefficients, quantization parameters and metadata such as set tags into binary data blocks as the compressed fitted trend data storage. When used, the fitted trend map and the environmental condition fitted vector can be quickly recovered through inverse transformation and inverse quantization.

[0054] In this embodiment of the application, step S300 involves performing client clustering based on the spatial attribute information, the client's historical data, and the fitted trend data to obtain a client rest cluster set and a client non-rest cluster set. Step S300 includes C1~C2: C1. Calculate the rest distance and non-rest distance between users based on the spatial attribute information, the user historical data, and the fitted trend data.

[0055] Specifically, suppose there are N households in the data set of the transformer substation. For any two households, such as the i-th household and the j-th household, the rest distance between the households is... The calculation formula is as follows: The formula for calculating the static distance component is: in, This represents the set of basic parameters for the i-th household, corresponding to the household population, household area, household floor, and obstruction coefficient, respectively. This represents the k-th basic parameter in the basic parameter set of the i-th client in the transformer area client data; The weights for the k-th basic parameter are set to values ​​of 0.15, 0.1, 0.1, and 0.2, respectively. This represents the number of people with the m-th population label, such as infants or students, in the user label vector of the i-th user. Let be the coordinates of the i-th client. This represents the maximum value of the coordinate distance between all client pairs; Calculate the dynamic distance component based on the trend of rest periods: in, The Jaccard similarity coefficient represents the set labels of the two households' rest areas. , For the nth rest set label of the i-th client, It is an indicator function , For the nth rest set label of the i-th client, Let N1 be the p-th rest set label of the j-th client, and be 1 when two clients have the same rest set label; N1 is the number of rest set labels of the i-th client; N2 is the number of rest set labels of the j-th client. This represents the trend chart of all electrical features fitted under a certain set of labels for two households. Fitting trend chart of equipment characteristic states The average of the time-normalized distances between them; and These are the quantities of electrical features and equipment features, respectively. The final client rest distance, which is the geometric mean of the static and dynamic distance components, is: Client non-rest distance The calculation formula is similar to that described above. All calculations are based on the non-rest set labels and their corresponding non-rest environment condition fitting trend charts.

[0056] C2. Cluster the client rest distance and client non-rest distance respectively to obtain the client rest cluster set and the client non-rest cluster set.

[0057] Specifically, with For example, using a hierarchical clustering algorithm, each client is initially a cluster. The two closest clusters are iteratively merged, and the distance between clusters is defined as the average distance between all client pairs within the two clusters. The iteration continues until the following clustering optimization objective function converges. in, It is an indicator function, which is 1 when client i and j belong to the same cluster, and 0 otherwise; This represents the e-th cluster. To balance the coefficients of intra-cluster compactness and inter-cluster separation, each cluster formed after convergence is a client-side rest cluster set; N5 is the total number of current clusters; right Repeat the clustering process to obtain the client-side non-resting cluster set.

[0058] In an optional implementation, step S300 may also employ a graph-based spectral clustering algorithm, the steps of which are: using... and As a foundation for constructing a similarity matrix, such as through a Gaussian kernel function... The distance is converted into similarity, a fully connected graph is constructed, the Laplacian matrix of the fully connected graph is calculated, and K-means clustering is performed to obtain a specified number of client-side resting cluster sets and client-side non-resting cluster sets. It can handle non-convex clusters and is more robust to noise in distance metrics.

[0059] In another alternative implementation, step S300 can also introduce more dimensional features, the step being: calculating the dynamic distance component In addition to considering the DTW distance between the electrical fitting trend chart and the equipment condition fitting trend chart, the similarity calculation of historical environmental condition vectors in historical monitoring data is also added. Sequences of historical environmental condition vectors from multiple historical monitoring data points for each user terminal under the same rest set label or non-rest set label are extracted, and the comprehensive distance between the sequences is calculated, such as based on time warping or Euclidean distance, and then weighted and incorporated. The calculation reflects the similarity of client behavior under specific environmental conditions.

[0060] In this embodiment of the application, S400 involves collecting real-time monitoring data of selected clients based on the client rest cluster set and the client non-rest cluster set; matching and comparing the real-time monitoring data with the corresponding fitted trend data to obtain real-time operation labels. Step S400 includes D1~D3: D1. Match the environmental operating condition information in the real-time monitoring data of the selected client with the operating condition benchmark information in the fitted trend data to obtain real-time set labels.

[0061] Specifically, the current period label is determined based on the current date. If the current period label is rest, then each client rest cluster is sampled. Based on the total number of clients contained in each client rest cluster, the random sample number of the client rest cluster is determined at a sampling ratio of, for example, 10%, and a corresponding number of random clients are randomly selected from the client rest cluster as selected clients. For each selected client, based on the first sampling interval such as 15 minutes, real-time sampling data at the current sampling time point within the real-time time period such as the current date is collected by its client line monitoring device at the first sampling interval. Collect current real-time environmental condition data, integrate real-time sampling data from the current and previous sampling time points to obtain real-time monitoring data for the selected client, and extract the real-time environmental condition vector from it. The real-time environmental condition vector of the selected client It fits the trend data In the middle, the environmental condition fitting vector of all rest environment condition sets corresponding to the current period label. Perform similarity comparison and calculate With each The cosine similarity is used to select the rest set label corresponding to the environmental condition fitting vector with the lowest similarity, which is then used as the real-time set label for the selected client. ; If the current cycle label is non-rest, the same operation is performed based on the client's non-rest cluster set to match and obtain the non-rest set label as the real-time set label.

[0062] D2. Compare the trend information in the real-time monitoring data with the fitted trend data corresponding to the obtained real-time set labels to obtain the comparison result.

[0063] Specifically, based on all real-time sampling data collected by the selected client within the current real-time period, a real-time electrical trend chart for each electrical characteristic is plotted. Real-time device status trend chart for each device feature ; Fitting trend data from the selected client In the middle, based on real-time collection tags Locate the corresponding electrical fitting trend chart Fitting trend chart of equipment status ; The time warping algorithm is used to calculate the real-time electrical trend charts. Corresponding electrical fitting trend chart Electrical alignment distance between and real-time device status trend chart Trend chart fitted with corresponding equipment status Equipment status regularization distance between ; Adjust electrical spacing Distance from equipment status normalization As a comparison result .

[0064] D3. Based on the comparison results, determine the real-time running tag of the selected client.

[0065] Specifically, based on the comparison results The calculated electrical regularization distances of all electrical characteristics And the device status regularization distance of all device characteristics Each is related to the preset electrical distance threshold. Distance threshold between preset devices Compare; When any electrical characteristic exists or any device feature If the electrical or equipment feature is not marked as an abnormal feature, the real-time operation label of the selected client in the current real-time period is determined to be an abnormal label; if all electrical regularization distances and state regularization distances do not exceed their corresponding preset thresholds, the real-time operation label of the selected client is determined to be a normal label.

[0066] In an optional implementation, a weighted matching strategy may be introduced in step S400, wherein when calculating the similarity between the real-time environmental condition vector and the historical environmental condition fitting vector, different weights are assigned to different environmental parameters based on historical data analysis or expert experience. If the temperature weight is set to 0.4 and the humidity weight is set to 0.3, the weighted similarity will be... Calculated as The historical set labels with the lowest weighted similarity are selected as the real-time set labels to reflect the key environmental factors that have the greatest impact on electricity consumption behavior.

[0067] In another optional implementation, step S400 can also introduce a confidence assessment mechanism after obtaining the real-time running labels, the steps of which are: calculating the confidence level. ,like ,in It is the total number of electrical and equipment features participating in the comparison. A confidence value is attached to the real-time running tag. For example, the confidence of the abnormal tag is 85%. When the confidence is lower than a certain level, such as 70%, it is marked as to be observed and a more intensive second sampling interval is triggered for short-term tracking and monitoring to reduce misjudgment caused by single sampling noise or instantaneous fluctuations.

[0068] In this embodiment of the application, S500 involves determining an abnormal client set based on the real-time operation tags, and generating predictive warning data by combining the client rest cluster set or client non-rest cluster set corresponding to the abnormal client set with real-time monitoring data. Step S500 includes E1~E3: E1. Group the selected clients that are running in real time and are labeled as abnormal into an abnormal client set.

[0069] Specifically, iterate through all selected clients with confirmed real-time running tags, filter out all clients with the real-time running tag set to "abnormal", and aggregate them into an "abnormal client set". ,in Select the number of clients for anomalies detected within the current real-time period.

[0070] E2. The client rest cluster set or client non-rest cluster set to which each abnormal client belongs in the abnormal client set is taken as the real-time client cluster set of the abnormal client.

[0071] Specifically, for the abnormal client set Each abnormal client in Based on the current cycle label, the time period category is determined. If the current cycle label is a rest period, then the rest period cluster set obtained in S300 is used. In, search for containing The client rest cluster set is marked as an abnormal client. Real-time client clustering set ; Includes with Other users with similar electricity consumption patterns during the corresponding time period; If the current period label is a non-rest period, then in all client non-rest cluster sets Perform the same search operation within it.

[0072] E3. Based on the real-time monitoring data of the abnormal client, the historical monitoring data with the historical operation tag of abnormal in the client's historical data, and the client's historical data of other clients in the real-time client cluster set, generate predictive warning data that includes warning level, predicted anomaly type and suggested handling measures.

[0073] Specifically, for each abnormal client Acquire real-time monitoring data, including real-time electrical trend graphs of all electrical characteristics marked as abnormal features. Real-time device status trend chart with device characteristics ;from From the client's historical data, extract all historical monitoring data with the historical operation tag marked as abnormal to form a historical abnormality dataset. ; For abnormal clients Real-time client clustering set Each of the other clients ( ), from other clients Extract all historical operation data from the client's historical data that are tagged as abnormal and whose historical period tags match the abnormal client data. The historical monitoring data with the same current period label constitutes each other client Real-time client reference data Historical monitoring data with the "abnormal" tag in the client's historical data is considered historical abnormal monitoring data. Quantitative real-time client clustering set Each other client With the current abnormal client The similarity is calculated for each Abnormal reference value : in, Other clients Real-time client reference data Number of historical anomaly monitoring data entries in China; The current time; yes The Middle The start time of the anomaly in the historical anomaly monitoring data; and These are abnormal client terminals. The number of abnormal electrical characteristics and abnormal equipment characteristics; and They are The Middle The first historical anomaly monitoring data Historical electrical trend chart of the first electrical feature and the first Historical device status trend chart for each device characteristic; and It is an indicator function, when The One abnormal electrical characteristic and The value is 1 if the corresponding electrical characteristics are the same, and 0 otherwise; the same applies to equipment characteristics. Indicates time-warped distance calculation; exponential term This is the time decay factor; abnormal client Real-time monitoring data; historical anomaly dataset Real-time client-side clustering set All other clients Real-time client reference data and the calculated abnormal reference values Input a comprehensive analysis module, which analyzes current anomaly patterns, matches historical anomaly types, references the handling experience of other clients, and considers factors such as time proximity to output the anomaly client. Forecast and early warning data; The predictive and early warning data includes: early warning levels such as high, medium, and low, which are determined based on a comprehensive assessment of the degree of abnormal deviation, the scope of impact, and the historical recurrence frequency; predicted anomaly types such as suspected line overload, suspected leakage, and suspected poor equipment contact, which are based on pattern matching between real-time and historical trend charts and statistics of historical anomaly types for similar clients; and suggested handling measures such as suggesting remote adjustment of the load on the branch line and arranging on-site verification within 2 hours, or suggesting notifying users to check high-power equipment first, which are generated based on the success rate of historical handling measures and effective handling experience for similar clients.

[0074] In an optional implementation, step S500 may also introduce an expert knowledge base for decision support. The steps are as follows: In the comprehensive analysis module, in addition to utilizing the data-driven analysis results, a predefined expert knowledge base is also accessed. This knowledge base stores the mapping relationships between domain experts and specific anomaly patterns, such as voltage drop curves of specific shapes, current harmonic characteristics and anomaly types, risk levels and handling measures, in the form of rules. When a rule pattern is matched with a rule pattern in the knowledge base, the rule-recommended early warning information will receive higher priority or weight. This information is then weighted and fused with the results of the data-driven analysis to generate the final prediction and early warning data, thereby improving the reliability and interpretability of the decision.

[0075] In another optional implementation, step S500 can also perform multi-dimensional verification and optimization of the early warning information. The steps are as follows: instead of immediately releasing the predicted early warning data, verification is first performed in a sandbox environment, using real-time client-side clustering sets. The system simulates the real-time data streams of other currently normal user terminals, mimicking the potential impact of implementing recommended handling measures. For example, it simulates voltage changes at relevant user terminals after reducing the load on a certain line. The warning information is cross-validated with the total incoming line monitoring data at the transformer substation level and weather forecast information such as high temperature and thunderstorm warnings for the next few hours. Only when the verification confirms that the warning does not conflict with the overall status of the transformer substation and has a high correlation with external environmental risks will the predicted warning data be officially released and pushed to maintenance personnel. Otherwise, the warning level will be downgraded or changed to observation status.

[0076] In summary, this invention collects multi-source data from low-voltage distribution areas and constructs a three-dimensional model of the distribution areas; analyzes historical data from customer terminals, distinguishes between rest periods and non-rest periods, and clusters them to obtain a set of environmental operating conditions, generating fitted trend data; integrates spatial attribute information, historical data from customer terminals, and fitted trends to perform customer terminal clustering, obtaining a set of customer terminal rest clusters and a set of customer terminal non-rest clusters; conducts sampling monitoring, and determines real-time operation labels through comparison; generates predictive early warning data containing warning levels, types, and measures, achieving accurate, efficient monitoring and proactive early warning of operational risks in complex low-voltage distribution areas.

[0077] Example 3 illustrates a schematic scheme for a low-voltage transformer area risk monitoring method based on multi-source data fusion. It should be noted that the technical solution of this low-voltage transformer area risk monitoring system based on multi-source data fusion belongs to the same concept as the technical solution of the aforementioned low-voltage transformer area risk monitoring method based on multi-source data fusion. Details not described in detail in the technical solution of the low-voltage transformer area risk monitoring system based on multi-source data fusion in this embodiment can be found in the description of the technical solution of the aforementioned low-voltage transformer area risk monitoring method based on multi-source data fusion.

[0078] This embodiment also provides a low-voltage distribution area risk monitoring system based on multi-source data fusion, including: The data acquisition module is used to collect multi-source data from the low-voltage distribution area and obtain spatial attribute information and historical data of each user terminal based on the multi-source data. The working condition analysis and fitting module is used to analyze the historical data of each user terminal to obtain the set of rest environment working conditions and the set of non-rest environment working conditions for each user terminal, and generate fitting trend data based on the set of rest environment working conditions and the set of non-rest environment working conditions. The client-side clustering calculation module is used to perform client-side clustering based on the spatial attribute information, the client-side historical data, and the fitted trend data to obtain a client-side rest cluster set and a client-side non-rest cluster set. The real-time monitoring and diagnosis module is used to collect real-time monitoring data of selected users based on the user rest cluster set and the user non-rest cluster set, and to match and compare the real-time monitoring data with the corresponding fitting trend data to obtain real-time operation labels; The early warning and handling module is used to determine the abnormal client set based on the real-time operation tag, and generate predictive early warning data by combining the client rest cluster set or client non-rest cluster set corresponding to the abnormal client set and real-time monitoring data.

[0079] This embodiment also provides an electronic device suitable for low-voltage distribution area risk monitoring based on multi-source data fusion, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the low-voltage distribution area risk monitoring method based on multi-source data fusion as proposed in the above embodiment.

[0080] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the low-voltage distribution area risk monitoring method based on multi-source data fusion as proposed in the above embodiments.

[0081] The storage medium proposed in this embodiment and the low-voltage distribution area risk monitoring method based on multi-source data fusion proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0082] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0083] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for low-voltage transformer area risk monitoring based on multi-source data fusion, characterized in that, include: Collect multi-source data from low-voltage distribution areas, and obtain spatial attribute information and historical data of each user terminal based on the multi-source data; Analyze the historical data of each client terminal to obtain the set of rest environment working conditions and the set of non-rest environment working conditions for each client terminal, and generate fitting trend data based on the set of rest environment working conditions and the set of non-rest environment working conditions. Based on the spatial attribute information, the client's historical data, and the fitted trend data, client clustering is performed to obtain a client rest cluster set and a client non-rest cluster set; Real-time monitoring data of selected clients are collected based on the client rest cluster set and the client non-rest cluster set; The real-time monitoring data is matched and compared with the corresponding fitted trend data to obtain the real-time running label; Based on the real-time running labels, an abnormal client set is determined, and combined with the client rest cluster set or client non-rest cluster set corresponding to the abnormal client set and real-time monitoring data, predictive warning data is generated.

2. The low-voltage distribution area risk monitoring method based on multi-source data fusion as described in claim 1, characterized in that, The steps for extracting spatial attribute information and historical data from each client include: Collect multi-source data of the low-voltage distribution area and construct a three-dimensional model of the distribution area based on the multi-source data; Based on the three-dimensional model of the transformer area, extract the spatial attribute information of each household terminal; Historical client data for each client is obtained from the multi-source data, wherein the historical client data includes historical operation tags, historical period tags, and historical monitoring data.

3. The low-voltage distribution area risk monitoring method based on multi-source data fusion as described in claim 2, characterized in that, The steps to obtain the set of rest environment working conditions and the set of non-rest environment working conditions for each client include: From the historical data of each client, filter out the historical operation records that are labeled as normal; The operation records are classified according to their corresponding historical period labels as rest or non-rest, resulting in rest period data and non-rest period data; Cluster analysis is performed on the environmental operating condition information in the rest period data and the non-rest period data to obtain the rest environmental operating condition set and the non-rest environmental operating condition set.

4. The low-voltage distribution area risk monitoring method based on multi-source data fusion as described in claim 3, characterized in that, The steps for generating fitted trend data based on the set of rest environment working conditions and the set of non-rest environment working conditions include: Trend fitting is performed on multiple historical monitoring data contained in each set of rest environment working conditions and non-rest environment working conditions to obtain the fitting trend map of rest environment working conditions and the fitting trend map of non-rest environment working conditions, as well as the corresponding working condition benchmark information. By combining the fitting trend charts of all rest environment working conditions and non-rest environment working conditions of each client terminal with the corresponding working condition baseline information, the fitting trend data of the client terminal is obtained.

5. The low-voltage distribution area risk monitoring method based on multi-source data fusion as described in claim 4, characterized in that, The steps to obtain the client-side rest cluster set and the client-side non-rest cluster set include: Based on the spatial attribute information, the historical data of the client terminals, and the fitted trend data, calculate the client rest distance and client non-rest distance between clients; Clustering is performed based on the client's rest distance and non-rest distance to obtain a client rest cluster set and a client non-rest cluster set.

6. The low-voltage distribution area risk monitoring method based on multi-source data fusion as described in claim 5, characterized in that, The steps for matching and comparing the real-time monitoring data with the corresponding fitted trend data to obtain the real-time running label include: The environmental operating condition information in the real-time monitoring data of the selected client is matched with the operating condition baseline information in the fitted trend data to obtain real-time set labels; The trend information in the real-time monitoring data is compared with the fitted trend data corresponding to the obtained real-time set labels to obtain the comparison results; Based on the comparison results, the real-time running tag of the selected client is determined.

7. The low-voltage distribution area risk monitoring method based on multi-source data fusion as described in claim 6, characterized in that, The steps for generating forecast and early warning data include: Selected clients that are running in real time and are labeled as abnormal are grouped into an abnormal client set. The set of client rest clusters or the set of client non-rest clusters to which each abnormal client belongs in the abnormal client set is taken as the real-time client cluster set of the abnormal client. Based on the real-time monitoring data of the abnormal client, the historical monitoring data with the historical operation label of abnormal in the client's historical data, and the client's historical data of other clients in the real-time client cluster set, predictive warning data containing warning level, predicted anomaly type and suggested handling measures is generated.

8. A low-voltage distribution area risk monitoring system based on multi-source data fusion, employing the method described in any one of claims 1-7, characterized in that, include: The data acquisition module is used to collect multi-source data from the low-voltage distribution area and obtain spatial attribute information and historical data of each user terminal based on the multi-source data. The working condition analysis and fitting module is used to analyze the historical data of each user terminal to obtain the set of rest environment working conditions and the set of non-rest environment working conditions for each user terminal, and generate fitting trend data based on the set of rest environment working conditions and the set of non-rest environment working conditions. The client-side clustering calculation module is used to perform client-side clustering based on the spatial attribute information, the client-side historical data, and the fitted trend data to obtain a client-side rest cluster set and a client-side non-rest cluster set. The real-time monitoring and diagnosis module is used to collect real-time monitoring data of selected users based on the user rest cluster set and the user non-rest cluster set, and to match and compare the real-time monitoring data with the corresponding fitting trend data to obtain real-time operation labels; The early warning and handling module is used to determine the abnormal client set based on the real-time operation tag, and generate predictive early warning data by combining the client rest cluster set or client non-rest cluster set corresponding to the abnormal client set and real-time monitoring data.

9. An electronic device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the low-voltage distribution area risk monitoring method based on multi-source data fusion as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the low-voltage distribution area risk monitoring method based on multi-source data fusion as described in any one of claims 1 to 7.

Images