Power grid risk assessment method and system
By constructing local and global perspective matrix, collaborative filtering and residual repair models, the problem of insufficient processing capabilities for abnormal data and residuals in smart grid risk assessment is solved, and higher evaluation reliability and data integrity are achieved.
Patent Information
- Application Number
- CN202411990808.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-31
- Publication Date
- 2025-04-29
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing smart grid risk assessment technology has weak ability to handle abnormal data and residuals in power monitoring data, resulting in error accumulation and low reliability of risk assessment.
By collecting power monitoring data, the distortion data in the network structure is determined, the local perspective matrix and the global perspective matrix are constructed, the attenuation coefficient is obtained through collaborative filtering, the residual repair model is constructed, the data is fine-grained based on the offset trend and attenuation coefficient, and the confidence evaluation is carried out to obtain the risk assessment results.
It improves the reliability of intelligent risk assessment in the power grid, reduces evaluation errors by effectively repairing residual and abnormal data, and improves data integrity and evaluation credibility.
Smart Images
Figure CN119904100A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of smart grid technology, and more specifically, to a grid risk assessment method and system. Background Art
[0002] Smart grid is a new type of power system that deeply integrates traditional power grid with modern information technology. It aims to realize intelligent operation and precise dispatch of power grid through sensor technology, big data analysis and artificial intelligence optimization. At present, smart grid technology is developing towards higher intelligence, distribution and low carbonization. In the future, smart grid will pay more attention to the application of artificial intelligence, big data and blockchain technology, further enhance the autonomy and adaptability of the system, and help achieve the goal of global energy transformation.
[0003] As an important part of smart grid operation, grid risk assessment technology can identify potential risks and predict grid operation status by analyzing power monitoring data. Among them, grid risk is assessed through historical data statistics and abnormal pattern recognition, and the risk of grid is predicted by big data training. Although it has a certain degree of accuracy, it has weak processing capabilities for abnormal data and residuals, which easily leads to error accumulation, making data repair difficult and risk assessment credibility low. Therefore, how to perform residual repair and confidence risk assessment on power monitoring data to improve the reliability of smart grid risk assessment is a difficult problem faced by the industry. Summary of the invention
[0004] The present application provides a power grid risk assessment method and system, which can perform residual repair and confidence risk assessment on power monitoring data to improve the reliability of power grid intelligent risk assessment.
[0005] In a first aspect, the present application provides a power grid risk assessment method, the assessment method comprising the following steps:
[0006] Collecting power monitoring data of the target power grid area and determining the network structure of the smart grid used by the dispatching center of the target power grid area;
[0007] Determining distorted data in the power monitoring data through an isolation device in the network structure, and constructing a local viewing angle matrix and a global viewing angle matrix based on the distorted data;
[0008] Determining the deviation trend of the smart grid during communication according to the global perspective matrix, performing collaborative filtering on the local perspective matrix, and thereby obtaining an attenuation coefficient of the smart grid during data isolation;
[0009] Acquire historical monitoring data of the target power grid area, and then construct a residual repair model of the smart grid during data isolation based on the historical monitoring data, and determine the repair granularity of the power monitoring data based on the offset trend and the attenuation coefficient through the residual repair model;
[0010] A confidence assessment is performed on the smart grid based on the repair granularity of the power monitoring data, thereby obtaining a risk assessment result for the target power grid area.
[0011] In this embodiment, the power monitoring data of the target power grid area is collected by a data collection server.
[0012] In this embodiment, determining the distorted data in the power monitoring data by the isolation device in the network structure specifically includes:
[0013] respectively collecting power transmission data before and after isolation of the isolation device in the network structure, and then determining the data transmission waveform through the power transmission data before and after isolation of the isolation device;
[0014] Distorted data in the power monitoring data is filtered out based on the data transmission waveform.
[0015] In this embodiment, constructing a local viewing matrix and a global viewing matrix based on the distortion data specifically includes:
[0016] Obtain dispatching flow information of the target power grid area;
[0017] Performing a fast Fourier transform on the distorted data to obtain a frequency domain space of the distorted data;
[0018] Based on the scheduling power flow information, spatial features of different power flow nodes are extracted from the frequency domain space, and then a local perspective matrix and a global perspective matrix are constructed according to all the spatial features.
[0019] In this embodiment, determining the deviation trend of the smart grid during communication according to the global viewing angle matrix specifically includes:
[0020] Mapping the data points in the global perspective matrix to a geometric space;
[0021] The data fluctuation curve is determined by the Euclidean distance of all data points in the geometric space, and then the trend of the data fluctuation curve is used as the deviation trend of the smart grid during communication.
[0022] In this embodiment, collaborative filtering is performed on the local perspective matrix to obtain the attenuation coefficient of the smart grid during data isolation, specifically including:
[0023] Performing time-space collaborative division on the local perspective matrix to obtain a time collaborative matrix and a space collaborative matrix;
[0024] Performing a primary filtering on the data points in the local perspective matrix according to the correlation coefficient of the time synergy matrix, and then performing a secondary filtering on the data points in the local perspective matrix based on the correlation coefficient of the space synergy matrix to obtain a filtered local perspective matrix;
[0025] The attenuation coefficient of the smart grid during data isolation is determined according to the filtered local viewing angle matrix and the local viewing angle matrix.
[0026] In this embodiment, historical monitoring data of the target power grid area is obtained from the dispatching database of the smart grid.
[0027] In this embodiment, the residual repair model is a neural network model built based on a multi-view attention mechanism.
[0028] In this embodiment, the confidence assessment of the smart grid is performed based on the repair granularity of the power monitoring data, and the risk assessment result of the target power grid area is obtained, which specifically includes:
[0029] Determining a repair error of the power monitoring data based on a repair granularity of the power monitoring data;
[0030] Determining the fluctuation coefficient and confidence level of the operation state of the smart grid through the repair error;
[0031] A weighted evaluation is performed on the fluctuation coefficient and the confidence level to obtain a risk assessment result of the target power grid area.
[0032] In a second aspect, the present application provides a power grid risk assessment system for executing a power grid risk assessment method, the assessment system comprising:
[0033] A collection module, used to collect power monitoring data of a target power grid area and determine the network structure of a dispatch center in the target power grid area using a smart grid;
[0034] An isolation and division module, used to determine the distorted data in the power monitoring data through the isolation device in the network structure, and construct a local viewing angle matrix and a global viewing angle matrix based on the distorted data;
[0035] A matrix processing module, used to determine the deviation trend of the smart grid during communication according to the global perspective matrix, perform collaborative filtering on the local perspective matrix, and then obtain the attenuation coefficient of the smart grid during data isolation;
[0036] A data repair module, used to obtain historical monitoring data of a target power grid area, and then construct a residual repair model of the smart grid during data isolation based on the historical monitoring data, and determine the repair granularity of the power monitoring data based on the offset trend and the attenuation coefficient through the residual repair model;
[0037] The confidence assessment module is used to perform confidence assessment on the smart grid based on the repair granularity of the power monitoring data, and then obtain a risk assessment result of the target power grid area.
[0038] The technical solution provided by the embodiments disclosed in this application has the following beneficial effects:
[0039] First, the power monitoring data of the target power grid area is collected, and the network structure of the smart grid used by the dispatching center of the target power grid area is determined; secondly, the distorted data in the power monitoring data is determined through the isolation device in the network structure, and a local perspective matrix and a global perspective matrix are constructed based on the distorted data; then, the offset trend of the smart grid during communication is determined according to the global perspective matrix, and the local perspective matrix is collaboratively filtered to obtain the attenuation coefficient of the smart grid during data isolation; then, the historical monitoring data of the target power grid area is obtained, and then a residual repair model of the smart grid during data isolation is constructed based on the historical monitoring data, and the repair granularity of the power monitoring data is determined based on the offset trend and the attenuation coefficient through the residual repair model; finally, the smart grid is confidence evaluated based on the repair granularity of the power monitoring data, and the risk assessment result of the target power grid area is obtained.
[0040] It can be seen that in this application, firstly, by collecting and screening distorted data, abnormal factors that may affect the accuracy of the evaluation are eliminated, and a purer data set is provided for constructing the matrix. After the distorted data is screened out, constructing a local perspective matrix is conducive to focusing on the impact of a single area or device, and constructing a global perspective matrix comprehensively reflects the operating status of the entire network, forming a multi-dimensional analysis framework, and providing data support for the analysis of the offset trend and the attenuation coefficient; secondly, the data in the local perspective matrix is collaboratively filtered and optimized, interference factors are eliminated, and more accurate regional features are extracted to obtain the attenuation coefficient, which is convenient for quantitatively describing the impact of data isolation on network performance and providing important parameters for residual repair; then, the residual repair model of the smart grid during data isolation is constructed through historical monitoring data. The residual repair model effectively repairs the distorted data by combining historical patterns and real-time characteristics, which is conducive to improving the integrity of the data, reducing the evaluation error caused by data anomalies, and improving the overall reliability; finally, the repair results are quantitatively analyzed through confidence evaluation. The confidence evaluation combines the repair error and the network characteristics of the smart grid, which is convenient for comprehensively measuring the risk status of the power grid area and providing a basis for decision-making, thereby helping to improve the effect of smart grid evaluation.
[0041] In summary, the technical solution adopted in this application can perform residual repair and confidence risk assessment on power monitoring data to improve the reliability of intelligent risk assessment of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:
[0043] Figure 1 is a flow chart of a power grid risk assessment method provided in this application;
[0044] Figure 2 is an exemplary flow chart for determining distorted data in power monitoring data provided by the present application;
[0045] Figure 3 is an exemplary flow chart for determining the attenuation coefficient of a smart grid during data isolation according to the present application;
[0046] Figure 4 It is a module structure diagram of the power grid risk assessment system provided in this application. DETAILED DESCRIPTION
[0047] In order to make the purpose, technical scheme and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments and drawings. The schematic implementation modes and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention. It should be noted that the present invention is already in the actual development and use stage.
[0048] Embodiment 1: In order to better understand the technical solution of the present application, the technical solution of the present application will be described in detail below in conjunction with the accompanying drawings and specific implementation methods of the specification. Figure 1 As shown, this figure is an exemplary flow chart of a power grid risk assessment method shown in this embodiment of the present application, and the assessment method includes the following steps:
[0049] In step S1, the power monitoring data of the target power grid area is collected, and the network structure of the smart grid used by the dispatching center of the target power grid area is determined.
[0050] In specific implementation, the power monitoring data of the target power grid area is collected through the data acquisition server, and the power monitoring data of the target power grid area can be collected by deploying a variety of acquisition terminals and then transmitted to the data acquisition server. Among them, the acquisition terminals include power sensors, smart meters and environmental monitoring equipment installed in substations, transmission lines and user ends. In actual implementation, it can be combined with data transmission, edge computing and big data analysis technologies to provide technical support for real-time perception and abnormal monitoring of power grid operation status. It should be noted that in this application, the collected power monitoring data not only includes power grid operation parameters, but also includes equipment status information and environmental influencing factors, etc., which provides a rich data basis for subsequent residual repair and confidence risk assessment.
[0051] In this embodiment, the network structure of the dispatching center in the target power grid area is determined by modeling the communication equipment, isolation devices and information flow topology relationships in the target area through the network analysis module of the dispatching center, thereby obtaining the network structure of the smart grid; in specific implementation, the communication network topology model of the target power grid area can be constructed based on the actual communication link and equipment configuration, and the key nodes (such as isolation devices) therein are marked and functionally analyzed. The network structure includes different communication nodes in the target power grid area and the association relationships between different communication nodes.
[0052] In step S2, the distorted data in the power monitoring data is determined by the isolation device in the network structure, and a local viewing angle matrix and a global viewing angle matrix are constructed based on the distorted data.
[0053] Preferably, in this embodiment, reference Figure 2As shown, this figure is an exemplary flow chart of determining distorted data in power monitoring data in an embodiment of the present application. In this embodiment, determining the distorted data in the power monitoring data by the isolation device in the network structure can be implemented by the following steps:
[0054] First, in step S21, power transmission data before and after isolation of the isolation device in the network structure are respectively collected, and then the data transmission waveform is determined by the power transmission data before and after isolation of the isolation device;
[0055] Then, in step S22, distorted data in the power monitoring data is screened out based on the data transmission waveform.
[0056] It should be noted that the isolation device in this embodiment refers to a device used for data collection and encryption in the smart grid, such as: a vertical encryption device, a reverse isolation device and a channel establishment module. The isolation device is installed at different network nodes of the network structure and is used for real-time collection, screening and encryption of power monitoring data.
[0057] In specific implementation, the transmission parameters of the power signal before entering the isolation device and the output parameters of the collected signal after being processed by the isolation device are collected through a high-speed communication link to obtain the power transmission data before and after the isolation of the isolation device, wherein the power transmission data includes: voltage, current and point power, etc., and the power transmission data before and after the isolation of the isolation device are denoised and normalized, which can eliminate abnormal signals caused by sampling noise or equipment errors. As a preferred embodiment, denoising can be performed by low-pass filtering and Gaussian filtering, and then the power transmission data before and after the isolation of the isolation device can be plotted into a waveform diagram through a visualization tool (such as the Matplotlib library in Python), thereby obtaining a data transmission waveform, which is convenient for intuitively displaying the characteristic changes in the signal transmission process and facilitating the screening of distorted data. In other embodiments, other existing technologies can also be used for denoising and normalization, which are not limited here.
[0058] In addition, in a specific implementation, the average phase of the data transmission waveform is obtained, and the average phase is used as an offset index. The data transmission waveform is symmetrically divided according to the data before and after isolation, and the phase offset between each power transmission data before isolation and the symmetrical power transmission data after isolation is calculated by dichotomy, and then the power transmission data after isolation whose phase offset is higher than the offset index is used as distorted data.
[0059] It should be noted that the distorted data in this application refers to abnormal data in the power monitoring data that deviates from the normal state due to equipment failure, signal interference or transmission line abnormality during the power transmission process. The abnormal data includes: phase disorder, amplitude abnormality, power offset and data jump (i.e., nonlinear jumps between continuous sampling data). In this application, by screening the distorted data, the error information can be quickly analyzed, which is conducive to improving the accuracy of data analysis and optimizing the data acquisition process.
[0060] In this embodiment, the local viewing angle matrix and the global viewing angle matrix are constructed based on the distortion data in the following manner, namely:
[0061] Obtain dispatching flow information of the target power grid area;
[0062] Performing a fast Fourier transform on the distorted data to obtain a frequency domain space of the distorted data;
[0063] Based on the scheduling power flow information, spatial features of different power flow nodes are extracted from the frequency domain space, and then a local perspective matrix and a global perspective matrix are constructed according to all the spatial features.
[0064] In specific implementation, first, the dispatching flow information of the target power grid area is obtained through the data platform of the power dispatching center. The dispatching flow information is information describing the power flow in the power grid. The dispatching flow information includes data such as voltage, frequency and power flow direction of different flow nodes in different flow lines, reflecting the operation status of the power grid; then, a fast Fourier transform algorithm model is initialized, and the distorted data is fast Fourier transformed through the fast Fourier transform algorithm model, and the distorted data obtained by the fast Fourier transform is converted into the frequency domain space through a visualization tool; finally, the matrix randomly composed of the spatial features corresponding to all the flow nodes on the same line in the frequency domain space can be used as the local perspective matrix, and the matrix composed of all the spatial features in the frequency domain space according to the acquisition time is used as the global perspective matrix, wherein the spatial features are the features describing the power flow and fluctuation of the distorted data in the frequency domain space, and the wavelet transform in the prior art can be used to extract the spatial features corresponding to each distorted data in the frequency domain space, which is conducive to detecting the instantaneous disturbance of the distorted data.
[0065] It should be noted that in this embodiment, by constructing a local perspective matrix, it is convenient to identify and analyze distorted data in a specific area, locate the source of the fault, respond to local problems in a timely manner, and enhance the local anti-interference ability of the power grid. Constructing a global perspective matrix is conducive to comprehensively reflecting the power flow status of each node in the power grid and providing data support for the overall scheduling and optimization of the power grid. In this application, by constructing local and global matrices, the smart grid can more efficiently provide decision support, optimize power flow, and thus improve the speed of fault response.
[0066] In step S3, the deviation trend of the smart grid during communication is determined according to the global perspective matrix, and collaborative filtering is performed on the local perspective matrix to obtain the attenuation coefficient of the smart grid during data isolation.
[0067] In this embodiment, the deviation trend of the smart grid during communication may be determined according to the global viewing angle matrix in the following manner, namely:
[0068] Mapping the data points in the global perspective matrix to a geometric space;
[0069] The data fluctuation curve is determined by the Euclidean distance of all data points in the geometric space, and then the trend of the data fluctuation curve is used as the deviation trend of the smart grid during communication.
[0070] In the specific implementation, first, the global perspective matrix is a matrix that reflects the overall power transmission characteristics of the smart grid, which includes various spatial characteristics of multiple nodes in the power grid. By reducing the dimension of each data point in the global perspective matrix and mapping it to the geometric space through principal component analysis (i.e. PCA), the mutual relationship between the nodes of the power grid and the power flow trend can be revealed; then, the Euclidean distance is an indicator that measures the straight-line distance between two data points in the geometric space, which can reflect the difference between the two data points in the same dimension. During the communication process of the power grid, the distance change between the data points can reveal the characteristics of power transmission fluctuations. The Euclidean distance can be recorded in a time series, and a data fluctuation curve can be obtained by visualizing the fluctuation diagram of the Euclidean distance in a time series. Then, a smoothing function is used to smooth and analyze the data fluctuation curve to obtain the trend of the data fluctuation curve, and the trend of the data fluctuation curve is used as the offset trend of the smart grid during communication. By extracting the trend of the data fluctuation curve, the potential offset problem of the smart grid in communication can be accurately identified, and the possible failure or transmission anomaly of the power grid can be predicted in advance to ensure the security of the smart grid.
[0071] It should be noted that the offset trend in this application refers to the trend of change in which the power monitoring data gradually deviates from its normal state due to various factors during the smart grid communication process. The offset trend reflects the fluctuation of certain parameters (such as voltage, current, and power, etc.) in the smart grid. By monitoring the offset trend, the operating status of the smart grid during the communication and data transmission process can be judged. For example: if the offset trend continues to rise, it may indicate that the grid load is gradually increasing and there is a risk of overload. If the offset trend decreases, it may be that the communication equipment is faulty or damaged.
[0072] Preferably, in this embodiment, reference Figure 3As shown, this figure is an exemplary flow chart of determining the attenuation coefficient of the smart grid during data isolation in an embodiment of the present application. In this embodiment, collaborative filtering is performed on the local perspective matrix to obtain the attenuation coefficient of the smart grid during data isolation, which can be specifically implemented by the following steps:
[0073] First, in step S31, the local view matrix is divided into time and space to obtain a time coordination matrix and a space coordination matrix;
[0074] Secondly, in step S32, the data points in the local perspective matrix are initially filtered according to the correlation coefficient of the time synergy matrix, and then the data points in the local perspective matrix are filtered again based on the correlation coefficient of the space synergy matrix to obtain a filtered local perspective matrix;
[0075] Finally, in step S33, the attenuation coefficient of the smart grid during data isolation is determined according to the filtered local viewing angle matrix and the local viewing angle matrix.
[0076] In the specific implementation, first, the Spearman correlation coefficient is used to analyze the linear or nonlinear relationship between the data at different time points in the local perspective matrix, and the bidirectional long short-term memory network based on the Spearman correlation coefficient analysis result is used to perform time series modeling on the data in the local perspective matrix to obtain a time coordination matrix, which is convenient for identifying the law of change of distorted data over time, such as load fluctuation, power offset, etc., so as to optimize the scheduling of the smart grid; secondly, the cosine similarity is used to analyze the similarity between different spatial nodes in the local perspective matrix, and the bidirectional long short-term memory network based on the similarity is used to perform spatial sequence modeling on the data in the local perspective matrix to obtain a spatial coordination matrix, wherein the spatial coordination matrix can reflect the anomalies of distorted data in the communication process of the smart grid, and through the spatial coordination matrix, the coordination relationship between different monitoring nodes in the target power grid area is analyzed, which is conducive to optimizing the load distribution of the power grid.
[0077] In addition, in the specific implementation, the data points in the local perspective matrix are initially filtered according to the correlation coefficient of the time collaboration matrix, and then the data points in the local perspective matrix are filtered again based on the correlation coefficient of the spatial collaboration matrix to obtain the filtered local perspective matrix. The specific implementation can be implemented in the following way, namely: a visualization tool can be used to construct a line graph of the Spearman correlation coefficient between all the data in the local perspective matrix, and all the valley values in the line graph are filtered, to construct a line graph of the cosine similarity between all the data in the local perspective matrix, and all the valley values in the line graph are filtered again to obtain the filtered local perspective matrix.
[0078] In a specific implementation, the attenuation coefficient of the smart grid during data isolation can be determined according to the filtered local perspective matrix and the local perspective matrix, wherein the attenuation coefficient can be calculated using the following formula:
[0079]
[0080] Among them, L represents the local perspective matrix, L′ represents the filtered local perspective matrix, α represents the attenuation coefficient, and the value range of α is between 0 and 1. If α is closer to 1, it means that the data is over-isolated, which will affect the communication of the smart grid. If α is closer to 0, it means that the data changes less during the isolation process. By calculating the attenuation coefficient α, the loss of data during the isolation process can be quantitatively evaluated, which is convenient for analyzing the stability of communication between different nodes in the smart grid. In addition, ||L|| F represents the Frobenus norm of the local view matrix L, ||LL′|| F The Frobenus norm representing the difference between the local view matrix and the filtered local view matrix, ||L|| F It can be calculated by the square root of the sum of the squares of all the data in the matrix L, that is:
[0081]
[0082] In addition, ||LL′|| F See also ||L|| F The calculation formula is used for calculation, which will not be repeated here, where L(i, j) represents the value of the i-th row and j-th column in the local perspective matrix L.
[0083] It should be noted that in this application, by calculating the attenuation coefficient of the smart grid during data isolation, it is helpful to quantify the degree of information loss in the smart grid during data isolation. The attenuation coefficient can be used to evaluate the impact of each node in the smart grid during data transmission to ensure the integrity of data communication in the smart grid.
[0084] In step S4, historical monitoring data of the target power grid area is obtained, and then a residual repair model of the smart grid during data isolation is constructed based on the historical monitoring data. The residual repair model is used to determine the repair granularity of the power monitoring data based on the offset trend and the attenuation coefficient.
[0085] In this embodiment, historical monitoring data of the target power grid area can be obtained from the dispatching database of the smart grid, wherein the historical monitoring data includes the voltage, current, power, fault occurrence time, fault repair time, load distribution and flow information historically monitored by the dispatching center of the target power grid area. The historical monitoring data provides a comprehensive historical perspective on the operating status and signal changes of the power grid, providing a training basis for subsequent model construction.
[0086] It should be noted that, in this embodiment, the residual repair model is a neural network model constructed based on a multi-view attention mechanism, wherein the residual repair model of the smart grid during data isolation based on the historical monitoring data can be specifically implemented in the following manner, namely: the residual repair model adopts a neural network architecture, combines the advantages of a multi-view attention mechanism, and focuses on capturing the correlation between data from different perspectives. The residual repair model includes: an input layer, an encoder, a multi-view attention module (including a temporal attention mechanism and a spatial attention mechanism), a residual prediction module, and an output layer.
[0087] In a specific implementation, after initializing the residual repair model, the historical monitoring data may be used to train, evaluate and repair the initial residual repair model.
[0088] In this embodiment, the repair granularity of the power monitoring data is determined based on the offset trend and the attenuation coefficient by the residual repair model in the following manner, namely:
[0089] The offset trend and the attenuation coefficient are both used as constraint parameters of the residual repair model;
[0090] The power monitoring data is repaired by a residual repair model with set constraint parameters, and then the repair granularity of the power monitoring data in the repair process is extracted.
[0091] It should be noted that using the offset trend and the attenuation coefficient as constraint parameters of the residual repair model can be used to guide the residual repair model to correct the global offset during the data repair process, ensuring that the repair result remains consistent with the actual system behavior. It can also be used to constrain the intensity adjustment of the model repair signal to avoid over-repair or under-repair. That is, by setting constraint parameters, global directional constraints can be provided for the repair of power financial holding data to improve the credibility of the repair results.
[0092] In specific implementation, the repair granularity of the power monitoring data in the repair process can be extracted through a visualization tool, and the repair granularity can be calculated using the following formula, namely:
[0093]
[0094] Among them, G is the repair granularity, N is the total number of power monitoring data, For the power monitoring data after repair, X t is the power monitoring data before repair, W1 is the constraint weight of the deviation trend, W2 is the constraint weight of the attenuation coefficient, and W1 and W2 can be allocated according to the ratio of the deviation trend and the attenuation coefficient.
[0095] It should be noted that the repair granularity in this application refers to the degree of precise adjustment of power monitoring data. The repair granularity can identify abnormal fluctuation points and determine the severity of signal loss caused by data isolation, which is convenient for evaluating the stability of different operations of smart grids. The repair granularity can also serve as an important basis for smart grid risk assessment, that is, the higher the repair granularity, the lower the risk assessment level of the smart grid, and the lower the repair granularity, the higher the risk assessment level of the smart grid, which is beneficial to improving the response efficiency of risk assessment during smart grid operation.
[0096] In step S5, a confidence assessment is performed on the smart grid based on the repair granularity of the power monitoring data, thereby obtaining a risk assessment result of the target power grid area.
[0097] In this embodiment, the confidence assessment of the smart grid is performed based on the repair granularity of the power monitoring data, and then the risk assessment result of the target power grid area is obtained in the following manner, namely:
[0098] Determining a repair error of the power monitoring data based on a repair granularity of the power monitoring data;
[0099] Determining the fluctuation coefficient and confidence level of the operation state of the smart grid through the repair error;
[0100] A weighted evaluation is performed on the fluctuation coefficient and the confidence level to obtain a risk assessment result of the target power grid area.
[0101] In specific implementation, first, the inverse of the repair granularity of the power monitoring data can be used as the repair error of the power monitoring data, that is, the higher the repair granularity, the lower the repair error, and the lower the repair granularity, the higher the repair error, that is, the repair error is positively correlated with the risk assessment level of the smart grid; then, the mean of the repair errors of the power monitoring data is used as the error mean, and the standard deviation of the repair errors of the power monitoring data is used as the error standard deviation, and the fluctuation coefficient of the operating state of the smart grid is calculated by the ratio of the error standard deviation to the error mean, that is, fluctuation coefficient = 1-(standard deviation / mean), and then the maximum repair error is screened out, and the confidence is calculated by the ratio of the maximum repair error to the error mean, that is, confidence = 1-error mean / maximum repair error; finally, the weighted sum of the fluctuation coefficient and the confidence is used as the risk assessment result of the target power grid area, wherein the weight value in the weighted summation process can be adaptively allocated in combination with the regression analysis model, which is conducive to improving the adaptability of the risk assessment result calculation.
[0102] It should be noted that in this application, the weighted sum of the fluctuation coefficient and the confidence level is used as the risk assessment result of the target power grid area. The fluctuation coefficient and the confidence level can be combined to comprehensively reflect the risk assessment result of the use of the smart grid by the dispatching center in the target power grid area. The higher the risk assessment result, the higher the risk, indicating that there is a greater risk in the communication of the smart grid. The lower the risk assessment result, the lower the risk, indicating that the communication of the smart grid is safe and reliable. By combining the two key indicators of the fluctuation coefficient and the confidence level into a numerical value, it is convenient to adjust the assessment results according to actual conditions, so as to adapt to the operation requirements of smart grids in different regions.
[0103] It can be seen that in this application, firstly, by collecting and screening distorted data, abnormal factors that may affect the accuracy of the evaluation are eliminated, and a purer data set is provided for constructing the matrix. After the distorted data is screened out, constructing a local perspective matrix is conducive to focusing on the impact of a single area or device, and constructing a global perspective matrix comprehensively reflects the operating status of the entire network, forming a multi-dimensional analysis framework, and providing data support for the analysis of the offset trend and the attenuation coefficient; secondly, the data in the local perspective matrix is collaboratively filtered and optimized, interference factors are eliminated, and more accurate regional features are extracted to obtain the attenuation coefficient, which is convenient for quantitatively describing the impact of data isolation on network performance and providing important parameters for residual repair; then, the residual repair model of the smart grid during data isolation is constructed through historical monitoring data. The residual repair model effectively repairs the distorted data by combining historical patterns and real-time characteristics, which is conducive to improving the integrity of the data, reducing the evaluation error caused by data anomalies, and improving the overall reliability; finally, the repair results are quantitatively analyzed through confidence evaluation. The confidence evaluation combines the repair error and the network characteristics of the smart grid, which is convenient for comprehensively measuring the risk status of the power grid area and providing a basis for decision-making, thereby helping to improve the effect of smart grid evaluation.
[0104] In summary, the technical solution adopted in this application can perform residual repair and confidence risk assessment on power monitoring data to improve the reliability of intelligent risk assessment of the power grid.
[0105] In this embodiment, after the confidence assessment of the smart grid is performed based on the repair granularity of the power monitoring data, and then the risk assessment result of the target power grid area is obtained, the resources of the power grid are optimized based on a pre-configured resource allocation optimization model. The resource allocation optimization model is constructed based on a comprehensive loss index and an emergency resource information set, and the comprehensive loss index is determined based on a partition risk assessment model and power grid load data.
[0106] In an optional embodiment of the present application, the process of constructing the partition risk assessment model includes:
[0107] Determine key statistical features in each preset time period, wherein the key statistical features may include statistical features such as variance, average value, difference between maximum and minimum values of various historical power parameter values in the time period, health status of each power grid device, various statistical features of various environmental parameters in the time period, and quantitative values of all geographic information features, such as altitude, slope, land use type, terrain classification index, digital elevation and other information;
[0108] A Bayesian network model is used as a partition risk assessment model, wherein the partition risk assessment model takes key statistical features of each time period as input;
[0109] Use structural learning algorithms to build the structure of the Bayesian network to represent the conditional dependencies between key statistical features and manually add known physical or engineering constraints;
[0110] The partition risk assessment model outputs an estimated probability of failure occurring in each time period according to the key statistical characteristics of the time period; it iteratively optimizes the parameters of each conditional probability table in the Bayesian network by using the maximum likelihood estimation or Bayesian estimation method for each time period;
[0111] In an optional embodiment of the present application, the resource demand ratio Bi is calculated according to the following formula:
[0112]
[0113] Wherein, i represents the serial number mark of each power grid partition; Ri represents the estimated resource demand mark in the i-th power grid partition, and the estimated resource demand Ri=a1×Li+a2×Ni+a3×Ri; wherein a1, a2 and a3 are all preset proportional coefficients; wherein Li represents the real-time power load mark of the i-th power grid partition; Ni represents the number mark of the key power facilities in the i-th power grid partition; FPi represents the failure probability mark of the i-th power grid partition estimated by the partition risk assessment model; Bi represents the resource demand ratio mark in the i-th power grid partition.
[0114] The grid loss index Wherein, a4 and a5 are preset proportional coefficients, and Ltotal is the total load of all grid partitions.
[0115] In an optional embodiment of the present application, the resource allocation optimization model is constructed by minimizing the target optimization function M as the optimization target, and taking the first constraint condition and the second constraint condition as the constraint condition set:
[0116] The expression of the objective optimization function M of the resource allocation optimization model is: M = ∑ i(Si×(1-xi)+Pi×xi); where Pi is the cost of allocating emergency resources to the i-th grid partition; xi represents the allocation ratio variable set for each grid partition; the total amount of emergency resources is 1.
[0117] Embodiment 2: This application provides a power grid risk assessment system, referring to Figure 4 As shown, this figure is a schematic diagram of an evaluation system according to this embodiment of the present application, and the evaluation system includes:
[0118] The acquisition module 100 is used to collect power monitoring data of the target power grid area and determine the network structure of the smart grid used by the dispatching center of the target power grid area;
[0119] An isolation and division module 200, configured to determine distorted data in the power monitoring data through an isolation device in the network structure, and construct a local viewing angle matrix and a global viewing angle matrix based on the distorted data;
[0120] A matrix processing module 300, for determining the deviation trend of the smart grid during communication according to the global perspective matrix, performing collaborative filtering on the local perspective matrix, and then obtaining an attenuation coefficient of the smart grid during data isolation;
[0121] A data repair module 400 is used to obtain historical monitoring data of a target power grid area, and then construct a residual repair model of the smart grid during data isolation based on the historical monitoring data, and determine the repair granularity of the power monitoring data based on the offset trend and the attenuation coefficient through the residual repair model;
[0122] The confidence assessment module 500 is used to perform confidence assessment on the smart grid based on the repair granularity of the power monitoring data, and then obtain a risk assessment result of the target power grid area.
[0123] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. A power grid risk assessment method, characterized in that: The evaluation method comprises the following steps: Collecting power monitoring data of the target power grid area and determining the network structure of the smart grid used by the dispatching center of the target power grid area; Determining distorted data in the power monitoring data through an isolation device in the network structure, and constructing a local viewing angle matrix and a global viewing angle matrix based on the distorted data; Determining the deviation trend of the smart grid during communication according to the global perspective matrix, performing collaborative filtering on the local perspective matrix, and thereby obtaining an attenuation coefficient of the smart grid during data isolation; Acquire historical monitoring data of the target power grid area, and then construct a residual repair model of the smart grid during data isolation based on the historical monitoring data, and determine the repair granularity of the power monitoring data based on the offset trend and the attenuation coefficient through the residual repair model; A confidence assessment is performed on the smart grid based on the repair granularity of the power monitoring data, thereby obtaining a risk assessment result for the target power grid area.
2. A power grid risk assessment method according to claim 1, characterized in that: The power monitoring data of the target power grid area is collected through the data acquisition server.
3. A power grid risk assessment method according to claim 1, characterized in that: Determining the distorted data in the power monitoring data by the isolation device in the network structure specifically includes: respectively collecting power transmission data before and after isolation of the isolation device in the network structure, and then determining the data transmission waveform through the power transmission data before and after isolation of the isolation device; Distorted data in the power monitoring data is filtered out based on the data transmission waveform.
4. A power grid risk assessment method according to claim 1, characterized in that: Constructing a local viewing angle matrix and a global viewing angle matrix based on the distortion data specifically includes: Obtain dispatching flow information of the target power grid area; Performing a fast Fourier transform on the distorted data to obtain a frequency domain space of the distorted data; Based on the scheduling power flow information, spatial features of different power flow nodes are extracted from the frequency domain space, and then a local perspective matrix and a global perspective matrix are constructed according to all the spatial features.
5. A power grid risk assessment method according to claim 1, characterized in that: Determining the deviation trend of the smart grid during communication according to the global perspective matrix specifically includes: Mapping the data points in the global perspective matrix to a geometric space; The data fluctuation curve is determined by the Euclidean distance of all data points in the geometric space, and then the trend of the data fluctuation curve is used as the deviation trend of the smart grid during communication.
6. A power grid risk assessment method according to claim 1, characterized in that: Performing collaborative filtering on the local perspective matrix to obtain the attenuation coefficient of the smart grid during data isolation specifically includes: Performing time-space collaborative division on the local perspective matrix to obtain a time collaborative matrix and a space collaborative matrix; Performing a primary filtering on the data points in the local perspective matrix according to the correlation coefficient of the time synergy matrix, and then performing a secondary filtering on the data points in the local perspective matrix based on the correlation coefficient of the space synergy matrix to obtain a filtered local perspective matrix; The attenuation coefficient of the smart grid during data isolation is determined according to the filtered local viewing angle matrix and the local viewing angle matrix.
7. A power grid risk assessment method according to claim 1, characterized in that: Obtain historical monitoring data of the target grid area from the dispatching database of the smart grid.
8. A power grid risk assessment method according to claim 1, characterized in that: The residual repair model is a neural network model built based on a multi-view attention mechanism.
9. A power grid risk assessment method according to claim 1, characterized in that: The confidence assessment of the smart grid is performed based on the repair granularity of the power monitoring data, and the risk assessment results of the target power grid area are obtained, which specifically include: Determining a repair error of the power monitoring data based on a repair granularity of the power monitoring data; Determining the fluctuation coefficient and confidence level of the operation state of the smart grid through the repair error; A weighted evaluation is performed on the fluctuation coefficient and the confidence level to obtain a risk assessment result of the target power grid area.
10. A power grid risk assessment system, used to execute a power grid risk assessment method according to any one of claims 1 to 9, characterized in that: The evaluation system comprises: A collection module, used to collect power monitoring data of a target power grid area and determine the network structure of a dispatch center in the target power grid area using a smart grid; An isolation and division module, used to determine the distorted data in the power monitoring data through the isolation device in the network structure, and construct a local viewing angle matrix and a global viewing angle matrix based on the distorted data; A matrix processing module, used to determine the deviation trend of the smart grid during communication according to the global perspective matrix, perform collaborative filtering on the local perspective matrix, and then obtain the attenuation coefficient of the smart grid during data isolation; A data repair module, used to obtain historical monitoring data of a target power grid area, and then construct a residual repair model of the smart grid during data isolation based on the historical monitoring data, and determine the repair granularity of the power monitoring data based on the offset trend and the attenuation coefficient through the residual repair model; The confidence assessment module is used to perform confidence assessment on the smart grid based on the repair granularity of the power monitoring data, and then obtain a risk assessment result of the target power grid area.