Method and system for evaluating reliability of medium and low voltage power supply based on HLNN model

Abstract

The present application relates to the technical field of machine learning, and more particularly to a medium and low voltage power supply reliability evaluation method and system based on a HLNN model. The method comprises the following steps: obtaining power supply use time sequence data corresponding to medium voltage lines, distribution transformers and low voltage users, and constructing a hierarchical long short-term memory neural network (HLNN) model for hierarchical multi-scale feature analysis to obtain line-transformer-user corresponding hierarchical different time scale power failure features; based on line-transformer-user corresponding hierarchical different time scale power failure feature analysis, the topology connection construction of line-transformer-user corresponding power supply space topology relationship is analyzed to generate a line-transformer-user power supply topology relationship graph model; based on the line-transformer-user power supply topology relationship graph model, low voltage power failure influence evaluation and user number hierarchical calculation are performed, and medium and low voltage power supply reliability evaluation is simultaneously performed to obtain line-transformer-user corresponding medium and low voltage power supply reliability measurement. The present application can ensure the accuracy of power system reliability calculation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of machine learning, in particular to a middle and low voltage power supply reliability evaluation method and system based on a HLNN model. BACKGROUND

[0002] With the continuous development of social economy, the demand for electricity is increasing, and the stability and reliability of the power system have become an important foundation to ensure the normal operation of society, especially the middle and low voltage power supply system, as an important part of the power network, directly affects the normal operation of residents' life and industrial production, how to improve the reliability of the middle and low voltage power supply system, reduce power failure accidents, and accurately evaluate the reliability of the system, has become a key problem to be solved in the power industry.

[0003] In recent years, with the rapid development of artificial intelligence technology, power supply reliability evaluation methods based on machine learning and deep learning have been gradually introduced, showing good prediction accuracy and intelligent characteristics, especially long short-term memory neural networks, which can effectively handle long-term dependencies in time series data, have become an important tool for predicting power system failures and reliability. Traditional reliability evaluation methods mainly rely on statistical analysis based on historical data, common indicators include power outage frequency, power outage duration and power supply quality. However, these traditional methods usually ignore the interaction of various complex factors in the power system, and have weak prediction ability for dynamically changing power system environment, in addition, the standard memory neural network model still has certain limitations when processing multi-dimensional data of complex power systems, such as insufficient modeling capability for multi-level and complex structure data, thereby reducing the accuracy and efficiency of power supply reliability evaluation. SUMMARY

[0004] Therefore, it is necessary to provide a middle and low voltage power supply reliability evaluation method and system based on a HLNN model to solve at least one of the above technical problems.

[0005] To achieve the above purpose, a middle and low voltage power supply reliability evaluation method based on a HLNN model comprises the following steps:

[0006] Step S1: Obtain the power supply use time series data corresponding to the middle voltage line, distribution transformer and low voltage user, and construct a hierarchical long short-term memory neural network HLNN model, and simultaneously perform hierarchical multi-scale feature analysis on the power supply use time series data corresponding to the middle voltage line, distribution transformer and low voltage user based on the hierarchical long short-term memory neural network HLNN model, to obtain different time scale power failure features corresponding to the line-transformer-user hierarchy;

[0007] Step S2: based on the line-variable-house corresponding layered different time scale power failure feature analysis of line-variable-house corresponding power supply space topology relationship, and according to the line-variable-house corresponding power supply space topology relationship corresponding medium voltage line, distribution transformer and low voltage user as a graph node topological connection construction, to generate line-variable-house power topology relationship graph model;

[0008] Step S3: based on line-variable-house power topology relationship graph model and combined with the corresponding graph node of medium voltage line and distribution transformer, the low voltage user corresponding graph node is evaluated for low voltage power failure influence, to obtain the power failure influence range of line-variable corresponding to low voltage user;According to the power failure influence range of line-variable corresponding to low voltage user, the time household number layer calculation is carried out, and the power failure influence time household number corresponding to distribution line, distribution transformer and meter is obtained;

[0009] Step S4: based on the power failure influence time household number corresponding to distribution line, distribution transformer and meter, the graph node of line-variable-house power topology relationship graph model corresponding to medium voltage line, distribution transformer and low voltage user is evaluated for medium and low voltage power supply reliability, to obtain the medium and low voltage power supply reliability measurement of line-variable-house corresponding.

[0010] Further, step S1 includes the following steps:

[0011] Step S11: obtaining the power supply use time sequence data corresponding to the medium voltage line;

[0012] Step S12: obtaining the power supply use time sequence data corresponding to the distribution transformer;

[0013] Step S13: obtaining the power supply use time sequence data corresponding to the low voltage user;

[0014] Step S14: constructing a hierarchical long short-term memory neural network HLNN model, which includes low, medium and high layer corresponding long short-term memory neural network architecture, the long short-term memory neural network architecture corresponding to the low layer is used to extract the different time scale power failure features corresponding to the low voltage user, the long short-term memory neural network architecture corresponding to the medium layer is used to extract the different time scale power failure features corresponding to the distribution transformer, and the long short-term memory neural network architecture corresponding to the high layer is used to extract the different time scale power failure features corresponding to the medium voltage line, wherein the long short-term memory neural network architecture corresponding to each layer is used to capture transient power failure sub-features corresponding to voltage sag and inrush in short time scale, to extract event power failure sub-features corresponding to load mutation and equipment start in medium time scale, and to identify power failure state trend sub-features corresponding to planned power failure and equipment aging in long time scale;

[0015] Step S15: Based on the hierarchical long short-term memory neural network HLNN model, the power supply usage time series data corresponding to the medium voltage line, distribution transformer and low voltage user are analyzed for hierarchical multi-scale feature analysis, and different time scale power failure features corresponding to the line-transformer-user are obtained.

[0016] Further, step S15 includes the following steps:

[0017] Step S151: The power supply usage time series data corresponding to the medium voltage line, distribution transformer and low voltage user are respectively input into the corresponding low, medium and high layer long short-term memory neural network architecture in the hierarchical long short-term memory neural network HLNN model for dimension tensor reconstruction, so as to decompose the voltage, current and power physical quantities in the power supply usage time series data into three power failure feature factor matrices of time evolution mode, device response mode and basic fluctuation mode according to different scale time windows, device levels and measurement points, and obtain three power failure feature tensor factor matrices corresponding to the medium voltage line, distribution transformer and low voltage user;

[0018] Step S152: The three power failure feature tensor factor matrices corresponding to the medium voltage line, distribution transformer and low voltage user are mapped for power failure feature hierarchical mapping, so as to map the three power failure feature tensor factor matrices to a line-transformer-user three-level cube, wherein the X-axis of the cube represents the device level, the Y-axis represents the time scale, and the Z-axis represents the power failure feature type, and a line-transformer-user corresponding hierarchical power failure feature three-dimensional mapping cube is obtained;

[0019] Step S153: The matrix eigenvalue and eigenvector corresponding to the power failure feature cube are calculated through the line-transformer-user corresponding hierarchical power failure feature three-dimensional mapping cube, and the power failure feature significance of each level under different time scales is determined according to the matrix eigenvalue and eigenvector corresponding to the power failure feature cube;

[0020] Step S154: The self-attention mechanism corresponding to each level long short-term memory neural network architecture in the hierarchical long short-term memory neural network HLNN model is introduced, and the line-transformer-user corresponding hierarchical power failure feature three-dimensional mapping cube is analyzed for multi-scale feature fusion according to the power failure feature significance of each level under different time scales, so as to obtain different time scale power failure features corresponding to the line-transformer-user.

[0021] Further, step S2 includes the following steps:

[0022] Step S21: The power supply space topological relationship corresponding to the line-transformer-user is statistically analyzed based on the line-transformer-user corresponding hierarchical different time scale power failure features.

[0023] Step S22: generating the power supply topology state constraint condition between any two graph nodes by taking the corresponding medium-voltage line, distribution transformer and low-voltage user as corresponding graph nodes, and performing power topology change constraint analysis between any two graph nodes with a topological relationship based on the power supply space topological relationship of the line-transformer-user correspondence, so as to calculate the condition probability that the state change of one graph node causes the state change of another graph node, and take it as the constraint condition between the two graph nodes, and generate the power supply topology state constraint condition between any two graph nodes;

[0024] Step S23: constructing the topology connection between the graph nodes corresponding to the medium-voltage line, distribution transformer and low-voltage user based on the power supply topology state constraint condition between any two graph nodes and combining the power supply space topological relationship of the line-transformer-user correspondence, to generate the line-transformer-user power supply topology relationship graph model.

[0025] Further, step S21 includes the following steps:

[0026] Performing power outage fluctuation time delay analysis on the power outage characteristics of the line-transformer-user corresponding hierarchical different time scales to obtain the time sequence power outage fluctuation time delay between the line-transformer-user corresponding different level devices;

[0027] Obtaining the power outage characteristic phase difference between the line-transformer-user corresponding different level devices through the line-transformer-user corresponding hierarchical different time scale power outage characteristics, and performing correlation and causality evaluation according to the time sequence power outage fluctuation time delay and the power outage characteristic phase difference between the line-transformer-user corresponding different level devices, to obtain the correlation and causality influence strength between the line-transformer-user corresponding different level devices;

[0028] Obtaining the energy flow direction between the line-transformer-user corresponding different level devices through the line-transformer-user corresponding hierarchical different time scale power outage characteristics, and performing power supply topology relationship connection analysis between the line-transformer-user corresponding different level devices based on the energy flow direction between the line-transformer-user corresponding different level devices and combining the correlation and causality influence strength between the line-transformer-user corresponding different level devices, to take the line-transformer-user corresponding different level devices as power supply space nodes, and the edge weight between each power supply space node represents the correlation and causality influence strength, and the direction of the edge represents the energy flow direction, to obtain the power supply space topological relationship of the line-transformer-user correspondence.

[0029] Further, step S3 includes the following steps:

[0030] Step S31: obtaining the topological-energy coupling tensor between the graph nodes corresponding to the medium-voltage line, distribution transformer and low-voltage user according to the line-transformer-user power supply topology relationship graph model;

[0031] Step S32: Calculate the energy loss amount when the current flows from the medium-voltage line and the distribution transformer corresponding graph nodes to the low-voltage user corresponding graph node based on the topology-energy coupling tensor between the medium-voltage line, the distribution transformer and the low-voltage user corresponding graph nodes and combined with the corresponding Ohm's law and energy loss in the power system;

[0032] Step S33: Perform low-voltage power failure impact assessment on the low-voltage user corresponding graph node based on the energy loss amount when the current flows from the medium-voltage line and the distribution transformer corresponding graph nodes to the low-voltage user corresponding graph node, to obtain the line-transformer corresponding power failure impact range of the low-voltage user;

[0033] Step S34: Perform time household number stratification calculation according to the line-transformer corresponding power failure impact range of the low-voltage user, to obtain the power failure impact time household number corresponding to the distribution line, the distribution transformer and the meter.

[0034] Further, step S33 includes the following steps:

[0035] Obtain the graph node connection relationship between the medium-voltage line, the distribution transformer and the low-voltage user through the line-transformer-house power supply topology relationship graph model;

[0036] According to the energy loss amount when the current flows from the medium-voltage line and the distribution transformer corresponding graph nodes to the low-voltage user corresponding graph node, perform energy attenuation coefficient transformation to calculate the energy attenuation ratio value between the medium-voltage line and the distribution transformer corresponding graph nodes and the low-voltage user corresponding graph nodes, and obtain the energy attenuation coefficient between the medium-voltage line and the distribution transformer corresponding graph nodes and the low-voltage user corresponding graph nodes;

[0037] Based on the graph node connection relationship between the medium-voltage line, the distribution transformer and the low-voltage user and combined with the energy attenuation coefficient between the medium-voltage line and the distribution transformer corresponding graph nodes and the low-voltage user corresponding graph nodes, perform node energy attenuation matrix construction to generate the corresponding node energy attenuation coefficient matrix between the medium-voltage line, the distribution transformer and the low-voltage user;

[0038] When the graph node corresponding to the medium-voltage line or the distribution transformer is powered off, based on the corresponding node energy attenuation coefficient matrix between the medium-voltage line, the distribution transformer and the low-voltage user, simulate the corresponding propagation process of the power failure disturbance wave in the line-transformer-house power supply topology relationship graph model, and iteratively calculate the propagation intensity of the power failure disturbance wave between the graph nodes, and simultaneously according to the propagation intensity of the power failure disturbance wave between the graph nodes, use the analytic hierarchy process to calculate the user side disturbance sensitivity value of each low-voltage user corresponding graph node;

[0039] The low-voltage user corresponding graph node is evaluated for low-voltage power failure influence based on the user side disturbed sensitivity value of each low-voltage user corresponding graph node and in combination with the corresponding node energy attenuation coefficient matrix, to obtain a line-transformer-low-voltage user corresponding power failure influence range.

[0040] Further, step S34 includes the following steps:

[0041] Step S341: constructing a corresponding four-dimensional tensor space-time influence model according to the line-transformer-low-voltage user corresponding power failure influence range in the time dimension, the space dimension, the equipment level dimension, and the influence intensity dimension, wherein each model sub-voxel represents the power failure influence intensity between the line-transformer-low-voltage user equipment level under a certain space-time dimension;

[0042] Step S342: calculating the time delay between the power failure conduction from the distribution line node to the distribution transformer node and then to the low-voltage user node based on the line-transformer-low-voltage user power supply topology relationship graph model, and constructing a line-transformer-low-voltage user corresponding topology conduction time delay matrix, wherein each matrix element represents the time required for the power failure to propagate from one node to another node;

[0043] Step S343: calculating the energy entropy weight between each equipment level in the line-transformer-low-voltage user power supply topology relationship graph model based on the line-transformer-low-voltage user corresponding topology conduction time delay matrix, to obtain the energy entropy weight value of the line-transformer-low-voltage user different equipment level in the power failure, which represents the uncertainty of the equipment level in the corresponding propagation process of the power failure;

[0044] Step S344: performing time-house number integral calculation in the time, space, and equipment level dimensions based on the four-dimensional tensor space-time influence model, the line-transformer-low-voltage user corresponding topology conduction time delay matrix, and the energy entropy weight value of the line-transformer-low-voltage user different equipment level in the power failure, and using the Monte Carlo integral method, to follow the principle of upper level covering lower level and automatically exclude repeated statistical time-house numbers by setting logical judgment conditions, to obtain the power failure influence time-house number of the distribution line, the distribution transformer, and the meter.

[0045] Further, step S4 includes the following steps:

[0046] Step S41: performing node-level power failure probability analysis on the line-transformer-low-voltage user corresponding graph nodes in the line-transformer-low-voltage user power supply topology relationship graph model based on the power failure influence time-house number of the distribution line, the distribution transformer, and the meter, to obtain the line-transformer-low-voltage user corresponding node-level power failure probability;

[0047] Step S42: based on the outage influence household number corresponding to the distribution line, distribution transformer and meter, the average outage time index between the edges of the graph nodes existing in the topology relationship in the line-transformer-house supply topology relationship graph model is calculated, and the average outage time index of the line-transformer-house corresponding feeder level is obtained;

[0048] Step S43: based on the outage influence household number corresponding to the distribution line, distribution transformer and meter, the annual outage frequency distribution statistics of the entire line-transformer-house topology network in the line-transformer-house supply topology relationship graph model is performed, and the annual outage frequency distribution of the line-transformer-house corresponding network level is obtained;

[0049] Step S44: according to the outage failure probability of the line-transformer-house corresponding node level, the average outage time index of the line-transformer-house corresponding feeder level, and the annual outage frequency distribution of the line-transformer-house corresponding network level, the graph nodes corresponding to the medium and low voltage lines, distribution transformers and low voltage users in the line-transformer-house supply topology relationship graph model are evaluated for medium and low voltage power supply reliability, to obtain the medium and low voltage power supply reliability measurement corresponding to the line-transformer-house.

[0050] Further, the application also provides a medium and low voltage power supply reliability evaluation system based on the HLNN model, which is used to execute the medium and low voltage power supply reliability evaluation method based on the HLNN model as described above, and the medium and low voltage power supply reliability evaluation system based on the HLNN model comprises:

[0051] A multi-scale feature analysis module is configured to acquire power supply use time sequence data corresponding to the medium voltage line, distribution transformer and low voltage user, construct a hierarchical long short-term memory neural network (HLNN) model, and perform hierarchical multi-scale feature analysis on the power supply use time sequence data corresponding to the medium voltage line, distribution transformer and low voltage user based on the hierarchical long short-term memory neural network (HLNN) model, so as to obtain different time scale outage features corresponding to the line-transformer-house at different hierarchical levels.

[0052] A topology connection construction module is configured to analyze the power supply spatial topology relationship corresponding to the line-transformer-house based on the different time scale outage features corresponding to the line-transformer-house at different hierarchical levels, and construct the topology connection of the corresponding medium voltage line, distribution transformer and low voltage user as graph nodes according to the power supply spatial topology relationship corresponding to the line-transformer-house, so as to generate a line-transformer-house supply topology relationship graph model.

[0053] A household number hierarchical calculation module is configured to perform low voltage outage influence evaluation on the graph nodes corresponding to the low voltage users based on the line-transformer-house supply topology relationship graph model and in combination with the graph nodes corresponding to the medium voltage line and distribution transformer, so as to obtain the outage influence range of the low voltage users corresponding to the line-transformer-house; and perform household number hierarchical calculation according to the outage influence range of the low voltage users corresponding to the line-transformer-house, so as to obtain the outage influence household number corresponding to the distribution line, distribution transformer and meter.

[0054] The middle-low voltage reliability evaluation module is configured to evaluate the middle-low voltage power supply reliability of the graph nodes corresponding to the middle voltage line, the distribution transformer and the low voltage user in the line-transformer-house power supply topology graph model based on the outage influence household number corresponding to the distribution line, the distribution transformer and the meter, so as to obtain the middle-low voltage power supply reliability measurement corresponding to the line-transformer-house.

[0055] The present application has the following beneficial effects:

[0056] 1、The low-voltage power supply reliability evaluation method based on the HLNN model has the beneficial effects that, compared with the prior art, by acquiring the power supply time sequence data of the medium-voltage line, distribution transformer and low-voltage user, and constructing a hierarchical long short-term memory neural network (HLNN, Hierarchical LSTM Neural Network) model, the behavior characteristics of each power system component at different time scales can be effectively captured, the HLNN model has the advantage of capturing long-term dependencies in time series data, therefore, the power network in the complex dynamic changes of the medium-voltage line, distribution transformer and low-voltage user can be deeply understood, the hierarchical multi-scale feature analysis method can process features of different time scales, reveal the behavior differences and rules of the power system in the short term and the long term, help further identify potential problems in the power system, based on this multi-scale analysis, the characteristics of different power facilities and users in the time dimension can be analyzed more carefully, data support is provided for subsequent power dispatching and maintenance decision-making, thereby improving the power supply safety and stability of the medium and low voltage power system. Secondly, by converting the power failure features of different time scales into the power supply space topology relationship corresponding to the line, transformer and user, a more accurate and clear power network topology structure graph model can be established, this process can accurately model the relationship between the medium-voltage line, distribution transformer and low-voltage user, and can clearly determine the connection mode and mutual influence between different power facilities, for example, when a line fails, it can quickly identify which distribution transformers and low-voltage users are affected and predict the power failure impact range, this topology relationship graph model provides a dynamic, spatially related network view for the power dispatching center, which helps to accurately identify and locate potential fault points in the power network, at the same time, the graph model can intuitively reflect the influence and dependence relationship between each node in the power system, providing support for system health monitoring and optimization, ensuring the efficient operation of the power network, thereby improving the long-term operation efficiency of the medium and low voltage power system. Then, based on the line-transformer-user power supply topology relationship graph model, the power failure impact of the low-voltage user is evaluated, which can provide an accurate analysis framework based on graph nodes, this evaluation not only includes the impact prediction after the failure occurs, but also includes the calculation of the power interruption impact range under different load conditions, the key of this step is to quantify the specific impact of power failure on different regions and different users through the impact range analysis of the low-voltage user node, thereby providing accurate guidance for power restoration and fault repair, for example, through the hierarchical calculation of the number of households during power failure, it can be analyzed which regions need to be restored to power supply first and which regions can be self-recovered in a short time to minimize the impact of power failure and improve user satisfaction, and ensure the rapid response capability of the power system when encountering failures, thereby being able to more fully analyze the interaction of various complex factors in the power system.Finally, by calculating the number of households affected by a power outage, the reliability of power supply to distribution lines, distribution transformers, and meters can be effectively assessed. This assessment, combined with the power system topology diagram, provides a deeper understanding of the reliability performance of various components of the power network under fault conditions. Through this assessment method, the vulnerable links of different power facilities can be identified. For example, which distribution lines and transformers have a greater impact on users during power outages, which areas require more redundancy in the design phase, or whether equipment maintenance and monitoring need to be strengthened. The core value of this step is that it provides a forward-looking and safer optimization path for the power system, thereby improving the accuracy and efficiency of the overall medium and low voltage power supply reliability assessment and reducing the impact of power outage events on low-voltage users.

[0057] 2. The medium- and low-voltage power supply reliability assessment system based on the HLNN model proposed in this invention consists of a multi-scale feature analysis module, a topology connection construction module, a time-to-household hierarchical calculation module, and a medium- and low-voltage reliability assessment module. It can realize any medium- and low-voltage power supply reliability assessment method based on the HLNN model described in this invention. It is used to combine the operations between the computer programs running on each module to realize the medium- and low-voltage power supply reliability assessment method based on the HLNN model. The internal structure of the system cooperates with each other, which can greatly reduce repetitive work and manpower input, and can quickly and effectively provide a more accurate and efficient medium- and low-voltage power supply reliability assessment process based on the HLNN model, thereby simplifying the operation process of the medium- and low-voltage power supply reliability assessment system based on the HLNN model. Attached Figure Description

[0058] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0059] Figure 1 This is a schematic diagram of the steps in the medium and low voltage power supply reliability assessment method based on the HLNN model of the present invention.

[0060] Figure 2 for Figure 1 A detailed flowchart of step S1;

[0061] Figure 3 for Figure 2 A detailed flowchart of step S15. Detailed Implementation

[0062] The technical method of the present application will be described clearly and completely in combination with the drawings. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0063] To achieve the above-mentioned purpose, please refer to Figures 1 to 3 The present application provides a medium and low voltage power supply reliability evaluation method based on an HLNN model, which comprises the following steps:

[0064] Step S1: Obtain the power supply time sequence data corresponding to the medium voltage line, distribution transformer and low voltage user, and construct a hierarchical long short-term memory neural network (HLNN) model. Meanwhile, based on the hierarchical long short-term memory neural network (HLNN) model, the power supply time sequence data corresponding to the medium voltage line, distribution transformer and low voltage user is analyzed in a hierarchical multi-scale manner to obtain the line-transformer-user corresponding hierarchical different time scale power failure characteristics.

[0065] Step S2: Based on the line-transformer-user corresponding hierarchical different time scale power failure characteristic analysis, the power supply spatial topological relationship corresponding to the line-transformer-user is analyzed, and the corresponding medium voltage line, distribution transformer and low voltage user are connected as graph nodes to construct a line-transformer-user power supply topological relationship graph model.

[0066] Step S3: Based on the line-transformer-user power supply topological relationship graph model and in combination with the graph nodes corresponding to the medium voltage line and distribution transformer, the low voltage user corresponding graph node is evaluated for low voltage power failure influence to obtain the line-transformer corresponding power failure influence range of the low voltage user. According to the line-transformer corresponding power failure influence range of the low voltage user, the time and household number layer calculation is performed to obtain the power failure influence time and household number corresponding to the distribution line, distribution transformer and meter.

[0067] Step S4: Based on the power failure influence time and household number corresponding to the distribution line, distribution transformer and meter, the graph nodes corresponding to the medium voltage line, distribution transformer and low voltage user in the line-transformer-user power supply topological relationship graph model are evaluated for medium and low voltage power supply reliability to obtain the medium and low voltage power supply reliability measurement corresponding to the line-transformer-user.

[0068] In the embodiments of the present application, please refer to Figure 1 The step flowchart of the medium and low voltage power supply reliability evaluation method based on the HLNN model of the present application is shown in the figure. In this example, the medium and low voltage power supply reliability evaluation method based on the HLNN model comprises the following steps:

[0069] Step S1: Obtain power supply usage time series data corresponding to the medium voltage line, distribution transformer and low voltage user, and construct a hierarchical long short-term memory neural network HLNN model, and simultaneously perform hierarchical multi-scale feature analysis on the power supply usage time series data corresponding to the medium voltage line, distribution transformer and low voltage user based on the hierarchical long short-term memory neural network HLNN model to obtain line-transformer-user corresponding hierarchical different time scale power failure features;

[0070] In the embodiment of the application, by obtaining power supply usage time series data corresponding to the medium voltage line, distribution transformer and low voltage user, these data contain voltage, current, power and other information, and the time resolution is 15 minutes. Taking a medium voltage line as an example, its power supply usage time series data records the voltage value, current value and power value of every 15 minutes in the past year. A hierarchical long short-term memory neural network HLNN model is constructed, which includes a low, medium and high level long short-term memory neural network architecture. The low layer is used to extract low voltage user corresponding different time scale power failure features, the medium layer is used to extract distribution transformer corresponding different time scale power failure features, and the high layer is used to extract medium voltage line corresponding different time scale power failure features. The power supply usage time series data corresponding to the medium voltage line, distribution transformer and low voltage user is analyzed for hierarchical multi-scale features. Taking a low voltage user as an example, through the low layer long short-term memory neural network architecture, the transient power failure sub-features corresponding to voltage sag and inrush current are captured in a short time scale, the event power failure sub-features corresponding to load mutation and equipment start are extracted in a medium time scale, and the power failure state trend sub-features corresponding to planned power failure and equipment aging are identified in a long time scale. The same method is applied to the data of the distribution transformer and the medium voltage line. Finally, line-transformer-user corresponding hierarchical different time scale power failure features are obtained.

[0071] Step S2: Analyze the line-transformer-user corresponding power supply space topological relationship based on the line-transformer-user corresponding hierarchical different time scale power failure features, and topologically connect the corresponding medium voltage line, distribution transformer and low voltage user as graph nodes to construct a line-transformer-user power supply topological relationship graph model;

[0072] In the embodiment of the present application, by analyzing the power supply space topology relationship of the line-transformer-user correspondence based on the different time scale power failure characteristics of the line-transformer-user correspondence hierarchy, the connection relationship and influence path between the medium voltage line, distribution transformer and low voltage user can be determined through the analysis of the power failure characteristics, so as to construct the topology connection according to the power supply space topology relationship of the line-transformer-user correspondence, taking the medium voltage line L1 as an example, it is connected with the distribution transformers T1 and T2, T1 is connected with the low voltage users U1, U2 and U3, and T2 is connected with the low voltage users U4, U5 and U6, in this way, the line-transformer-user power supply topology relationship graph model is constructed.

[0073] Step S3: Based on the line-transformer-user power supply topology relationship graph model and combined with the graph nodes corresponding to the medium voltage line and distribution transformer, the low voltage user corresponding graph node is evaluated for low voltage power failure influence, to obtain the power failure influence range of the line-transformer-user corresponding low voltage user; according to the power failure influence range of the line-transformer-user corresponding low voltage user, the time household number layering calculation is performed to obtain the power failure influence time household number corresponding to the distribution line, distribution transformer and meter;

[0074] In the embodiment of the present application, based on the line-transformer-user power supply topology relationship graph model, combined with the graph nodes corresponding to the medium voltage line and distribution transformer, the low voltage user corresponding graph node is evaluated for low voltage power failure influence, when the medium voltage line L1 fails, the affected distribution transformers T1 and T2 and low voltage users U1-U6 can be determined through the topology relationship graph, according to the power failure influence range of the line-transformer-user corresponding low voltage user, the time household number layering calculation is performed, taking the distribution line L1 as an example, its power failure influence range includes T1 and T2 and the connected low voltage users, the power failure time sum of these affected users is calculated to obtain the power failure influence time household number of the distribution line L1, the same method is applied to the distribution transformer and meter, and finally the power failure influence time household number corresponding to the distribution line, distribution transformer and meter is obtained.

[0075] Step S4: Based on the power failure influence time household number corresponding to the distribution line, distribution transformer and meter, the graph nodes corresponding to the medium voltage line, distribution transformer and low voltage user in the line-transformer-user power supply topology relationship graph model are evaluated for medium and low voltage power supply reliability, to obtain the medium and low voltage power supply reliability measurement corresponding to the line-transformer-user.

[0076] In the embodiment of the present application, based on the power failure influence time household number corresponding to the distribution line, distribution transformer and meter, the graph nodes corresponding to the medium voltage line, distribution transformer and low voltage user in the line-transformer-user power supply topology relationship graph model are evaluated for medium and low voltage power supply reliability, the comprehensive scoring method is adopted, the node level power failure fault probability weight =0.4, the feeder level average power failure time index weight = 0.3, network level annual outage frequency distribution weight = 0.3, reliability measure calculation formula , wherein is the node level outage failure probability, is the feeder level average outage time index, is the network level high outage frequency corresponding frequency, taking the medium voltage line L1 as an example, it is known that = 0.005, the = 0.001 of a certain edge related to L1, = 0.1, substituting the formula can obtain = 0.9677, in the same way, the graph nodes of all medium voltage lines, distribution transformers and low voltage users in the power supply topology relationship graph model are calculated, and finally the medium and low voltage power supply reliability measure value corresponding to the line-transformer-user is obtained. The higher the value, the stronger the power supply reliability, so as to comprehensively evaluate the reliability level of the medium and low voltage power supply system.

[0077] Further, as an embodiment of the present application, referring to Fig. 1, it is a detailed step flow diagram of step S1 in the embodiment, and step S1 in the embodiment includes the following steps: Figure 2 Figure 1

[0078] Step S11: acquiring power supply use time sequence data corresponding to the medium voltage line;

[0079] In the embodiment of the present application, through the SCADA (data acquisition and monitoring control system) of the power system, the power supply use time sequence data corresponding to the medium voltage line is acquired with a sampling interval of 1 minute. The specific collection contents include line three-phase voltage value (unit: kilovolt), three-phase current value (unit: ampere), active power (unit: megawatt), reactive power (unit: megavar) and switch state (1 represents closed and 0 represents open) and other information. For example, for a medium voltage line numbered ZYL-001, 1440 groups of data are acquired in a continuous 24-hour collection process, wherein at the 300th sampling time, the A-phase voltage is 10.2 kilovolts, the B-phase voltage is 10.1 kilovolts, the C-phase voltage is 10.3 kilovolts, the A-phase current is 80 amperes, the B-phase current is 82 amperes, the C-phase current is 78 amperes, the active power is 1.2 megawatts, the reactive power is 0.5 megavar, and the switch state is 1. These data completely record the power supply operation state of the medium voltage line at different time, providing basic data support for subsequent analysis.

[0080] Step S12: acquiring power supply use time sequence data corresponding to the distribution transformer;

[0081] ​​In the embodiment of the present application, by using the intelligent monitoring terminal installed on the distribution transformer, the corresponding power supply use time series data of the distribution transformer is obtained with a sampling interval of 1 minute, and the collected parameters cover the input side three-phase voltage, current, output side three-phase voltage, current, winding temperature (unit: Celsius), load rate and other information of the transformer. Taking one distribution transformer as an example, 1440 groups of data are obtained in one day of monitoring. For example, at the 600th sampling moment, the input side A-phase voltage is 10 kV, the B-phase voltage is 9.9 kV, the C-phase voltage is 10.1 kV, the input side A-phase current is 30 A, the B-phase current is 32 A, the C-phase current is 28 A, the output side A-phase voltage is 0.4 kV, the B-phase voltage is 0.39 kV, the C-phase voltage is 0.41 kV, the output side A-phase current is 120 A, the B-phase current is 125 A, the C-phase current is 115 A, the winding temperature is 55 degrees Celsius, and the load rate is 60%. These data reflect the electrical parameters and working state changes of the distribution transformer during operation.

[0082] Step S13: obtaining the power supply use time series data corresponding to the low-voltage user;

[0083] In the embodiment of the present application, by installing an intelligent electric meter on the low-voltage user side, the corresponding power supply use time series data of the low-voltage user is obtained with a sampling interval of 15 minutes, and the collected information includes the user's real-time active power (unit: kilowatt), reactive power (unit: kilovolt-ampere), power consumption (unit: kilowatt-hour) and the like. For example, for a certain residential user, 2880 groups of data are obtained in a monitoring period of one month. In a certain 15-minute period, the active power of the user is 3 kilowatts, the reactive power is 1 kilovolt-ampere, and the power consumption is 0.75 kilowatt-hours. Through the collection of a large amount of low-voltage user data, the user's power consumption behavior and load characteristics can be comprehensively understood, and a basis for analyzing the low-voltage power supply situation is provided.

[0084] Step S14: constructing a hierarchical long short-term memory neural network HLNN model, which includes the long short-term memory neural network architecture corresponding to the low, medium and high layers. The long short-term memory neural network architecture corresponding to the low layer is used to extract different time scale power failure features corresponding to the low-voltage user, the long short-term memory neural network architecture corresponding to the medium layer is used to extract different time scale power failure features corresponding to the distribution transformer, and the long short-term memory neural network architecture corresponding to the high layer is used to extract different time scale power failure features corresponding to the medium-voltage line. The long short-term memory neural network architecture corresponding to each layer is used to capture transient power failure sub-features corresponding to voltage sag and inrush current at a short time scale, to extract event power failure sub-features corresponding to load mutation and equipment start-up at a medium time scale, and to identify power failure state trend sub-features corresponding to planned power failure and equipment aging at a long time scale;

[0085] In the embodiment of the present application, by constructing a hierarchical long short-term memory neural network HLNN model, the low-layer long short-term memory neural network architecture is set to contain 2 hidden layers, each hidden layer has 64 neurons, the input layer receives the standardized values of the low-voltage user power supply usage time series data, the standardization formula is , wherein is the original data, is the data after standardization, this layer captures the transient characteristics corresponding to voltage sag (such as the case where the voltage value is less than 80% of the rated value), inrush (current instantaneously increases greatly) under short time scale (15 minutes-1 hour) through memory cells and gating mechanism; under medium time scale (1-12 hours), event outage characteristics corresponding to load mutation (such as the sudden increase in power caused by the start of high-power electrical appliances of users), equipment start-up (such as the power change when the elevator in the cell starts) are extracted; under long time scale (12 hours-1 day), power trend sub-features such as power trend changes caused by user work and rest regularity are identified, the middle-layer long short-term memory neural network architecture is also set to contain 2 hidden layers, each hidden layer has 80 neurons, receives the standardized values of the power distribution transformer power supply usage time series data, under the corresponding time scale, transformer-related power outage features are extracted, the high-layer long short-term memory neural network architecture is set to contain 2 hidden layers, each hidden layer has 100 neurons, processes the standardized values of the medium-voltage line power supply usage time series data, and completes the feature extraction under the corresponding time scale, the data is transmitted between layers through the connection layer, and the complete HLNN model architecture is constructed.

[0086] Step S15: based on the hierarchical long short-term memory neural network HLNN model, the power supply usage time series data corresponding to the medium-voltage line, the power distribution transformer and the low-voltage user are analyzed in a hierarchical multi-scale manner to obtain the hierarchical different time scale power outage features corresponding to the line-transformer-user.

[0087] In the embodiment of the present application, by inputting the acquired power supply use time sequence data corresponding to the medium voltage line, distribution transformer and low voltage user into the hierarchical long short-term memory neural network HLNN model after preprocessing according to the standardized formula, the low layer network processes the low voltage user data. For example, when analyzing the user data of a certain community, it is identified that at 20 o'clock on a certain day, due to the start of the air conditioner of a certain household, there is a short-time voltage drop (from 0.22 kV to 0.18 kV), and the transient power failure characteristic is captured. In the analysis of the data of a continuous week, it is found that there is a load mutation at 7-8 o'clock every morning, and the power failure characteristic of the event is extracted. At the same time, it is observed that with the passage of time, the power consumption of the community has a gradually increasing trend, and the power failure state trend sub-feature is identified. The middle layer network analyzes the distribution transformer data, such as detecting that the winding temperature of a certain transformer rises from 50 degrees Celsius to 65 degrees Celsius in a short time when a device starts, and extracting the power failure characteristic of the related event. Long-term observation finds that the load rate gradually increases, and the power failure state trend sub-feature caused by equipment aging is identified. The high layer network processes the medium voltage line data, finds the transient power failure characteristic corresponding to the inrush phenomenon of the line due to switch operation at a certain time period, and the trend sub-feature of long-time power failure due to planned maintenance, and finally, through the hierarchical processing of the HLNN model, the different time scale power failure characteristics corresponding to the line-transformer-user are obtained, which provides key data support for the reliability evaluation of medium and low voltage power supply.

[0088] Further, as an embodiment of the present application, referring to Figure 3 , the detailed step flowchart of step S15 is shown in Figure 2 , which includes the following steps:

[0089] Step S151: input the power supply use time sequence data corresponding to the medium voltage line, distribution transformer and low voltage user into the corresponding low, middle and high layer long short-term memory neural network architecture in the hierarchical long short-term memory neural network HLNN model respectively, and perform dimension tensor reconstruction, so as to decompose the voltage, current and power physical quantities in the power supply use time sequence data into three power failure characteristic factor matrices of time evolution mode, device response mode and basic fluctuation mode according to different scale time windows, device levels and measurement points, and obtain three power failure characteristic tensor factor matrices corresponding to the medium voltage line, distribution transformer and low voltage user;

[0090] In the embodiment of the present application, by inputting the acquired and standardized power supply use time sequence data corresponding to the medium voltage line, distribution transformer and low voltage user into the corresponding high, medium and low layer long short term memory neural network architecture in the hierarchical long short term memory neural network HLNN model, taking the medium voltage line data as an example, assuming that 24 hours of continuous, 1 group per minute of three-phase voltage (unit: kilovolts), three-phase current (unit: amperes), active power (unit: megawatts) and reactive power (unit: megavars) data are collected, a total of 1440 groups, when performing dimension tensor reconstruction, set the short time scale time window to 15 minutes (i.e. 15 groups of data), the medium time scale time window to 1 hour (60 groups of data), and the long time scale time window to 12 hours (720 groups of data), according to the device level (medium voltage line) and the measurement point (different measurement quantities such as three-phase voltage, current and power), the data is decomposed, taking the A-phase voltage as an example, in the short time scale 15-minute window, the data is constructed into a time evolution pattern matrix by means of a sliding window, each row represents the change of the A-phase voltage with time within 15 minutes; the device response pattern matrix arranges the data relationship of the A-phase voltage and other measurement quantities (such as A-phase current) under the same time window from the perspective of different measurement points; the basic fluctuation pattern matrix extracts statistical characteristics such as the mean and variance of the A-phase voltage within the 15 minutes. Similarly, the same operation is performed on the distribution transformer and low voltage user data, such as 1440 groups of data of a distribution transformer, which are decomposed into three power failure characteristic factor matrices under different time windows, and finally three power failure characteristic tensor factor matrices corresponding to the medium voltage line, distribution transformer and low voltage user are obtained, completing the conversion of the data from the original time sequence form to the feature matrix form.

[0091] Step S152: performing power failure feature hierarchical mapping on the three power failure feature tensor factor matrices corresponding to the medium voltage line, distribution transformer and low voltage user, to map the three power failure feature tensor factor matrices to a line-transform-user three-level cubic body, wherein the X-axis of the cubic body represents the device level, the Y-axis represents the time scale, and the Z-axis represents the power failure feature type, to obtain a power failure feature three-dimensional mapping cubic body corresponding to the hierarchical line-transform-user;

[0092] In the embodiment of the present application, by performing power failure feature hierarchical mapping on three power failure feature tensor factor matrices corresponding to the medium voltage line, distribution transformer and low voltage user, the three power failure feature tensor factor matrices of the medium voltage line are mapped into a three-dimensional space, wherein the X axis represents the device level and takes the value of the medium voltage line; the Y axis represents the time scale and corresponds to the short time scale, the medium time scale and the long time scale respectively; and the Z axis represents the power failure feature type and corresponds to the time evolution mode, the device response mode and the basic fluctuation mode. For example, each element in the time evolution mode matrix of the medium voltage line under the short time scale corresponds to a point in the three-dimensional space according to its position in the matrix, and a plane is formed after mapping all the elements. Similarly, the power failure feature tensor factor matrices of the distribution transformer and the low voltage user are mapped to form their respective three-dimensional space structures, and finally, the corresponding hierarchical power failure feature three-dimensional mapping cube of the line-transform-user is obtained, which clearly shows the data relationship between different device levels, time scales and power failure feature types. For example, the feature distribution of the device response mode of a certain low voltage user under the medium time scale can be directly observed in the cube.

[0093] Step S153: The matrix eigenvalues and eigenvectors corresponding to the power failure feature cube are calculated through the line-transform-user corresponding hierarchical power failure feature three-dimensional mapping cube, and the power failure feature significance of each level under different time scales is determined according to the matrix eigenvalues and eigenvectors corresponding to the power failure feature cube.

[0094] In the embodiment of the present application, by regarding the line-transform-user corresponding hierarchical power failure feature three-dimensional mapping cube as a high-order matrix, the singular value decomposition (SVD) method is used to calculate the matrix eigenvalues and eigenvectors. Taking the power failure feature three-dimensional mapping cube of the medium voltage line as an example, it is converted into a matrix form , and the SVD decomposition is performed , wherein and are orthogonal matrices, is a diagonal matrix, and the elements on the diagonal are eigenvalues . According to the size of the eigenvalues, the power failure feature significance of each level under different time scales is determined. The larger the eigenvalue is, the higher the importance of the corresponding feature in describing the power failure feature is. For example, if the eigenvalue = 15 related to the time evolution mode under the long time scale is greater than other eigenvalues, it means that the feature is more significant in the power failure feature description under the long time scale. The power failure feature three-dimensional mapping cubes of the distribution transformer and the low voltage user are calculated in the same way to determine the significance of different power failure feature types of each level under different time scales, which provides a quantitative basis for subsequent feature fusion.

[0095] Step S154: introducing a corresponding self-attention mechanism between each hierarchical long short-term memory neural network architecture in the hierarchical long short-term memory neural network HLNN model, and combining the power failure feature significance of each layer at different time scales to perform multi-scale feature fusion analysis on the power failure feature three-dimensional mapping cube of the line-transform-customer corresponding layer, to obtain the power failure features of the line-transform-customer corresponding layer at different time scales.

[0096] In the embodiment of the application, by introducing a self-attention mechanism between each hierarchical long short-term memory neural network architecture in the hierarchical long short-term memory neural network HLNN model, taking the three-layer architecture of the medium-voltage line, the distribution transformer and the low-voltage user as an example, for the power failure features output by the high-layer architecture of the medium-voltage line, the attention weights between the output features of the medium-voltage line high-layer architecture and the output features of the distribution transformer middle-layer architecture and the low-voltage user low-layer architecture are calculated, the output feature vector of the medium-voltage line high-layer architecture is denoted as , the output feature vector of the distribution transformer middle-layer architecture is denoted as , and the output feature vector of the low-voltage user low-layer architecture is denoted as , and the attention weight calculation formula is , wherein Dot product is used for calculation, that is, and by combining the determined power failure feature significance of each layer at different time scales, the attention weights are adjusted, if the power failure feature significance of a medium-voltage line at a long time scale is high, the weight of the medium-voltage line when fused with the features of other layers is increased, and by weighted summation, the power failure features of different layers, different time scales and different feature types are fused, such as , wherein denotes the power failure feature vector of a device layer , a time scale , and a feature type , and finally the power failure features of the line-transform-customer corresponding layer at different time scales are obtained, which fuse multiple aspects of information and more comprehensively reflect the power failure of the medium-low voltage power supply system.

[0097] Further, step S2 includes the following steps:

[0098] Step S21: statistically analyzing the power supply space topological relationship of the line-transform-customer based on the power failure features of the line-transform-customer corresponding layer at different time scales;

[0099] In the embodiment of the present application, by performing statistical analysis on the power supply space topology relationship based on the previously obtained line-variable-user corresponding hierarchical different time scale power failure characteristics, taking a low-voltage power supply system in a certain area as an example, assuming that the area has 3 medium-voltage lines (L1, L2, L3), 5 distribution transformers (T1-T5) and 10 low-voltage users (U1-U10), for the medium-voltage line, by analyzing the change rule of its power failure characteristics at different time scales and the association with other devices, it is found that L1 supplies power to T1, T2 and T3, L2 supplies power to T4 and T5, and L3 is not directly associated with the currently statistical distribution transformer, for the distribution transformer, T1 supplies power to U1-U3, T2 supplies power to U4-U6, T3 supplies power to U7-U8, T4 supplies power to U9, and T5 supplies power to U10, in this way, the connection relationship between each device is clear, and the power supply space topology relationship is constructed, for example, taking the medium-voltage line as the starting point, the connection line between the medium-voltage line and the distribution transformer is drawn, and then the distribution transformer is connected to the corresponding low-voltage user, forming a tree-shaped topology relationship diagram, which clearly shows the power supply connection between the line-variable-user.

[0100] Step S22: by taking the corresponding medium-voltage line, distribution transformer and low-voltage user as corresponding graph nodes, and based on the power supply space topology relationship corresponding to the line-variable-user, performing power topology change constraint analysis between any two graph nodes corresponding to the topology relationship to calculate the conditional probability that the state change of one of the graph nodes causes the state change of the other graph node, and taking it as the constraint condition between the two graph nodes, to generate the power supply topology state constraint condition between any two graph nodes;

[0101] In the embodiment of the present application, by taking the medium-voltage line, distribution transformer and low-voltage user as graph nodes, taking L1, T1 and U1 as an example, based on the previously determined power supply space topology relationship, performing power topology change constraint analysis, assuming that the state change (such as power failure) of L1 will affect T1, and the state change of T1 will affect U1, calculating the conditional probability that the power failure of L1 causes the power failure of T1, through historical data statistics, in the past 100 L1 power failure events, the number of times that T1 follows the power failure is 80 times, then the conditional probability that the power failure of L1 causes the power failure of T1 =80 / 100=0.8, similarly, calculating the conditional probability that the power failure of T1 causes the power failure of U1, assuming that in the past 50 T1 power failure events, the number of times that U1 follows the power failure is 40 times, then =40 / 50=0.8, taking these conditional probabilities as the constraint conditions between L1 and T1, T1 and U1, and so on, calculating the conditional probability between any two graph nodes with topology relationship, such as the conditional probability between L2 and T4, T4 and U9, and finally generating the power supply topology state constraint condition between all graph nodes, providing a basis for subsequent topology connection construction.

[0102] Step S23: based on the power supply topology state constraint condition between any two graph nodes and combined with the power supply space topology relationship corresponding to the line-transformer-user, the topology connection between the graph nodes corresponding to the medium voltage line, distribution transformer and low voltage user is constructed to generate a line-transformer-user power supply topology relationship graph model.

[0103] In the embodiment of the application, by constructing the topology connection between the graph nodes corresponding to the medium voltage line, distribution transformer and low voltage user based on the previously obtained power supply topology state constraint condition between any two graph nodes and combined with the determined power supply space topology relationship corresponding to the line-transformer-user, taking L1, T1 and U1 as an example, since the conditional probability between L1 and T1 is 0.8 and the conditional probability between T1 and U1 is 0.8, when constructing the topology connection, a directed edge is used to represent the relationship, from L1 to T1 and marked with the conditional probability 0.8, from T1 to U1, also marked with the conditional probability 0.8, according to this method, the connection of all graph nodes is constructed, for example, the conditional probability between L2 and T4 is 0.75 and the conditional probability between T4 and U9 is 0.9, then from L2 to T4 is marked with 0.75 and from T4 to U9 is marked with 0.9, finally a line-transformer-user power supply topology relationship graph model is generated, which clearly shows the topology connection between the devices in the medium and low voltage power supply system and the constraint condition of the state change, and provides an intuitive visual model and analysis basis for the medium and low voltage power supply reliability evaluation.

[0104] Further, step S21 includes the following steps:

[0105] The power outage fluctuation time delay analysis is performed on the power outage characteristics of different time scales corresponding to the line-transformer-user hierarchical layers to obtain the time sequence power outage fluctuation time delay between the devices corresponding to the line-transformer-user at different levels;

[0106] In the embodiment of the present application, by analyzing the power failure fluctuation time delay through the power failure characteristics of different time scales corresponding to the line-transform-customer hierarchical layer, taking a certain power supply area as an example, which contains a medium voltage line L1, a distribution transformer T1 and three low voltage users U1, U2 and U3, from the obtained hierarchical multi-scale power failure characteristic data, the power failure data at a short time scale (15 minutes) is selected for analysis. It is assumed that at a certain time, the medium voltage line L1 has a voltage drop due to a fault, resulting in a change in the power failure characteristic. Ten minutes later, the distribution transformer T1 detects an abnormal input voltage, and its power failure characteristic changes accordingly. Three minutes later, the supply voltage of the low voltage user U1 is lower than the normal range, and the power failure characteristic fluctuates. By comparing the time points of the power failure characteristic changes of the devices at each level, the time sequence power failure fluctuation time delay between the medium voltage line L1 and the distribution transformer T1 is calculated as 10 minutes, and the time sequence power failure fluctuation time delay between the distribution transformer T1 and the low voltage user U1 is calculated as 3 minutes. Using the same method, the power failure characteristic data between the medium voltage line L1 and other low voltage users, and between the distribution transformer T1 and other low voltage users are analyzed to obtain all the time sequence power failure fluctuation time delays between different level devices at a short time scale. For the medium time scale (1 hour) and the long time scale (12 hours), the above operation is repeated to obtain all the time sequence power failure fluctuation time delay data at different time scales, forming a time sequence power failure fluctuation time delay set between different level devices corresponding to the line-transform-customer.

[0107] Preferably, the power failure characteristic phase difference between different level devices corresponding to the line-transform-customer is obtained through the power failure characteristics of different time scales corresponding to the line-transform-customer hierarchical layer, and the correlation and causality evaluation is performed according to the time sequence power failure fluctuation time delay and the power failure characteristic phase difference between different level devices corresponding to the line-transform-customer to obtain the correlation and causality influence strength between different level devices corresponding to the line-transform-customer.

[0108] In the embodiment of the present application, the power failure characteristic phase difference between different level devices is obtained through the power failure characteristics of different time scales corresponding to the line-transform-customer hierarchical layer. Taking the data of the medium voltage line L1 and the distribution transformer T1 at a medium time scale as an example, the power failure characteristic data of the two is converted into a periodic signal (such as converting the data of voltage, current and the like changing with time into a frequency domain signal through Fourier transform). It is assumed that the power failure characteristic signal of the medium voltage line L1 is , the power failure characteristic signal of the distribution transformer T1 is , and the power failure characteristic phase difference of the two is . It is assumed that the power failure characteristic phase difference of the medium voltage line L1 and the distribution transformer T1 at a certain medium time scale is calculated as = 30°. Combined with the previously obtained time sequence power failure fluctuation time delay (set as 15 minutes) between the two, the correlation and causality evaluation is performed, and the causality strength calculation formula is used wherein is the time-varying power failure fluctuation time delay), the correlation and causality influence strength between the medium-voltage line L1 and the distribution transformer T1 is calculated According to the method, the correlation and causality influence strength matrix between the line, the transformer and the user corresponding to the different hierarchical devices at different time scales is obtained by calculating the phase difference of the power failure characteristics and the time-varying power failure fluctuation time delay between the different hierarchical devices at different time scales, which clearly reflects the strength of the correlation and causality between the devices.

[0109] Preferably, the energy flow direction between the line, the transformer and the user corresponding to the different hierarchical devices is obtained from the different time-scale power failure characteristics of the line, the transformer and the user corresponding to the hierarchical devices, and the power supply topology connection analysis is performed based on the energy flow direction between the line, the transformer and the user corresponding to the different hierarchical devices and in combination with the correlation and causality influence strength between the line, the transformer and the user corresponding to the different hierarchical devices, so as to take the line, the transformer and the user corresponding to the different hierarchical devices as power supply space nodes, the edge weight between the nodes represents the correlation and causality influence strength, and the direction of the edge represents the energy flow direction, so as to obtain the power supply space topology relationship of the line, the transformer and the user.

[0110] In the embodiment of the present application, the energy flow direction between the different hierarchical devices is obtained from the different time-scale power failure characteristics of the line, the transformer and the user corresponding to the hierarchical devices. Under normal power supply conditions, the energy flows from the high-voltage side to the low-voltage side, i.e. from the medium-voltage line to the distribution transformer, and then from the distribution transformer to the low-voltage user. For example, the medium-voltage line L1 supplies power to the distribution transformer T1, and the energy flow direction is from L1 to T1; the distribution transformer T1 supplies power to the low-voltage users U1, U2 and U3, and the energy flow direction is from T1 to U1, U2 and U3, respectively. The power supply topology connection analysis is performed based on the previously obtained correlation and causality influence strength between the different hierarchical devices and in combination with the energy flow direction, the medium-voltage line L1, the distribution transformer T1 and the low-voltage users U1, U2 and U3 are taken as power supply space nodes, a directed edge is drawn between the nodes L1 and T1, the direction of the edge is from L1 to T1, representing the energy flow direction, and the edge weight is set as the correlation and causality influence strength between the two (e.g. 2° / min); a directed edge is drawn between the nodes T1 and U1, the direction of the edge is from T1 to U1, and the edge weight is the correlation and causality influence strength between T1 and U1 (assuming that the calculation result is 1.5° / min). In the same way, the connection between all the different hierarchical devices is performed, and the complete power supply space topology relationship diagram of the line, the transformer and the user is constructed, which takes the nodes to represent the devices, takes the edges with weights and directions to represent the energy flow relationship and the causality influence strength between the devices, and intuitively presents the mutual relationship between the devices in the medium-low voltage power supply system, thereby providing an accurate model basis for subsequent power supply reliability evaluation based on the topology relationship.

[0111] Further, step S3 comprises the following steps:

[0112] Step S31: obtaining the topological-energy coupling tensor between the corresponding graph nodes of the medium-voltage line, distribution transformer and low-voltage user according to the line-transform-user power supply topological relationship graph model;

[0113] In the embodiment of the present application, taking a certain regional power supply system as an example, the system contains 2 medium-voltage lines (L1, L2), 3 distribution transformers (T1, T2, T3) and 10 low-voltage users (U1-U10), for each node connection in the graph, determine its topological-energy coupling parameters, for example, the connection between the medium-voltage line L1 and the distribution transformer T1, obtain the resistance of the connection line =0.5 ohm, the reactance =0.3 ohm, and the active power =200 kilowatts, the reactive power =100 kilovars transmitted by the connection in unit time, combine these parameters into a three-dimensional tensor element, in which the three dimensions correspond to the starting node (medium-voltage line L1), the terminal node (distribution transformer T1) and the energy parameter type (resistance, reactance, active power, reactive power, etc.) of the topological connection respectively, according to this way, parameters are obtained and tensor elements are constructed for all node pairs with connection relationship in the graph, for example, the resistance =0.1 ohm, the reactance =0.05 ohm, the active power =20 kilowatts, the reactive power =10 kilovars of the connection between the distribution transformer T1 and the low-voltage user U1, finally, the complete topological-energy coupling tensor between the corresponding graph nodes of the medium-voltage line, distribution transformer and low-voltage user is constructed, which comprehensively describes the topological structure and energy transmission relationship between the nodes in the power supply system.

[0114] Step S32: based on the topological-energy coupling tensor between the corresponding graph nodes of the medium-voltage line, distribution transformer and low-voltage user and combining the Ohm's law and energy loss calculation in the power system, calculate the energy loss amount when the current flows through the corresponding graph nodes of the low-voltage user from the corresponding graph nodes of the medium-voltage line and distribution transformer;

[0115] In the embodiment of the present application, based on the previously obtained topological-energy coupling tensor, combining the Ohm's law and energy loss calculation formula in the power system, the energy loss amount when the current flows through the corresponding graph nodes of the low-voltage user from the corresponding graph nodes of the medium-voltage line and distribution transformer is calculated, taking the power supply of the low-voltage user U1 by the medium-voltage line L1 through the distribution transformer T1 as an example, the resistance of the L1-T1 line is known = 0.5 ohm, T1-U1 line resistance = 0.1 ohm, according to Ohm's law (hypothetical L1-T1 segment voltage = 10 kV, power factor = 0.9, active power = 200 kW), the current ≈ 22.2 A, the energy loss calculation formula is , then the energy loss of the L1-T1 segment line = 22.2 2 × 0.5 = 246.42 W, the energy loss of the T1-U1 segment line = 22.2 2 × 0.1 = 49.284 W, the total energy loss of the path from L1 to U1 through T1 is = 246.42 + 49.284 = 295.704 W, for all paths from the medium voltage line and the distribution transformer node to the low voltage user node, the energy loss is calculated according to the above method, and finally the energy loss data set of each power supply path in the entire power supply system is obtained.

[0116] Step S33: Perform low-voltage power failure impact assessment on the low-voltage user corresponding graph node based on the energy loss of the current flowing from the medium voltage line and the distribution transformer corresponding graph node to the low-voltage user corresponding graph node, to obtain the low-voltage user corresponding power failure impact range of the line-transformer;

[0117] In the embodiment of the application, the low-voltage user corresponding graph node is assessed for low-voltage power failure impact based on the previously calculated energy loss of the current flowing from the medium voltage line and the distribution transformer corresponding graph node to the low-voltage user corresponding graph node, an energy loss threshold is set, and it is assumed that when the energy loss of the power supply path of a certain low-voltage user node exceeds 30% of the power required for normal operation of the node, the node is considered to be affected by power failure. Taking the low-voltage user U1 as an example, its normal operation power is = 20 kW, if the total energy loss of the path of the node = 295.704 W, exceeds 20 × 10 3X30% = 6000 watts threshold, then determine that U1 is affected by power failure, by judging all 10 low-voltage user nodes one by one, determine the affected low-voltage user set. Further analyze the topological relationship between the affected low-voltage users, if multiple low-voltage users are powered by the same distribution transformer and are affected, then the power supply area corresponding to the distribution transformer is the power failure influence range; if the low-voltage users of multiple distribution transformers are affected and these distribution transformers are powered by the same medium-voltage line, then the power supply area corresponding to the medium-voltage line is the power failure influence range, so as to obtain the power failure influence range corresponding to the line-transformer-low-voltage user, such as determining that the medium-voltage line L1 and the low-voltage users U1-U6 connected by the distribution transformers T1 and T2 powered by the medium-voltage line L1 are a power failure influence range.

[0118] Step S34: According to the power failure influence range corresponding to the line-transformer-low-voltage user, the time household number is calculated, and the power failure influence time household number corresponding to the distribution line, distribution transformer and meter is obtained.

[0119] In the embodiment of the application, the time household number is calculated according to the previously obtained power failure influence range corresponding to the line-transformer-low-voltage user, taking the power failure influence range corresponding to the medium-voltage line L1 as an example, the range includes the distribution transformers T1 and T2 and the low-voltage users U1-U6, for the distribution line (i.e. the medium-voltage line L1), the total duration of power failure influence is counted, assuming that L1 causes power failure in the range due to failure, the power failure duration = 2 hours, then the power failure influence time household number of the distribution line L1 is the number of low-voltage users in the range multiplied by the power failure duration, i.e. 6x2 = 12 time household numbers, for the distribution transformer T1, the low-voltage users powered by T1 are U1-U3, if T1 fails to cause U1-U3 to be powered off, the power failure duration = 1.5 hours, then the power failure influence time household number of T1 is 3x1.5 = 4.5 time household numbers, for the meter (corresponding to each low-voltage user), taking U1 as an example, if the power failure duration of U1 is = 2 hours, then the power failure influence time household number of the meter corresponding to U1 is 1x2 = 2 time household numbers, according to this method, the time household number is calculated for all distribution lines, distribution transformers and meters in the respective power failure influence ranges, and the complete power failure influence time household number data corresponding to the distribution lines, distribution transformers and meters is obtained, which provides a quantitative index for medium and low voltage power supply reliability evaluation, and facilitates the analysis of weak links and reliability level of the power supply system.

[0120] Further, step S33 includes the following steps:

[0121] Obtain the graph node connection relationship between the medium-voltage line, distribution transformer and low-voltage user through the line-transformer-household power supply topological relationship graph model;

[0122] In this embodiment of the invention, the connection relationships between medium-voltage lines, distribution transformers, and low-voltage users are extracted from the constructed line-transformer-user power supply topology graph model. Taking a power supply system containing 3 medium-voltage lines (L1, L2, L3), 5 distribution transformers (T1-T5), and 20 low-voltage users (U1-U20) as an example, the connection methods between each node are clarified by traversing the node and edge information of the graph model. For example, medium-voltage line L1 is connected to distribution transformers T1 and T2 via cables. T1 further connects to low-voltage users U1-U5, and T2 connects to U6-U10; medium-voltage line L2 connects to T3 and T4, T3 connects to U11-U15, and T4 connects to U16-U18; medium-voltage line L3 connects to T5, and T5 connects to U19-U20. These connection relationships are recorded in the form of an adjacency matrix, where the elements of the matrix are... If node With nodes If a connection exists, then =1, otherwise =0, thus obtaining the complete graph node connection relationship of the entire power supply system.

[0123] Preferably, the energy attenuation coefficient is converted based on the energy loss when the current flows from the corresponding node of the medium-voltage line and distribution transformer through the corresponding node of the low-voltage user, so as to calculate the energy attenuation ratio between the corresponding node of the medium-voltage line and distribution transformer and the corresponding node of the low-voltage user, and thus obtain the energy attenuation coefficient between the corresponding node of the medium-voltage line and distribution transformer and the corresponding node of the low-voltage user.

[0124] In this embodiment of the invention, the energy attenuation coefficient is converted based on the energy loss of the current flowing from the medium-voltage line and the corresponding node of the distribution transformer through the corresponding node of the low-voltage user, which was previously calculated. Taking the path from the medium-voltage line L1 through the distribution transformer T1 to the low-voltage user U1 as an example, the total energy loss of this path is known. =295.704 watts, active power input to this path by medium-voltage line L1. =200×10 3 The formula for calculating the energy attenuation ratio is: Substituting the data, we can obtain =0.00147852, this value is the energy attenuation ratio from the corresponding node of medium-voltage line L1 and distribution transformer T1 to the corresponding node of low-voltage user U1, which is also the energy attenuation coefficient. For all paths in the power supply system from the medium-voltage line and distribution transformer node to the low-voltage user node, the energy attenuation coefficient is calculated one by one according to the above method. For example, the path from medium-voltage line L2 through distribution transformer T3 to low-voltage user U11, the input power is known. =150×10 3 Watts, total energy loss =200 watts, then the energy attenuation coefficient of this path ≈0.001333, finally obtaining the set of energy attenuation coefficients for all paths.

[0125] Preferably, a node energy attenuation matrix is ​​constructed based on the graph node connection relationship between medium-voltage lines, distribution transformers and low-voltage users, and combined with the energy attenuation coefficient between the corresponding graph nodes of medium-voltage lines and distribution transformers and the corresponding graph nodes of low-voltage users, so as to generate the corresponding node energy attenuation coefficient matrix between medium-voltage lines, distribution transformers and low-voltage users.

[0126] In this embodiment of the invention, based on the previously obtained graph node connection relationships between medium-voltage lines, distribution transformers, and low-voltage users, and combined with the previously obtained energy attenuation coefficients, a node energy attenuation matrix is ​​constructed. Using the adjacency matrix as a basis, for node pairs with connection relationships (i.e., those in the adjacency matrix)... For positions where =1, fill the corresponding energy attenuation coefficient into the newly constructed matrix; for node pairs that do not have a connection relationship ( =0), the element at this position is set to 0. For example, in the node energy attenuation coefficient matrix, if the medium-voltage line L1 is connected to the distribution transformer T1 and the energy attenuation coefficient is 0.00147852, then the element at the corresponding L1 and T1 positions in the matrix is ​​0. =0.00147852; If there is no direct connection between medium-voltage line L1 and low-voltage user U11, then =0. In this way, a complete node energy attenuation coefficient matrix is ​​constructed between medium-voltage lines, distribution transformers and low-voltage users. This matrix clearly presents the attenuation characteristics of energy transmission between nodes in the power supply system, providing a quantitative basis for subsequent analysis.

[0127] Preferably, when a power outage occurs at a node corresponding to a medium-voltage line or distribution transformer, the propagation process of the power outage disturbance wave in the online-transformer-user power supply topology diagram model is simulated based on the node energy attenuation coefficient matrix between the medium-voltage line, distribution transformer, and low-voltage user. The propagation intensity of the power outage disturbance wave between each node is iteratively calculated. At the same time, the user-side disturbance sensitivity value of each low-voltage user's corresponding node is statistically analyzed using the analytic hierarchy process based on the propagation intensity of the power outage disturbance wave between each node.

[0128] In the embodiment of the present application, when the power failure occurs in the graph node corresponding to the medium voltage line L1, the propagation process of the power failure disturbance wave is simulated based on the node energy attenuation coefficient matrix, the initial state is set, the power failure disturbance intensity of the L1 node is set to 1, and the power failure disturbance intensity of other nodes is set to 0, the iterative calculation is performed according to the node energy attenuation coefficient matrix, and the calculation formula of the disturbance intensity received by the distribution transformer T1 and T2 nodes connected to the L1 node is , (assuming = 0.0015), then = 1 x 0.00147852 = 0.00147852, = 1 x 0.0015 = 0.0015, then the T1 and T2 nodes further propagate the disturbance to the low-voltage user nodes connected thereto, such as the U1-U5 nodes connected to the T1, (assuming = 0.001), = 0.00147852 x 0.001 = 1.47852 x 10 −6 , and so on, to complete the calculation of the disturbance intensity of each node in the entire power supply topology graph. Meanwhile, the user-side disturbance sensitivity value of each low-voltage user corresponding graph node is calculated by using the analytic hierarchy process, a judgment matrix is constructed, the disturbance of each low-voltage user node is compared from the dimensions of the duration of power failure influence, the degree of energy loss, the influence on the production and life of users, and the like, the weight is calculated, and the comprehensive user-side disturbance sensitivity value is obtained. For example, for the U1 and U2 nodes, the user-side disturbance sensitivity value of the U1 is 0.7 and the user-side disturbance sensitivity value of the U2 is 0.6, which are obtained by the analytic hierarchy process, and the sensitivity calculation of all low-voltage user nodes is completed.

[0129] Preferably, the low-voltage power failure influence of the low-voltage user corresponding graph node is evaluated based on the user-side disturbance sensitivity value of each low-voltage user corresponding graph node and in combination with the corresponding node energy attenuation coefficient matrix, to obtain the power failure influence range of the line-transformer-low-voltage user.

[0130] In the embodiment of the present application, the low-voltage user corresponding graph node is evaluated for low-voltage power failure influence by combining the previously obtained user side disturbance sensitivity value of each low-voltage user corresponding graph node with the node energy attenuation coefficient matrix, and the sensitivity threshold is set to 0.5. When the user side disturbance sensitivity value of a certain low-voltage user node is greater than or equal to 0.5, it is determined that the node is obviously affected by power failure. For example, the user side disturbance sensitivity values of low-voltage users U1, U3 and U4 are 0.7, 0.55 and 0.6 respectively, all of which are greater than the threshold, indicating that these three users are affected by power failure. The sensitivity values of U2 and U5 are less than 0.5, and they are not obviously affected. Further analysis of the topological relationship between the affected low-voltage users shows that U1, U3 and U4 are all powered by distribution transformer T1, and thus it is determined that distribution transformer T1 and the affected low-voltage users (U1, U3 and U4) connected thereto constitute a power failure influence range. If there are multiple distribution transformers whose low-voltage users are affected and these distribution transformers are powered by the same medium-voltage line, the power supply area corresponding to the medium-voltage line is a larger power failure influence area. In this way, the power failure influence range of the line-transformer pair corresponding to the low-voltage users is finally obtained, which provides accurate decision-making reference for the operation and reliability improvement of the power supply system.

[0131] Further, step S34 includes the following steps:

[0132] Step S341: constructing a corresponding four-dimensional tensor space-time influence model according to the time dimension, the space dimension, the equipment level dimension and the influence intensity dimension of the line-transformer pair low-voltage user corresponding power failure influence range, wherein each model sub-voxel represents the power failure influence intensity between the line-transformer pair equipment level and the low-voltage user under a certain space-time dimension;

[0133] In the embodiment of the present application, the four-dimensional tensor space-time influence model is constructed based on the previously obtained power outage influence range corresponding to the line-transformer-user, the time dimension is divided by hours, 24 hours a day is divided into 24 time intervals; the space dimension is in units of geographic coordinate grid of the power supply area, assuming that the power supply area is divided into a 10x10 grid; the equipment level dimension is divided into three layers of distribution line, distribution transformer and low-voltage user; the influence intensity dimension is divided into five levels according to the user side disturbance sensitivity value (0-0.2 is low, 0.2-0.4 is lower, 0.4-0.6 is medium, 0.6-0.8 is higher, and 0.8-1 is high). For example, in a certain power outage event, the distribution line L1 fails in the 3-4 hour interval, affecting the distribution transformers T1 and T2 supplied by it, and further affecting the low-voltage users U1-U5 supplied by T1 and the low-voltage users U6-U10 supplied by T2. In the space-time dimension, L1 corresponds to the space grid (3, 5), T1 corresponds to (3, 6), and U1 corresponds to (3, 6). The user side disturbance sensitivity value of U1 in this period is 0.7, which belongs to the higher influence intensity level. Corresponding to the four-dimensional tensor space-time influence model, in the time dimension of 3-4 hours, the space dimension of (3, 6) grid, the equipment level of low-voltage user U1, and the influence intensity of higher level, the influence intensity of the power outage event on U1 is recorded as 0.7. In the same way, fill in the information in all power outage influence ranges into the model, and complete the construction of the four-dimensional tensor space-time influence model, which clearly presents the power outage influence in different space-time dimensions and equipment levels.

[0134] Step S342: The time delay of the power outage failure from the distribution line node to the distribution transformer node and then to the low-voltage user node is calculated based on the line-transformer-user power supply topology relationship graph model, and a topology conduction time delay matrix corresponding to the line-transformer-user is constructed, wherein each matrix element represents the time required for the power outage failure to propagate from one node to another node;

[0135] In the embodiment of the present application, the conduction time delay of the power outage failure is calculated based on the existing line-transformer-user power supply topology relationship graph model. Taking the failure of the distribution line L1 to the distribution transformer T1 and then to the low-voltage user U1 as an example, through historical power outage data statistics, in the past 50 L1 failure events, the average time required for L1 failure to T1 is 12 minutes, and the average time required for T1 failure to U1 is 8 minutes. These time data are arranged into a topology conduction time delay matrix, wherein the elements in the matrix represent the average time required for the power outage failure to propagate from node to node . For example, the element = 12 minutes, =8 minutes, the conduction time delay of all node pairs in the power supply topology graph model is calculated according to this method, such as the conduction time of the distribution line L2 to the distribution transformer T3 is 10 minutes, that is =10 minutes, and finally a complete topology conduction time delay matrix corresponding to the line-transformer-user is constructed, which provides time parameters for subsequent analysis of the power failure fault propagation process.

[0136] Step S343: Energy entropy weight calculation is performed between each device level in the line-transformer-user power supply topology graph model based on the topology conduction time delay matrix corresponding to the line-transformer-user, to obtain energy entropy weight values of different device levels of the line-transformer-user in the power failure fault, which represent the uncertainty of the device level in the corresponding propagation process of the power failure fault.

[0137] In the embodiment of the present application, the energy entropy weight values of different device levels of the line-transformer-user in the power failure fault are calculated based on the topology conduction time delay matrix corresponding to the line-transformer-user, and the entropy weight method is used for calculation. Taking the three device levels of the distribution line, the distribution transformer and the low-voltage user as an example, first, the probability distribution of each device level in the power failure fault propagation process is calculated. Assuming that in 100 power failure events, the number of times of failure caused by the distribution line is 30, the number of times of failure caused by the distribution transformer is 40, and the number of times of failure caused by the low-voltage user is 30, then the probability of the distribution line is =30 / 100=0.3, the probability of the distribution transformer is =40 / 100=0.4, and the probability of the low-voltage user is =30 / 100=0.3. The entropy value calculation formula is (wherein =1 / ln(n), n is the number of device levels, and here n=3), k=1 / ln(3)≈0.91 is calculated first, then the entropy value of the distribution line is =−0.91×(0.3×ln(0.3)+0.4×ln(0.4)+0.3×ln(0.3))≈0.98, and the entropy values of the distribution transformer and the low-voltage user are 、 respectively. The energy entropy weight value calculation formula is , the energy entropy weight value of the distribution line is , the energy entropy weight value of the distribution transformer is , and the energy entropy weight value of the low-voltage user is . These values represent the uncertainty of each device level in the power failure fault propagation process. The greater the entropy weight value, the higher the uncertainty of the level in the fault propagation.

[0138] Step S344: Based on the four-dimensional tensor space-time influence model, the topological conduction time delay matrix corresponding to the line-variable-house, and the energy entropy weight value corresponding to the line-variable-house different equipment levels in the power failure fault, and using the Monte Carlo integral method to perform time-house number integral calculation in the time, space and equipment level three dimensions, to follow the principle of upper level covering lower level and automatically exclude the repeated statistical time-house number by setting the logical judgment condition, to obtain the power distribution line, distribution transformer and meter corresponding power failure influence time-house number.

[0139] In the embodiment of the present application, by using the Monte Carlo integral method to perform time-house number integral calculation based on the four-dimensional tensor space-time influence model, the topological conduction time delay matrix and the energy entropy weight value, setting the simulation number to 10000 times, in each simulation, the propagation time of the power failure fault between the nodes of each equipment level is determined according to the topological conduction time delay matrix, the power failure influence intensity in different space-time dimensions is obtained by combining the four-dimensional tensor space-time influence model, and then the influence of each equipment level is weighted according to the energy entropy weight value. For example, in one simulation, it is assumed that the power distribution line L1 fails at the 2nd hour, according to the topological conduction time delay matrix, it is conducted to the distribution transformer T1 after 10 minutes, and to the low-voltage user U1 after 20 minutes, in the four-dimensional tensor space-time influence model, the influence intensity of U1 in this space-time dimension is 0.6, and the energy entropy weight value of the power distribution line, the distribution transformer and the low-voltage user is 、 、 , the weighted influence value of U1 is calculated, in the process of calculating the time-house number, the principle of upper level covering lower level is followed, that is, if the power distribution line power failure affects all the distribution transformers and low-voltage users under it, only the power distribution line power failure time-house number is counted, and the time-house number of the subordinate equipment is not counted repeatedly, by setting the logical judgment condition, such as judging whether the equipment level relationship and the power failure time interval overlap, the repeatedly counted time-house number is automatically excluded, after 10000 simulations, the results are statistically averaged, and finally the power distribution line, distribution transformer and meter corresponding power failure influence time-house number is obtained, which provides accurate quantitative data for medium and low voltage power supply reliability evaluation.

[0140] Further, step S4 comprises the following steps:

[0141] Step S41: Based on the power distribution line, distribution transformer and meter corresponding power failure influence time-house number, the node level power failure probability analysis is performed on the graph nodes corresponding to the medium voltage line, distribution transformer and low voltage user in the line-variable-house power supply topology relationship graph model, to obtain the line-variable-house corresponding node level power failure probability.

[0142] In the embodiment of the present application, the node-level power failure probability of the graph nodes in the line-transformer-user power supply topology graph model is analyzed based on the number of households affected by power failure of the distribution line, distribution transformer and meter, assuming that there are two medium voltage lines (L1, L2), three distribution transformers (T1, T2, T3) and 15 low voltage users (U1-U15) in the power supply system, for the medium voltage line L1, the number of households affected by power failure of the corresponding distribution line is 50 households, the total time length for statistical analysis is 10000 hours, according to the node-level power failure probability formula (wherein is the number of households affected by power failure of the node, is the total time length for statistics), the power failure probability of L1 is =50 / 10000=0.005, for the distribution transformer T1, the number of households affected by power failure is 30 households, and the power failure probability is calculated by substituting the formula, and the power failure probability is =30 / 10000=0.003, for the low voltage user U1, the number of households affected by power failure is 5 households, and the power failure probability is calculated to be =5 / 10000=0.0005, according to this method, the graph nodes of all medium voltage lines, distribution transformers and low voltage users in the power supply topology graph model are calculated one by one, and finally the node-level power failure probability set of line-transformer-user corresponding nodes is obtained, which clearly presents the possibility of power failure of each node.

[0143] Step S42: based on the number of households affected by power failure of the distribution line, distribution transformer and meter, the average power failure time index of the graph node edges with topological relationship in the line-transformer-user power supply topology graph model is calculated, and the line-transformer-user corresponding feeder-level average power failure time index is obtained.

[0144] In the embodiment of the present application, the average power failure time index of the graph node edges with topological relationship in the line-transformer-user power supply topology graph model is calculated based on the number of households affected by power failure of the distribution line, distribution transformer and meter, taking the edge connecting the medium voltage line L1 and the distribution transformer T1 as an example, assuming that the number of households affected by power failure of L1 is 50 households, the number of households affected by power failure of T1 is 30 households, and the number of users connected by the edge is 8 households, the average power failure time index calculation formula is (wherein is the number of households affected by power failure of the node, is the number of users connected by the edge, is the total time length for statistics), then the average power failure time index of the edge is =50+30 / 8×10000=0.001. For all nodes with topological relationships in the power supply topology diagram model, such as the edge connecting T1 with low-voltage users U1-U5, the edge connecting L2 with T2, etc., the above formula is used for calculation. Finally, the average outage time index set of the feeder level corresponding to line-transformer-user is obtained. This index reflects the average impact of the feeder on the outage time during the power supply process.

[0145] Step S43: Based on the number of households affected by power outages corresponding to power distribution lines, transformers and meters, perform annual power outage frequency distribution statistics on the entire line-transformer-household topology network within the line-transformer-household power supply topology relationship diagram model to obtain the annual power outage frequency distribution at the network level corresponding to line-transformer-household.

[0146] In this embodiment of the invention, based on the number of households affected by power outages corresponding to distribution lines, transformers, and meters, the annual distribution of power outage frequency is statistically analyzed for the entire line-transformer-household power supply topology network within the line-transformer-household power supply topology diagram model. The statistical period is divided into a year (8760 hours). Assuming that 100 power outage events occur within the 10000 hours of the statistical period, the number of power outages is plotted on the x-axis, and the number of time intervals in which each outage occurs is plotted on the y-axis. For example, the statistics show that there are 20 time intervals with 5 power outages and 10 time intervals with 10 power outages in a year. A probability statistical method is used to calculate the frequency of different power outage frequencies, using the formula: (in The number of power outages is The frequency of occurrence The number of power outages is Number of time intervals (This refers to the total number of time intervals). Assuming the total number of time intervals is 100 and the number of power outages is 5, =20 / 100=0.2; When the number of power outages is 10, =10 / 100=0.1. By analyzing the entire statistical period, the annual power outage frequency distribution at the network level corresponding to the line-transformer-household is finally obtained, which intuitively shows the probability of different power outage frequencies in a year for this power supply topology network.

[0147] Step S44: Based on the outage fault probability at the node level corresponding to the line-transformer-customer, the average outage time index at the feeder level corresponding to the line-transformer-customer, and the annual outage frequency distribution at the network level corresponding to the line-transformer-customer, conduct a medium- and low-voltage power supply reliability assessment on the graph nodes corresponding to the medium-voltage lines, distribution transformers, and low-voltage users in the line-transformer-customer power supply topology graph model to obtain the medium- and low-voltage power supply reliability metric corresponding to the line-transformer-customer.

[0148] In the embodiment of the present application, the graph nodes corresponding to the medium-voltage line, distribution transformer and low-voltage user in the line-transformer-user power supply topology relationship graph model are evaluated for medium-low voltage power supply reliability by the line-transformer-user corresponding node level power failure probability, the line-transformer-user corresponding feeder level average power failure time index and the line-transformer-user corresponding network level annual power failure number distribution, the comprehensive scoring method is adopted, the node level power failure probability weight is set as 0.4, the feeder level average power failure time index weight is set as 0.3, the network level annual power failure number distribution weight is set as 0.3, and the reliability measurement calculation formula is = 0.4, the feeder level average power failure time index weight = 0.3, the network level annual power failure number distribution weight = 0.3, and the reliability measurement calculation formula is (wherein is the frequency corresponding to the high power failure number, and here the frequency of the power failure number greater than or equal to 10 times is taken), taking the medium-voltage line L1 as an example, it is known that = 0.005, the = 0.001 of a certain edge related to L1, = 0.1, and the formula can be obtained = 0.9677, in the same way, the graph nodes of all the medium-voltage lines, distribution transformers and low-voltage users in the power supply topology relationship graph model are calculated, and finally the medium-low voltage power supply reliability measurement value corresponding to the line-transformer-user is obtained, and the higher the value, the stronger the power supply reliability, so as to comprehensively evaluate the reliability level of the medium-low voltage power supply system.

[0149] Further, the present application also provides a medium-low voltage power supply reliability evaluation system based on the HLNN model, which is used for executing the medium-low voltage power supply reliability evaluation method based on the HLNN model as described above, and the medium-low voltage power supply reliability evaluation system based on the HLNN model comprises:

[0150] A multi-scale feature analysis module is configured to acquire the power supply use time sequence data corresponding to the medium-voltage line, distribution transformer and low-voltage user, construct a hierarchical long short-term memory neural network (HLNN) model, and simultaneously perform hierarchical multi-scale feature analysis on the power supply use time sequence data corresponding to the medium-voltage line, distribution transformer and low-voltage user based on the hierarchical long short-term memory neural network (HLNN) model, so as to obtain the line-transformer-user corresponding hierarchical different time scale power failure features;

[0151] A topology connection construction module is configured to analyze the line-transformer-user corresponding power supply space topology relationship based on the line-transformer-user corresponding hierarchical different time scale power failure feature, and construct the topology connection of the corresponding medium-voltage line, distribution transformer and low-voltage user as the graph nodes according to the line-transformer-user corresponding power supply space topology relationship, so as to generate a line-transformer-user power supply topology relationship graph model;

[0152] The time and household number hierarchical calculation module is configured to perform low-voltage power failure influence evaluation on the low-voltage user corresponding graph nodes based on the line-transformer-household power supply topology graph model and in combination with the corresponding graph nodes of the medium-voltage line and the distribution transformer, so as to obtain a line-transformer low-voltage user corresponding power failure influence range; and perform time and household number hierarchical calculation according to the line-transformer low-voltage user corresponding power failure influence range, so as to obtain the distribution line, distribution transformer and meter corresponding power failure influence time and household number.

[0153] The medium and low voltage reliability evaluation module is configured to perform medium and low voltage power supply reliability evaluation on the graph nodes of the corresponding medium-voltage line, distribution transformer and low-voltage user in the line-transformer-household power supply topology graph model based on the distribution line, distribution transformer and meter corresponding power failure influence time and household number, so as to obtain the line-transformer-household corresponding medium and low voltage power supply reliability metric.

[0154] The above description is merely that of this application, to enable those skilled in the art to understand or implement the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein can be applied to other embodiments without departing from the spirit or scope of the application. Accordingly, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for evaluating reliability of medium and low voltage power supply based on a HLNN model, characterized in that, The method comprises the following steps: Step S1: obtaining power supply time sequence data corresponding to the medium-voltage line, distribution transformer and low-voltage user, and constructing a hierarchical long short-term memory neural network HLNN model, and simultaneously performing hierarchical multi-scale feature analysis on the power supply time sequence data corresponding to the medium-voltage line, distribution transformer and low-voltage user based on the hierarchical long short-term memory neural network HLNN model to obtain line-transformer-user corresponding hierarchical different time scale power failure features; Step S2: analyzing the line-transformer-user corresponding power supply space topological relationship based on the line-transformer-user corresponding hierarchical different time scale power failure features, and topologically connecting the corresponding medium-voltage line, distribution transformer and low-voltage user as graph nodes to construct a line-transformer-user power supply topological relationship graph model; Step S3: performing low-voltage power failure influence evaluation on the low-voltage user corresponding graph node based on the line-transformer-user power supply topological relationship graph model and in combination with the medium-voltage line and distribution transformer corresponding graph nodes to obtain line-transformer corresponding low-voltage user power failure influence range; and performing time household number hierarchical calculation according to the line-transformer corresponding low-voltage user power failure influence range to obtain distribution line, distribution transformer and meter corresponding power failure influence time household number; Step S4: performing medium-low voltage power supply reliability evaluation on the line-transformer-user power supply topological relationship graph model based on the distribution line, distribution transformer and meter corresponding power failure influence time household number to obtain line-transformer-user corresponding medium-low voltage power supply reliability measurement.

2. The method for reliability evaluation of medium and low voltage power supply based on HLNN model according to claim 1, characterized in that, Step S1 comprises the following steps: Step S11: obtaining power supply time sequence data corresponding to the medium-voltage line; Step S12: obtaining power supply time sequence data corresponding to the distribution transformer; Step S13: obtaining power supply time sequence data corresponding to the low-voltage user; Step S14: constructing a hierarchical long short-term memory neural network HLNN model, which comprises low, medium and high layer corresponding long short-term memory neural network architecture, the low layer corresponding long short-term memory neural network architecture is used for extracting low-voltage user corresponding different time scale power failure features, the medium layer corresponding long short-term memory neural network architecture is used for extracting distribution transformer corresponding different time scale power failure features, and the high layer corresponding long short-term memory neural network architecture is used for extracting medium-voltage line corresponding different time scale power failure features, wherein each layer corresponding long short-term memory neural network architecture is used for capturing transient power failure sub-features corresponding to voltage sag and inrush current at a short time scale, extracting event power failure sub-features corresponding to load mutation and device start at a medium time scale, and identifying power failure state trend sub-features corresponding to planned power failure and device aging at a long time scale; Step S15: performing hierarchical multi-scale feature analysis on the power supply time sequence data corresponding to the medium-voltage line, distribution transformer and low-voltage user based on the hierarchical long short-term memory neural network HLNN model to obtain line-transformer-user corresponding hierarchical different time scale power failure features.

3. The method of claim 2, wherein the method further comprises: Step S15 comprises the following steps: Step S151: input the power supply usage time series data corresponding to the medium voltage line, distribution transformer and low voltage user into the corresponding low, medium and high layer long short term memory neural network architecture in the hierarchical long short term memory neural network HLNN model respectively for dimension tensor reconstruction, to decompose the voltage, current and power physical quantities in the power supply usage time series data into three power outage characteristic factor matrices of time evolution mode, device response mode and basic fluctuation mode according to different time window, device level and measurement point, to obtain three power outage characteristic tensor factor matrices corresponding to the medium voltage line, distribution transformer and low voltage user; Step S152: perform power outage characteristic hierarchical mapping on the three power outage characteristic tensor factor matrices corresponding to the medium voltage line, distribution transformer and low voltage user, to map the three power outage characteristic tensor factor matrices to a line-transform-user three level cube, wherein the X axis of the cube represents the device level, the Y axis represents the time scale, and the Z axis represents the power outage characteristic type, to obtain a line-transform-user corresponding hierarchical power outage characteristic three-dimensional mapping cube; Step S153: calculate the matrix eigenvalue and eigenvector corresponding to the power outage characteristic cube through the line-transform-user corresponding hierarchical power outage characteristic three-dimensional mapping cube, and determine the power outage characteristic significance of each level under different time scales according to the matrix eigenvalue and eigenvector corresponding to the power outage characteristic cube; Step S154: introduce corresponding self-attention mechanisms between the long short term memory neural network architectures of each level in the hierarchical long short term memory neural network HLNN model, and perform multi-scale feature fusion analysis on the line-transform-user corresponding hierarchical power outage characteristic three-dimensional mapping cube by combining the power outage characteristic significance of each level under different time scales, to obtain different time scale power outage characteristics corresponding to the line-transform-user hierarchical layer.

4. The method of claim 1, wherein the method further comprises: Step S2 includes the following steps: Step S21: statistically analyze the line-transform-user corresponding power supply space topology relationship based on the line-transform-user corresponding hierarchical different time scale power outage characteristics; Step S22: take the corresponding medium voltage line, distribution transformer and low voltage user as corresponding graph nodes, and based on the line-transform-user corresponding power supply space topology relationship, perform power topology change constraint analysis between any two graph nodes with topological relationship to calculate the conditional probability that the state change of one graph node causes the state change of another graph node, and take it as the constraint condition between the two graph nodes, to generate the power supply topology state constraint condition between any two graph nodes; Step S23: based on the power supply topology state constraint condition between any two graph nodes and combining the line-transform-user corresponding power supply space topology relationship, perform topology connection construction between the graph nodes corresponding to the medium voltage line, distribution transformer and low voltage user, to generate a line-transform-user power supply topology relationship graph model.

5. The method for reliability evaluation of medium and low voltage power supply based on HLNN model according to claim 4, characterized in that, Step S21 includes the following steps: Perform power outage fluctuation time delay analysis on the line-transform-user corresponding hierarchical different time scale power outage characteristics, to obtain the time sequence power outage fluctuation time delay between the line-transform-user corresponding different level devices; The power outage feature phase difference between the different levels of equipment corresponding to the line-transformer-user is obtained by the power outage feature of the line-transformer-user corresponding to the different time scales of the hierarchical layer, and the correlation and causality evaluation is carried out according to the time sequence power outage fluctuation time delay and the power outage feature phase difference between the different levels of equipment corresponding to the line-transformer-user, so as to obtain the correlation and causality influence strength between the different levels of equipment corresponding to the line-transformer-user. The energy flow direction between the different levels of equipment corresponding to the line-transformer-user is obtained by the power outage feature of the line-transformer-user corresponding to the different time scales of the hierarchical layer, and the power supply topology relationship connection analysis is carried out based on the energy flow direction between the different levels of equipment corresponding to the line-transformer-user and combined with the correlation and causality influence strength between the different levels of equipment corresponding to the line-transformer-user, so as to obtain the power supply space topology relationship of the line-transformer-user corresponding to the different levels of equipment as the power supply space node, the edge weight between each power supply space node represents the correlation and causality influence strength, and the direction of the edge represents the energy flow direction.

6. The method of claim 1, wherein the method further comprises: Step S3 includes the following steps: Step S31: obtaining the topology-energy coupling tensor between the medium voltage line, distribution transformer and low voltage user corresponding graph nodes according to the line-transformer-user power supply topology relationship graph model; Step S32: based on the topology-energy coupling tensor between the medium voltage line, distribution transformer and low voltage user corresponding graph nodes and combined with the corresponding ohm law and energy loss calculation current in the power system, the energy loss amount of the low voltage user corresponding graph node is calculated when the current flows from the medium voltage line and the distribution transformer corresponding graph node; Step S33: based on the energy loss amount of the current flowing from the medium voltage line and the distribution transformer corresponding graph node to the low voltage user corresponding graph node, the low voltage user corresponding graph node is evaluated for low voltage power outage influence, to obtain the power outage influence range of the line-transformer-user corresponding to the low voltage user; Step S34: according to the power outage influence range of the line-transformer-user corresponding to the low voltage user, the time household number layer calculation is carried out, to obtain the power outage influence time household number of the distribution line, distribution transformer and meter corresponding.

7. The method of claim 6, wherein the method further comprises: Step S33 includes the following steps: Obtaining the graph node connection relationship between the medium voltage line, distribution transformer and low voltage user through the line-transformer-user power supply topology relationship graph model; According to the energy loss amount of the current flowing from the medium voltage line and the distribution transformer corresponding graph node to the low voltage user corresponding graph node, the energy attenuation coefficient is converted to calculate the energy attenuation ratio value between the medium voltage line and the distribution transformer corresponding graph node to the low voltage user corresponding graph node, to obtain the energy attenuation coefficient between the medium voltage line and the distribution transformer corresponding graph node to the low voltage user corresponding graph node; Based on the graph node connection relationship between the medium voltage line, distribution transformer and low voltage user and combined with the energy attenuation coefficient between the medium voltage line and the distribution transformer corresponding graph node to the low voltage user corresponding graph node, the node energy attenuation matrix is constructed to generate the node energy attenuation coefficient matrix corresponding to the medium voltage line, distribution transformer and low voltage user. When the corresponding graph node of the medium-voltage line or distribution transformer is powered off, based on the node energy attenuation coefficient matrix corresponding to the medium-voltage line, distribution transformer and low-voltage user, the propagation process of the power-off disturbance wave in the line-transformer-user power supply topology graph model is simulated, and the propagation intensity of the power-off disturbance wave between the graph nodes is iteratively calculated. At the same time, according to the propagation intensity of the power-off disturbance wave between the graph nodes, the user-side disturbance sensitivity value of each low-voltage user corresponding graph node is calculated by using the analytic hierarchy process; Based on the user-side disturbance sensitivity value of each low-voltage user corresponding graph node and the corresponding node energy attenuation coefficient matrix, the low-voltage user corresponding graph node is evaluated for low-voltage power-off impact to obtain the power-off impact range of the line-transformer on the low-voltage user.

8. The method of claim 6, wherein the method further comprises: Step S34 includes the following steps: Step S341: constructing a corresponding four-dimensional tensor space-time impact model according to the time dimension, space dimension, device level dimension and impact intensity dimension of the power-off impact range of the line-transformer on the low-voltage user, wherein each model sub-voxel represents the power-off impact intensity between the line-transformer device level and the low-voltage user under a certain space-time dimension; Step S342: calculating the time delay of the power-off fault from the distribution line node to the distribution transformer node and then to the low-voltage user node based on the line-transformer-user power supply topology graph model, and constructing a topology conduction time delay matrix corresponding to the line-transformer-user, wherein each matrix element represents the time required for the power-off fault to propagate from one node to another node; Step S343: based on the topology conduction time delay matrix corresponding to the line-transformer-user, the energy entropy weight of each device level in the line-transformer-user power supply topology graph model is calculated to obtain the energy entropy weight value of the different device levels of the line-transformer-user in the power-off fault, which represents the uncertainty of the device level in the corresponding propagation process of the power-off fault; Step S344: based on the four-dimensional tensor space-time impact model, the topology conduction time delay matrix corresponding to the line-transformer-user, and the energy entropy weight value of the different device levels of the line-transformer-user in the power-off fault, and using the Monte Carlo integration method to perform time-house number integration calculation in the time, space and device level dimensions, the power-off impact time-house number corresponding to the distribution line, distribution transformer and meter is obtained by following the principle of upper level covering lower level and automatically excluding repeated statistical time-house numbers through setting logical judgment conditions.

9. The method of claim 1, wherein the method further comprises: Step S4 includes the following steps: Step S41: based on the power-off impact time-house number corresponding to the distribution line, distribution transformer and meter, the node-level power-off fault probability analysis of the corresponding graph nodes of the line-transformer-user in the line-transformer-user power supply topology graph model is performed, and the node-level power-off fault probability of the line-transformer-user is obtained; Step S42: based on the power-off impact time-house number corresponding to the distribution line, distribution transformer and meter, the average power-off time index calculation of the graph node edges with topological relationship in the line-transformer-user power supply topology graph model is performed, and the feeder-level average power-off time index of the line-transformer-user is obtained; Step S43: Based on the outage influence time household number corresponding to the distribution line, distribution transformer and meter, the annual outage frequency distribution of the entire line-transformer-household topology network in the line-transformer-household power supply topology graph model is statistically calculated, and the annual outage frequency distribution of the line-transformer-household corresponding network level is obtained; Step S44: According to the outage failure probability of the line-transformer-household corresponding node level, the average outage time index of the line-transformer-household corresponding feeder level, and the annual outage frequency distribution of the line-transformer-household corresponding network level, the graph nodes corresponding to the medium and low voltage lines, distribution transformers and low voltage users in the line-transformer-household power supply topology graph model are evaluated for medium and low voltage power supply reliability, to obtain the medium and low voltage power supply reliability measurement corresponding to the line-transformer-household.

10. A system for reliability evaluation of medium and low voltage power supply based on HLNN model, characterized in that, The medium and low voltage power supply reliability evaluation system based on the HLNN model for executing the method of claim 1 comprises: A multi-scale feature analysis module is configured to obtain power supply time sequence data corresponding to the medium voltage line, distribution transformer and low voltage user, construct a hierarchical long short-term memory neural network (HLNN) model, and perform hierarchical multi-scale feature analysis on the power supply time sequence data corresponding to the medium voltage line, distribution transformer and low voltage user based on the hierarchical long short-term memory neural network (HLNN) model, so as to obtain different time scale outage features corresponding to the line-transformer-household at different hierarchical levels. A topology connection construction module is configured to analyze the power supply spatial topology relationship corresponding to the line-transformer-household based on the different time scale outage features corresponding to the line-transformer-household at different hierarchical levels, and construct the topology connection of the corresponding graph nodes of the medium voltage line, distribution transformer and low voltage user according to the power supply spatial topology relationship corresponding to the line-transformer-household, to generate a line-transformer-household power supply topology graph model. A time household number hierarchical calculation module is configured to evaluate the low voltage outage influence of the graph nodes corresponding to the low voltage users based on the line-transformer-household power supply topology graph model and the graph nodes corresponding to the medium voltage line and distribution transformer, to obtain the outage influence range of the line-transformer-household corresponding to the low voltage users; and perform time household number hierarchical calculation according to the outage influence range of the line-transformer-household corresponding to the low voltage users, to obtain the outage influence time household number corresponding to the distribution line, distribution transformer and meter. A medium and low voltage reliability evaluation module is configured to evaluate the medium and low voltage power supply reliability of the graph nodes corresponding to the medium voltage line, distribution transformer and low voltage user in the line-transformer-household power supply topology graph model based on the outage influence time household number corresponding to the distribution line, distribution transformer and meter, to obtain the medium and low voltage power supply reliability measurement corresponding to the line-transformer-household.

Images