A method and system for collecting big data information

Abstract

The present invention relates to the field of data processing technology, specifically to a method and system for collecting big data information, S1, collecting real-time electricity consumption data information of residents through multiple front-end data collectors, S2, processing the collected electricity consumption data information through the front-end central processing unit and the communication algorithm, and then storing it in a distributed database through the communication network. In the present invention, the communication control unit controls and manages the data transmission buffer of different data channels through the active queue management algorithm, realizes the stability of the operation of the electricity consumption data information transmission network, improves the accuracy and speed of electricity consumption information collection, improves the service communication quality and the stability of the electricity consumption data information management system, and completes the electricity consumption information fusion result through the data processing unit, better solves the problem of electricity consumption information collection and processing when there are multiple source faults in the power grid, and can facilitate the subsequent reliable prediction of electricity demand and identification of abnormal electricity consumption behavior.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a big data information collection method and system. Background Art

[0002] With the development of communication technology, computer network technology and big data industry, the scale of smart grid equipment has grown rapidly. The electricity information collection system consists of a system master station, transmission channels, collection terminals and smart meters. As the scale of smart grids expands and the output load increases, power grid failures are inevitable during operation. The load distribution and line loss in the power system will change, which will have a great impact on the judgment of electricity information, making it impossible for the power system to accurately grasp electricity demand, resulting in a reduction in the rationality of electricity planning. Therefore, the key to terminal electricity planning in the power system lies in effectively collecting and managing electricity information under power grid failures. The power grid's electricity information is of great significance for power grid planning, electricity theft monitoring, and electricity demand forecasting.

[0003] However, the hardware of the traditional electricity consumption information data collection system has the problems of low collection accuracy and speed, and low computing efficiency. At the same time, due to the complex structure of the power system, the operating data has the dynamic characteristics of change, and the information generated by the power system under fault conditions has a certain degree of randomness and variability, resulting in poor integration of electricity consumption information. At present, there are still some unscrupulous operators who will steal electricity by any means. The theft of electricity not only damages the economic interests of power companies, but also endangers the safe operation of the power grid and hinders the development of the power industry. The existing anti-theft methods are highly dependent on people, and the accuracy and efficiency are difficult to meet the requirements of anti-theft work. The collection of electricity information data is also particularly important for judging suspected users of electricity theft.

[0004] In view of this, the present application proposes a method and system for collecting big data information. Summary of the invention

[0005] The purpose of the present invention is to provide a big data information collection method and system to solve the above problems.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for collecting big data information comprises the following steps:

[0008] S1. Collecting real-time electricity consumption data information of residents through multiple front-end data collectors;

[0009] S2, processing the collected power consumption data information through the front-end central processor and communication algorithm, and then storing it in a distributed database through the communication network;

[0010] S3, performing data preprocessing, multi-source data information classification, multi-source data feature extraction and information fusion processing on the electricity consumption data information in the distributed database;

[0011] S4. By constructing different electricity theft diagnosis and analysis models, the analysis and identification of suspected electricity theft users can be realized;

[0012] S5. Screen out the areas of suspected electricity theft users by means of target area screening;

[0013] S6. Construct a key user analysis model and use the hierarchical analysis method to analyze and identify suspected electricity theft users.

[0014] Furthermore, in step S3, the multi-source data information classification model is used to classify the power consumption data information in the distributed database, including the following steps:

[0015] A1. The multi-source electricity consumption information classification model is based on the original electricity consumption data after data processing. After completing the preliminary screening of the data, the small sample data is sampled and replicated;

[0016] A2. The multi-source electricity consumption information classification model performs decoupling and encoding processing on small sample data through the deep belief network DBM to establish a coded data set;

[0017] A3. After the new data is copied and encoded through the DBM network, the new data is vertically spliced ​​with the original data to obtain the expanded data of electricity consumption.

[0018] A4. Train the extreme learning machine network based on the data to achieve classification and processing of electricity consumption data and form a static data set and dynamic datasets .

[0019] Furthermore, in step S3, the multi-source data feature extraction of the electricity consumption data information in the distributed database includes the following steps:

[0020] B1. Feature extraction through data-driven mutual information principle, static data set Dynamic Datasets The mutual information calculation formula between is: ;

[0021] Where: for and The joint probability density function between and Are marginal probability density functions, corresponding to and ;

[0022] B2. Extraction based on mutual information and The feature quantity of the extraction result is: ;

[0023] Where: is the initial feature quantity, is the feature quantity with dimension m;

[0024] B3. Based on the above extraction result expression, the maximum mutual information result can be obtained, and feature selection is completed on the basis of ensuring that the category does not change;

[0025] B4. Use decomposition to process multi-dimensional mutual information to form one-dimensional mutual information. The calculation formula is: ;

[0026] Where: and for and Selected features of the dataset;

[0027] B5. Through steps B1-B4, the power consumption information feature set is quickly extracted;

[0028] B6, initial variable selection and second variable selection are: ;

[0029] B7. According to the expression in step B6, the selection of the remaining variables can be completed. Each time, the next feature is selected based on the principle of maximum mutual information. When making the selection, a threshold N is set. When the number of cycles reaches the threshold, the selection stops.

[0030] Furthermore, in step S5, the target substation screening method includes user change monitoring, file verification, abnormal monitoring analysis and abnormal indicator analysis of the substation. The file verification includes multiple assessment metering points, photovoltaic metering point main use errors and two-way metering abnormality verification. User change monitoring mainly screens substations that have not changed in the past week. Abnormal indicator analysis includes line loss rate abnormality, daily fluctuation of line loss rate, three-phase imbalance and power factor abnormality.

[0031] Furthermore, in step S6, the key user analysis model includes a power consumption comparison analysis model, a power consumption and line loss rate correlation analysis model, a zero-live current imbalance analysis model and a shunt analysis model. The hierarchical analysis method is used to score the suspected electricity theft users obtained from different models, which are divided into primary and secondary scores, and the suspected electricity theft users are divided into high-suspicion users and general suspected users according to different scores.

[0032] A big data information collection system, the system comprising a data collection unit, a communication control unit, a distributed data storage unit, a data processing unit, a unit for establishing a power theft diagnosis and analysis model, a target area screening unit and a scoring prediction unit; the data collection unit collects real-time power consumption data information of residents through multiple front-end data collectors, and stores it in the distributed data storage unit after being processed by the front-end central processing and communication algorithm, and is visually displayed on the system terminal at the same time;

[0033] The communication control unit is connected to the front data collector through the communication network to obtain the power consumption information data, and through the active queue management algorithm, namely the AQM algorithm, different data channels are controlled and managed to communicate and buffer the data;

[0034] The distributed data storage unit is connected to the communication control unit and is used to store the collected real-time residential electricity consumption data information in a distributed database;

[0035] The data processing unit is used to perform data preprocessing, multi-source data information classification, multi-source data feature extraction and information fusion processing on the power consumption data information in the distributed database;

[0036] The electricity theft diagnosis and analysis model establishment unit summarizes different electricity theft characteristics through big data analysis, constructs different electricity theft diagnosis and analysis models, and realizes analysis and identification of suspected electricity theft users;

[0037] The target area screening unit is used to monitor user changes, conduct file verification, perform abnormal monitoring and analysis, and perform abnormal index analysis on the area, and screen out areas where users suspected of electricity theft are found;

[0038] The score prediction unit constructs a key user analysis model and uses the hierarchical analysis method to score key users suspected of electricity theft in the substation area, find out the suspected electricity theft users, and realize the analysis and identification of the suspected electricity theft users.

[0039] Further, the data acquisition unit includes a data acquisition module, a concentrator module and a visualization module;

[0040] The data acquisition module receives data request instructions through the data port of the initialized front data collector to realize the real-time electricity consumption data information collection of residents;

[0041] The concentrator module is used to issue different data request instructions under the control of the system terminal, and to collect and centrally control the data of multiple front-end data collectors;

[0042] The visualization module displays data through the system terminal to achieve real-time data monitoring and massive data query functions.

[0043] Furthermore, the data processing unit includes a data preprocessing module, a multi-source data information classification module, a multi-source data feature extraction module and an information fusion processing module;

[0044] The data preprocessing module is used to preprocess the real-time electricity consumption data information of residents in the distributed database to realize real-time data stream processing, complex event stream processing, massive data offline processing and data mining functions;

[0045] The multi-source data information classification module constructs a multi-source data information classification model and completes the classification of multi-source electricity consumption information through the multi-source data information classification model;

[0046] The multi-source data feature extraction module completes feature extraction through the data-driven mutual information principle to obtain a feature set;

[0047] The information fusion processing module is used to summarize the extracted feature sets so that they are all located in the source node, and to process the redundant features in the feature set through the Kalman filter algorithm KF to complete the consistency fusion of the electricity consumption data.

[0048] Furthermore, the multi-source data information classification module completes the classification of multi-source electricity information through a multi-source data information classification model, including the following steps:

[0049] C1. Assume the original data is , the electricity consumption data collected by the i-th collection point is , the correlation coefficient matrix between the data rows is ,The number of visible layer nodes of the deep belief network DBM is 20;

[0050] C2. Yes After obtaining a small sample of data, the obtained sample data is filtered and normalized. express, and After training and learning through the deep belief network DBM, the outputs are and ;

[0051] C3, and They have the same dimension and low coupling degree. The two are processed by vertical splicing to obtain The sampling data results of ,and The expression is as follows: ;

[0052] C4. Based on the above data, the extreme learning machine network is trained to classify electricity consumption data and form a static data set. and dynamic datasets .

[0053] 1. Beneficial effects of the present invention: In the present invention, after real-time electricity consumption data information of residents is collected by multiple front-end data collectors, it is stored in a distributed database by a communication control unit. The communication control unit controls and manages data transmission buffers through different data channels through an active queue management algorithm, thereby achieving stable operation of the electricity consumption data information transmission network. The active queue management algorithm is more sensitive to fluctuating data and has a wider screening range. It controls the fluctuation of electricity consumption data information management data within a certain range to avoid affecting the entire electricity consumption data information transmission network, improves the accuracy and speed of electricity consumption information collection, improves the service communication quality and the stability of the electricity consumption data information management system, and realizes the reasonable allocation of communication resources and improves the efficiency of power data communication.

[0054] 2. In the present invention, by setting a data processing unit, after classifying and processing multi-source electricity consumption information, dynamic data and static data therein are obtained, feature sets of different categories of electricity consumption data are extracted through data drive, and redundant features in the feature set are removed by using the Kalman filter algorithm to complete the consistent feature description of the electricity consumption information, thereby obtaining the electricity consumption information fusion result, which can complete the statistics of the electricity consumption of end users, better solve the problem of electricity consumption information collection and processing when the power grid has multi-source faults, and at the same time, the electricity consumption information fusion result obtained by using this technology can facilitate the subsequent reliable prediction of electricity demand and identification of abnormal electricity consumption behavior;

[0055] 3. In the present invention, based on the collected and processed electricity information data, the target area screening method is used to screen out the areas where suspected electricity theft users exist. At the same time, for the areas where suspected electricity theft users exist, multiple key user analysis models are used to find out the suspected electricity theft users through the analytic hierarchy process scoring method, so as to realize the analysis and identification of the suspected electricity theft users. According to the model scoring rules, the suspected users are divided into two levels, high and general, according to the scoring range, to provide electricity inspection personnel with high-risk suspected electricity theft target users, thereby improving the quality and efficiency of anti-theft inspection work. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 A method flow chart of a big data information collection method of the present invention;

[0057] Figure 2 This is a system principle diagram of a big data information collection system of the present invention;

[0058] Figure 3 This is a flow chart of an active queue management algorithm for a big data information collection system of the present invention.

[0059] In the figure: 1. Data acquisition unit; 2. Communication control unit; 3. Distributed data storage unit; 4. Data processing unit; 5. Unit for establishing electricity theft diagnosis and analysis model; 6. Target area screening unit; 7. Score prediction unit; 11. Data acquisition module; 12. Concentrator module; 13. Visualization module; 41. Data preprocessing module; 42. Multi-source data information classification module; 43. Multi-source data feature extraction module; 44. Information fusion processing module. DETAILED DESCRIPTION

[0060] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Embodiment 1

[0061] See also Figure 1 and Figure 3 ,This design proposes an implementation method, a big data information collection method, including the following steps:

[0062] Step 1: Collecting real-time electricity consumption data information of residents through multiple front-end data collectors;

[0063] Step 2: Process the collected power consumption data information through the front-end central processor and communication algorithm, and then store it in a distributed database through the communication network;

[0064] Step 3: Perform data preprocessing, multi-source data information classification, multi-source data feature extraction and information fusion processing on the electricity consumption data information in the distributed database;

[0065] The multi-source data information classification model is used to classify the power consumption data information in the distributed database, including the following steps:

[0066] A1. The multi-source electricity consumption information classification model is based on the original electricity consumption data after data processing. After completing the preliminary screening of the data, the small sample data is sampled and replicated;

[0067] A2. The multi-source electricity consumption information classification model performs decoupling and encoding processing on small sample data through the deep belief network DBM to establish a coded data set;

[0068] A3. After the new data is copied and encoded through the DBM network, the new data is vertically spliced ​​with the original data to obtain the expanded data of electricity consumption.

[0069] A4. Train the extreme learning machine network based on the data to achieve classification and processing of electricity consumption data and form a static data set and dynamic datasets ; Among them, extracting multi-source data features from the electricity consumption data information in the distributed database includes the following steps:

[0070] B1. Feature extraction through data-driven mutual information principle, static data set and dynamic datasets The mutual information calculation formula between is: ;

[0071] Where: for and The joint probability density function between and Are marginal probability density functions, corresponding to and ;

[0072] B2. Extraction based on mutual information and The feature quantity of the extraction result is: ; Where: is the initial feature quantity, is the feature quantity with dimension m;

[0073] B3. Based on the above extraction result expression, the maximum mutual information result can be obtained, and feature selection is completed on the basis of ensuring that the category does not change;

[0074] B4. Use decomposition to process multi-dimensional mutual information to form one-dimensional mutual information. The calculation formula is: ; Where: and for and Selected features of the dataset;

[0075] B5. Through steps B1-B4, the power consumption information feature set is quickly extracted;

[0076] B6, initial variable selection and second variable selection are: ; B7, according to the expression in step B6, the selection of the remaining variables can be completed. Each time, the next feature is selected based on the principle of maximum mutual information. When making the selection, a threshold N is set. When the number of cycles reaches the threshold, the selection stops;

[0077] Step 4: Analyze and identify suspected electricity theft users by building different electricity theft diagnosis and analysis models;

[0078] Step 5: Screen out the areas of suspected electricity theft users by screening target areas;

[0079] Among them, the target area screening method includes user change monitoring, file verification, abnormal monitoring analysis and abnormal indicator analysis of the area. The file verification includes multiple assessment metering points, main purpose errors of photovoltaic metering points and bidirectional metering abnormality verification. User change monitoring mainly screens the areas that have not changed in the past week. Abnormal indicator analysis includes abnormal line loss rate, daily fluctuation of line loss rate, three-phase imbalance and power factor abnormality.

[0080] Step 6. Construct a key user analysis model and use the hierarchical analysis method to score the suspected electricity theft users. The key user analysis model includes a power consumption comparison analysis model, a power consumption and line loss rate correlation analysis model, a zero-live current imbalance analysis model and a shunt analysis model. The hierarchical analysis method is used to score the suspected electricity theft users obtained from different models, which are divided into primary and secondary scores, and the suspected electricity theft users are divided into high-suspicion users and general suspected users according to different scores.

[0081] In this embodiment, after multiple front-end data collectors collect real-time electricity consumption data information of residents, the front-end central processor processes the collected data and stores it in the distributed database of the distributed data storage unit 3 through the communication network of the communication control unit 2; Figure 3 As shown, the communication control unit 2 is connected to the front-end data collector through the communication network to obtain the power consumption information data, and through the active queue management algorithm, namely the AQM algorithm, different data channels are controlled and managed to communicate and buffer, thereby realizing the stability of the power consumption data information transmission network operation. The active queue management algorithm, namely the AQM algorithm, is more sensitive to fluctuating data and has a wider screening range, and the fluctuation of the power consumption data information management data is controlled within a certain range to avoid affecting the entire power consumption data information transmission network, thereby improving the accuracy and speed of power consumption information collection, improving the business communication quality and the stability of the power consumption data information management system, and realizing the reasonable allocation of communication resources and improving the efficiency of power data communication;

[0082] The multi-source data information classification model of the multi-source data information classification module 42 is used to classify the power consumption data information in the distributed database to form a static data set. and dynamic datasets The multi-source data feature extraction module 43 then completes the feature extraction through the data-driven mutual information principle. The feature information of the feature set obtained through this technology is complete, and the category of the feature will not change. At the same time, the information fusion processing module 44 is used to summarize the extracted feature sets so that they are all located in the source node, and the redundant features in the feature set are processed through the Kalman filter algorithm KF to complete the consistent fusion of electricity consumption data and ensure a better electricity consumption information fusion effect. Embodiment 2

[0083] See also Figure 1 - Figure 3 ,This design proposes an implementation method, a big data information collection system, including a data collection unit 1, a communication control unit 2, a distributed data storage unit 3, a data processing unit 4, a power theft diagnosis and analysis model establishment unit 5, a target area screening unit 6 and a score prediction unit 7;

[0084] The data collection unit 1 collects the real-time electricity consumption data information of residents through multiple front-end data collectors, and stores it in the distributed data storage unit 3 after processing by the front-end central processing and communication algorithm, and displays it visually at the system terminal at the same time; the communication control unit 2 is connected to the front-end data collector through the communication network to obtain the electricity consumption information data, and through the active queue management algorithm, namely the AQM algorithm, different data channels are controlled and managed to communicate and buffer the data;

[0085] The distributed data storage unit 3 is connected to the communication control unit 2 and is used to store the collected real-time electricity consumption data information of residents in a distributed database;

[0086] The data processing unit 4 is used to perform data preprocessing, multi-source data information classification, multi-source data feature extraction and information fusion processing on the power consumption data information in the distributed database;

[0087] The electricity theft diagnosis and analysis model building unit 5 summarizes different electricity theft characteristics through big data analysis, builds different electricity theft diagnosis and analysis models, and realizes the analysis and identification of suspected electricity theft users;

[0088] The target area screening unit 6 is used to monitor the user changes, check the files, perform abnormal monitoring and analysis, and analyze abnormal indicators in the area, and screen out the areas of suspected electricity theft users;

[0089] The score prediction unit 7 constructs a key user analysis model and uses the hierarchical analysis method to score key users suspected of electricity theft in the substation area, find out the suspected electricity theft users, and realize the analysis and identification of the suspected electricity theft users.

[0090] In this embodiment, preferably, the data acquisition unit 1 includes a data acquisition module 11, a concentrator module 12 and a visualization module 13; the data acquisition module 11 receives data request instructions through the data port of the initialized front data collector to realize real-time electricity consumption data information collection of residents; the concentrator module 12 is used to issue different data request instructions under the control of the system terminal, and at the same time aggregate and centrally control the data of multiple front data collectors; the visualization module 13 displays data through the system terminal to realize real-time data monitoring and massive data query functions.

[0091] In this embodiment, preferably, the data processing unit 4 includes a data preprocessing module 41, a multi-source data information classification module 42, a multi-source data feature extraction module 43 and an information fusion processing module 44; the data preprocessing module 41 is used to preprocess the real-time electricity consumption data information of residents in the distributed database to realize real-time data stream processing, complex event stream processing, massive data offline processing and data mining functions; the multi-source data information classification module 42 constructs a multi-source data information classification model and completes the classification of multi-source electricity consumption information through the multi-source data information classification model; the multi-source data feature extraction module 43 completes feature extraction through the data-driven mutual information principle to obtain a feature set; the information fusion processing module 44 is used to summarize the extracted feature set so that they are all located in the source node, and the redundant features in the feature set are processed through the Kalman filter algorithm KF to complete the consistency fusion of the electricity consumption data;

[0092] The multi-source data information classification module 42 completes the classification of multi-source electricity information through a multi-source data information classification model, including the following steps:

[0093] C1. Assume the original data is , the electricity consumption data collected by the i-th collection point is , the correlation coefficient matrix between the data rows is The number of visible layer nodes of the deep belief network DBM is 20;

[0094] C2. Yes After obtaining the small sample data, the obtained sample data is used for screening and normalization. express, and After training and learning through the deep belief network DBM, the outputs are and ;

[0095] C3, and They have the same dimension and low coupling degree. The two are processed by vertical splicing to obtain The sampling data results of ,and The expression is as follows:

[0096] ;

[0097] C4. Based on the above data, the extreme learning machine network is trained to classify electricity consumption data and form a static data set. and dynamic datasets .

[0098] In this embodiment, based on the collected and processed electricity information data, the target area screening method is used to screen out the areas where suspected electricity theft users exist; at the same time, for the areas where suspected electricity theft users exist, multiple key user analysis models are used to find out the suspected electricity theft users through the hierarchical analysis method scoring method, so as to realize the analysis and identification of the suspected electricity theft users (the key user analysis model mainly screens out abnormal users based on the user's mutation value compared with the historical electricity consumption, the correlation between electricity consumption and line loss rate, the zero-live line current imbalance, and the shunt analysis. The hierarchical analysis method is used to score the suspected users obtained by different models, which are divided into primary and secondary points, and the suspected electricity theft users are divided into high-suspicion users and general-suspicion users according to different scores). According to the model scoring rules, the suspected users are divided into high and general grades according to the scoring range, providing electricity inspection personnel with high-risk suspected electricity theft target users, thereby improving the quality and efficiency of anti-theft inspection work.

[0099] In the present invention, when collecting power grid data, firstly, various data ports of multiple front-end data collectors in the data collection unit 1 are initialized, and different data request commands are issued through the system terminal setting concentrator, and the multiple front-end data collectors receive the issued data request instructions. After the real-time electricity consumption data information of residents' electricity consumption is collected by multiple front-end data collectors, the front-end central processor processes the collected data and stores it in the distributed database of the distributed data storage unit 3 through the communication network of the communication control unit 2; the communication control unit 2 controls the data transmission buffer of different data channels through the active queue management algorithm, thereby realizing the stability of the operation of the electricity consumption data information transmission network. The active queue management algorithm is more sensitive to fluctuating data and has a wider screening range. The fluctuation of the electricity consumption data information management data is controlled within a certain range to avoid affecting the entire electricity consumption data information transmission network, thereby improving the accuracy and speed of electricity consumption information collection, improving the service communication quality and the stability of the electricity consumption data information management system, and realizing the reasonable allocation of communication resources and improving the efficiency of power data communication.

[0100] The data processing unit 4 can perform data preprocessing, multi-source data information classification, multi-source data feature extraction and information fusion processing on the power consumption data information in the distributed database, and classify the multi-source data information in the distributed database through the multi-source data information classification model of the multi-source data information classification module 42 to form a static data set. and dynamic datasets , the multi-source data feature extraction module 43 then completes the feature extraction through the data-driven mutual information principle. The feature information of the feature set obtained by this technology is complete, and the category of the feature will not change. At the same time, the information fusion processing module 44 is used to summarize the extracted feature sets so that they are all located in the source node, and the redundant features in the feature set are processed by the Kalman filter algorithm KF to complete the consistency fusion of the power consumption data, thereby ensuring a better power consumption information fusion effect; by setting the data processing unit 4, after classifying and processing the multi-source power consumption information, the dynamic data and static data therein are obtained, and the feature sets of different categories of power consumption data are extracted through data-driven, and the Kalman filter algorithm KF is used to remove the redundant features in the feature set to complete the consistent feature description of the power consumption information, thereby obtaining the power consumption information fusion result, which can complete the statistics of the terminal user's power consumption, and better solve the problem of power consumption information collection and processing during multi-source faults in the power grid. At the same time, the power consumption information fusion result obtained by using this technology can facilitate the subsequent reliable prediction of power demand and identification of abnormal power consumption behavior;

[0101] After the data is processed, the areas with suspected electricity theft users are screened out through target area screening based on the collected and processed electricity consumption information data. At the same time, for the areas with suspected electricity theft users, multiple key user analysis models are used to find out the suspected electricity theft users through the analytic hierarchy process scoring method, so as to analyze and identify the suspected electricity theft users. According to the model scoring rules, the suspected users are divided into high and general levels according to the scoring range, providing electricity inspection personnel with high-risk suspected electricity theft target users, thereby improving the quality and efficiency of anti-theft inspection work.

[0102] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for collecting big data information, comprising the following steps: S1. Collecting real-time electricity consumption data information of residents through multiple front-end data collectors; S2, processing the collected power consumption data information through the front-end central processor and the communication algorithm, and then storing it in a distributed database through the communication network, wherein the communication algorithm is a Kalman filter algorithm; S3, performing data preprocessing, multi-source data information classification, multi-source data feature extraction and information fusion processing on the power consumption data information in the distributed database, and performing multi-source data information classification on the power consumption data information in the distributed database through a multi-source data information classification model, including the following steps: A1. The multi-source electricity consumption information classification model is based on the original electricity consumption data after data processing. After completing the preliminary screening of the data, the small sample data is sampled and replicated; A2. The multi-source electricity consumption information classification model performs decoupling and encoding processing on small sample data through the deep belief network DBM to establish a coded data set; A3. After the new data is copied and encoded through the DBM network, the new data is vertically spliced ​​with the original data to obtain the expanded data of electricity consumption. A4. Train the extreme learning machine network based on the data to achieve classification and processing of electricity consumption data and form a static data set and dynamic datasets ; The method of extracting multi-source data features from the electricity consumption data information in the distributed database includes the following steps: B1. Feature extraction through data-driven mutual information principle, static data set and dynamic datasets The mutual information calculation formula between is: ； Where: for and The joint probability density function between and Are marginal probability density functions, corresponding to and ; B2. Extraction based on mutual information and The feature quantity of the extraction result is: ； Where: is the initial feature quantity, is the feature quantity with dimension m; B3. Based on the above extraction result expression, the maximum mutual information result can be obtained, and feature selection is completed on the basis of ensuring that the category does not change; B4. Use decomposition to process multi-dimensional mutual information to form one-dimensional mutual information. The calculation formula is: ； Where: and for and Selected features of the dataset; B5. Through steps B1-B4, the power consumption information feature set is quickly extracted; B6, initial variable selection and second variable selection are: ； B7. According to the expression in step B6, the selection of the remaining variables can be completed. Each time, the next feature is selected based on the principle of maximum mutual information. When selecting, a threshold N is set. When the number of cycles reaches the threshold, the selection stops. S4. By constructing different electricity theft diagnosis and analysis models, the analysis and identification of suspected electricity theft users can be realized; S5. Screen out the areas of suspected electricity theft users by means of target area screening, wherein the target area screening method includes user change monitoring, file verification, abnormal monitoring analysis and abnormal index analysis of the areas. The file verification includes multiple assessment metering points, main purpose errors of photovoltaic metering points and bidirectional metering abnormality verification. User change monitoring mainly screens areas that have not changed in the past week. Abnormal index analysis includes abnormal line loss rate, daily fluctuation of line loss rate, three-phase imbalance and abnormal power factor. S6. Construct a key user analysis model and use the analytic hierarchy process scoring method to analyze and identify suspected electricity theft users. The target substation screening method includes user change monitoring, file verification, abnormal monitoring analysis and abnormal indicator analysis. The file verification includes multiple assessment metering points, main purpose errors of photovoltaic metering points and two-way metering abnormality verification. User change monitoring mainly screens substations that have not changed in the past week. Abnormal indicator analysis includes abnormal line loss rate, daily fluctuation of line loss rate, three-phase imbalance and power factor abnormality.

2. A big data information collection system, used to implement a big data information collection method as described in claim 1, characterized in that: The system comprises a data acquisition unit (1), a communication control unit (2), a distributed data storage unit (3), a data processing unit (4), a unit for establishing a power theft diagnosis and analysis model (5), a target area screening unit (6) and a scoring prediction unit (7); The data collection unit (1) collects real-time electricity consumption data information of residents through multiple front-end data collectors, and stores the collected data in a distributed data storage unit (3) after processing by a front-end central processing and communication algorithm, and simultaneously performs visual display on a system terminal. The data collection unit (1) comprises a data collection module (11), a concentrator module (12) and a visualization module (13); The data acquisition module (11) receives data request instructions through the data port of the initialized front-end data acquisition device, thereby realizing the real-time electricity consumption data information collection of residents; The concentrator module (12) is used to issue different data request instructions under the control of the system terminal, and to simultaneously aggregate and centrally control the data of multiple front-end data collectors; The visualization module (13) displays data through the system terminal, realizing real-time data monitoring and massive data query functions; The communication control unit (2) is connected to the front-end data collector via a communication network to obtain power consumption information data, and controls and manages data transmission buffers of different data channels through an active queue management algorithm, namely an AQM algorithm; The distributed data storage unit (3) is connected to the communication control unit (2) and is used to store the collected real-time electricity consumption data information of residents in a distributed database; The data processing unit (4) is used to perform data preprocessing, multi-source data information classification, multi-source data feature extraction and information fusion processing on the electricity consumption data information in the distributed database, and the data processing unit (4) comprises a data preprocessing module (41), a multi-source data information classification module (42), a multi-source data feature extraction module (43) and an information fusion processing module (44); The data preprocessing module (41) is used to preprocess the real-time residential electricity consumption data information in the distributed database to realize real-time data stream processing, complex event stream processing, massive data offline processing and data mining functions; The multi-source data information classification module (42) constructs a multi-source data information classification model and completes the classification of multi-source electricity consumption information through the multi-source data information classification model; The multi-source data feature extraction module (43) completes feature extraction through the data-driven mutual information principle to obtain a feature set; The information fusion processing module (44) is used to aggregate the extracted feature sets so that they are all located in the source node, and to process the redundant features in the feature sets through the Kalman filter algorithm KF to complete the consistency fusion of the power consumption data; The electricity theft diagnosis and analysis model establishment unit (5) summarizes different electricity theft characteristics through big data analysis, constructs different electricity theft diagnosis and analysis models, and realizes analysis and identification of suspected electricity theft users; The target area screening unit (6) is used to monitor user changes, conduct file verification, perform abnormal monitoring analysis and abnormal index analysis on the area, and screen out areas with suspected electricity theft users; The score prediction unit (7) constructs a key user analysis model and uses a hierarchical analysis method to score key users suspected of electricity theft in the substation area, find out the suspected electricity theft users, and analyze and identify the suspected electricity theft users.

3. A big data information collection system according to claim 2, characterized in that: The multi-source data information classification module (42) completes the classification of multi-source electricity usage information through a multi-source data information classification model, comprising the following steps: C1. Assume the original data is , the electricity consumption data collected by the i-th collection point is , the correlation coefficient matrix between the data rows is ,The number of visible layer nodes of the deep belief network DBM is 20; C2. Yes After obtaining a small sample of data, the obtained sample data is filtered and normalized. express, and After training and learning through the deep belief network DBM, the outputs are and ; C3, and They have the same dimension and low coupling degree. The two are processed by vertical splicing to obtain The sampling data results of ,and The expression is as follows: ； C4. Based on the above data, the extreme learning machine network is trained to classify electricity consumption data and form a static data set. and dynamic datasets .

Images