A data representation method for emergency repair sites in monitoring areas

Abstract

The present invention belongs to the field of power grid on-site monitoring, and specifically relates to a data characterization method for emergency repair sites in monitoring areas. It includes: monitoring data collection; dividing data into two categories: structured data represented by attribute data and unstructured data represented by image data; using different data characterization strategies to characterize the two different types of monitoring data, standardizing the characterized rough data, and then using an object-oriented metamodel to reprocess, model and store the data; this characterization method effectively avoids data distortion, and data processing takes less time, making it easy to use a big data analysis model based on cloud-edge collaboration to perform different data analyses on stored data, and obtain conclusions such as early warning results and emergency repair decisions.

Description

Technical Field

[0001] The invention belongs to the field of power grid on-site monitoring, and specifically discloses a data characterization method for an emergency repair site in a monitoring area. Background Art

[0002] Disaster site monitoring data based on mobile Internet, including power grid and equipment operation data obtained by fixed-point monitoring, individual operation tools, intelligent positioning and other technical means, is an important disaster prediction parameter and one of the important data sources for disaster early warning and improving emergency repair efficiency. Correlation analysis and in-depth mining of such data to improve its real-time monitoring capability of the power grid is a key supporting technology to ensure the safe and stable operation of the power grid, and is of great significance to improving the overall disaster handling efficiency and control capabilities of the power grid.

[0003] At present, in the power grid monitoring system, there are many monitoring devices distributed in different locations, systems and applications. The monitoring data sources are different, the formats are diverse, the storage mechanism is not unified, the association method is unclear, and there is a lack of effective sharing mechanism, resulting in low data utilization and the formation of information islands. Traditional field monitoring data collection technology requires a lot of manual data sorting, manual extraction of valid data, and integration and grouping before further utilization and mining analysis can be carried out. Even intelligent technology has problems such as low data representation efficiency, inaccurate feature extraction, and high data redundancy, and has not truly achieved unified representation of multi-source data. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art in monitoring data representation, the present invention provides a data representation method for an emergency repair site in a monitoring area, comprising the following steps:

[0005] S1, monitoring data collection;

[0006] S2, divides the data into two categories: structured data represented by attribute data and unstructured data represented by image data;

[0007] S3, standardizing the data;

[0008] S4, using a principal component analysis model to perform dimensionality reduction processing on the structured data to complete the structured data representation;

[0009] S5, performing artificial intelligence learning on the unstructured data using an independent component analysis model to obtain a training basis function;

[0010] S6, obtaining characteristic coefficients of the unstructured data by means of basis function response;

[0011] This method classifies structured data and unstructured data and models them in a way that matches the data type, reducing data redundancy while making the data better reflect the information characteristics of the monitoring site.

[0012] Preferably, in step S3, the continuous attribute data in the structured data is mapped to corresponding values ​​in the interval [0, 1] using a min-max standardization method, and the discrete attribute data is retained in its original state;

[0013] Furthermore, in step S4, the principal component analysis model is used to perform dimensionality reduction processing on the structured data, including the following steps:

[0014] The original sample data is organized into an m-row and n-column matrix X0, where m represents the row number of a record and n represents the column number of the corresponding attribute feature;

[0015] Calculate the transposed matrix X of the matrix X0 n*m , for standardization;

[0016] Calculate the covariance matrix R of the two sets of sample data and find its eigenvalues ​​and corresponding eigenvectors;

[0017] Arrange the eigenvectors into a matrix from top to bottom according to the corresponding eigenvalues, and take the first k rows to form the eigenvector matrix P k*n ;

[0018] Calculate the data matrix Y after reducing to k dimensions k*m =P k*n X n*m , calculate the corresponding transposed matrix Y0, and obtain the dimensionality reduction processing result of the data;

[0019] Preferably, in step S3, the unstructured data is whitened and centered to obtain sample set data in a unified format.

[0020] Further, step S6 comprises the following steps:

[0021] The independent component analysis model is used to calculate the characteristic response coefficient of the basis function to the input sample X.

[0022] The characteristic response coefficient is screened by setting a threshold value to obtain a characteristic coefficient.

[0023] Compared with the prior art, the present invention has the following beneficial effects: the data characterization method of the present invention realizes the centralized processing, effective association and mining analysis of multi-source composite data, improves the utilization rate of emergency repair site data, and enables the corrected data to accurately characterize the status of the monitoring site, which is conducive to the unified coordination and analysis of big data at the disaster site, and improves the efficiency of disaster site repair. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a flow chart of the emergency repair site monitoring data characterization method in an embodiment of the present invention.

[0025] Figure 2 It is a data characterization flow chart of structured data in an embodiment of the present invention.

[0026] Figure 3 It is a flow chart of structured data dimensionality reduction in an embodiment of the present invention.

[0027] Figure 4 It is a data characterization flow chart for unstructured data in an embodiment of the present invention.

[0028] Figure 5 It is an object-oriented data association analysis model diagram in an embodiment of the present invention. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be described below in conjunction with the drawings in the embodiments of the present invention. The described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making innovative efforts are within the scope of protection of the present invention.

[0030] See the instruction manual Figure 1-5 , this embodiment is completed according to the following steps:

[0031] The overall technical process of the present invention is as follows Figure 1 As shown, that is: by collecting monitoring data of various types of information such as disasters, people, objects, and equipment at the emergency repair site, using the unified modeling, storage, and analysis framework of emergency repair site data, the data is stored in a unified manner, the efficiency of information transmission is improved, the quality of data monitoring is improved, and the emergency site command is further effectively guided and the efficiency of disaster repair is improved. Among them, the technical difficulties are concentrated on key technologies such as unified representation of data and standardized storage of data. This is achieved specifically through the following methods:

[0032] Collecting emergency repair site monitoring data sets includes collecting various meteorological data, disaster data, typical disaster cases, equipment record data, fault monitoring data, online monitoring of transmission lines, etc. from the power grid company in recent years, forming a monitoring data set centered on disasters, people, power equipment, vehicles, materials, etc. At the same time, the original storage method and expression method of these data are collected to understand the data type, original number and basic characteristics of emergency disaster site data, as monitoring data support for the unified modeling architecture.

[0033] Building a unified representation of emergency repair site data includes the following steps:

[0034] First, the collected data is classified and organized to distinguish between two major categories: structured attribute data characterized by discrete data and continuous data, and unstructured data such as graphs, pictures and videos. This allows for further storage and analysis based on data categories using different information processing methods, making the data more shareable and spreadable.

[0035] For attribute data represented by weather, temperature, humidity, equipment status (normal / abnormal), number of fault records, disaster frequency, disaster type, monitoring capability, emergency response capability, etc., a data characterization method based on principal component analysis (PCA) is used for data characterization. Specifically, the following steps are included:

[0036] Clean the collected data to remove duplicate data or some duplicate attribute features;

[0037] The missing values ​​are filled by using the same mean interpolation method, and the data matrix X0 = {x ij}(i∈[1,m],j∈[1,n]), where row m represents a record and column n represents an attribute feature;

[0038] According to different data types, attribute data can be divided into continuous data and discrete data. For continuous data, such as air humidity, average wind speed, rainfall, etc., the monitoring value is the encoded feature value; for discrete attribute data, such as topography, whether the line is reinforced, voltage level, etc., unique hot encoding is used for feature encoding. For example, whether the line is reinforced can be encoded in 0-1 encoding, where 0 means the line is not reinforced and 1 means the line is reinforced;

[0039] Data standardization. After feature encoding, the attribute data sample matrix X0 = {x ij}(i∈[1,m],j∈[1,n]), each sample data xij is mapped to the interval [0,1] by the min-max normalization method to obtain the mapping value.

[0040] Data dimensionality reduction: The principal component analysis method PCA is used to perform feature dimensionality reduction on the data sample X0 collected from each monitored object to obtain the reduced-dimensional data matrix Y0.

[0041] For image data represented by infrared and surveillance cameras, the data characterization method based on the independent component analysis method ICA method is used to characterize the image data. Specifically, the following steps are included:

[0042] Build an image library using the collected data.

[0043] Directly sample the picture data, preprocess the video data to obtain the image data, and finally form an image data sample set;

[0044] Perform whitening and centering processing on the image samples to obtain sample set data in a unified format, which will be used as the input data X of the subsequent model;

[0045] Select the objective function. According to the definition of the ICA analysis model, the input data X = AS, where A is the image feature vector, which is a reversible mixing matrix, and S is the image independent hidden random variable. The random variable S = WX can be calculated, where W = A-1, which is the objective function estimated by the ICA model;

[0046] Determine the non-Gaussianity of variable S. In this patent, the negative entropy of random variable S is used to determine its non-Gaussianity. In information theory, entropy is used to describe the degree of information contained in a random variable. If the randomness is stronger, the entropy is greater. The Gaussian variable has the largest entropy value among variables with the same variance. Therefore, negative entropy can be used to measure the non-Gaussianity of a random variable.

[0047] If the variable S has the strongest non-Gaussianity, then its corresponding basis function is the best basis function. Otherwise, the transformation matrix A is modified and the non-Gaussianity judgment is repeated until the best basis function is obtained and output.

[0048] Calculate the basis function response. Calculate the basis function response coefficient S to the input sample X according to the formula S = WX;

[0049] Screen the characteristic response coefficient S. Select a suitable threshold to screen the characteristic response coefficient, set the coefficients less than the threshold to 0, and obtain a new characteristic response coefficient S' as the main feature of the image data;

[0050] Characterize the image. Use the new characteristic response coefficient S' and the formula X = AS = W-1S to obtain the estimated value X' of the image data, which is used to characterize the image and obtain the characterized image data set.

[0051] The object-oriented data association analysis model is used to model and store monitoring data to realize the organization and management of monitoring data. By taking the monitoring object ID as the center, studying the relationship between different dimensions, and strengthening the association between monitoring data and objects on this basis, the data is stored, such as Figure 5 As shown, this is used to identify and store emergency repair site monitoring data.

[0052] After the data is represented and stored in a unified manner, the monitoring data is further analyzed and processed using a big data analysis model based on edge-cloud collaboration to obtain early warning results and guide on-site decision-making.

[0053] In summary, constructing a unified representation modeling method suitable for emergency repair site data in the monitoring area can enable unified storage and analysis and processing of on-site monitoring data, accelerate the efficiency of intelligent decision-making, and assist in the generation of emergency repair site command strategies in real time, thereby realizing the standardization and intelligence of disaster emergency repair.

[0054] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A data characterization method for monitoring emergency repair sites in an area, characterized in that The following steps are involved: S1, monitoring data collection; S2, divides the data into two categories: structured data represented by attribute data and unstructured data represented by image data; S3, standardizing the data; S4, using a principal component analysis model to perform dimensionality reduction processing on the structured data to complete the structured data representation; S5, performing artificial intelligence learning on the unstructured data using an independent component analysis model to obtain a training basis function; S6, obtaining characteristic coefficients of the unstructured data by means of basis function response; S7, substituting the characteristic coefficient into the training basis function to calculate an estimated value to characterize the unstructured data; In step S3, the continuous attribute data in the structured data is mapped to the corresponding value in the interval [0,1] using the min-max normalization method, and the discrete attribute data is retained in the original state; the unstructured data is whitened and centered to obtain sample set data with a unified format; In step S4, the principal component analysis model is used to perform dimensionality reduction processing on the structured data, including the following steps: The original sample data is organized into an m-row and n-column matrix X0, where m represents the row number of a record and n represents the column number of the corresponding attribute feature; Calculate the transposed matrix X of the matrix X0 n*m , for standardization; Calculate the covariance matrix R of the two sets of sample data and find its eigenvalues ​​and corresponding eigenvectors; Arrange the eigenvectors into a matrix from top to bottom according to the corresponding eigenvalues, and take the first k rows to form the eigenvector matrix P k*n ; Calculate the data matrix Y after reducing to k dimensions k*m =P k*n X n*m , calculate the corresponding transposed matrix Y0, and obtain the dimensionality reduction processing result of the data; Step S6 comprises the following steps: An independent component analysis model is used to calculate the characteristic response coefficient of the basis function to the input sample X; the characteristic response coefficient is screened by setting a threshold to obtain a characteristic coefficient.

Images