Distribution network area portraying method based on multi-source data fusion

Abstract

The invention relates to a distribution network area portraying method based on multi-source data fusion, and the method comprises the steps: obtaining first-aid repair data and weather data, and carrying out the data integration of the first-aid repair data and the weather data, and obtaining initial data; processing the initial data to obtain application data satisfying image modeling; constructing a court portrait model based on the application data; and rating the power distribution network transformer area state based on the transformer area portrait model. According to the method, a distribution network first-aid repair organization structure is taken as a background, each link in a first-aid repair process is analyzed based on the problem of low first-aid repair efficiency, a big data preprocessing technology is utilized to process a transformer area first-aid repair work order and machine account data, and a transformer area portrait label system is established; a distribution network big data system is deeply fused, characteristic indexes under different attributes are considered, integration is carried out from a multi-dimension and multi-target level, proper rating indexes are selected, a set of distribution network zone area clusters with universality is formulated, and a set of rating system suitable for distribution network zone area state grades is constructed.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution network substation area portrait technology, and in particular to a distribution network substation area portrait method based on multi-source data fusion. Background Art

[0002] With the improvement of high-voltage power network facilities, the construction of distribution networks has become a research focus, and the upgrading and transformation of distribution networks has become an important direction for the development of power grids. Therefore, following the development trend of ubiquitous power Internet of Things, the construction of distribution network area portraits has become an important research topic. In the case of unavoidable distribution network failures and planned power outages, the quality of power supply services can only be improved through timely and rapid repairs and advance notification of power outage information and locations. The status of the equipment has become the key to achieving the above results.

[0003] Therefore, the present invention focuses on the construction of the distribution network substation image system and other aspects to take measures to provide a feasible and effective reference for improving proactive service and emergency repair to further shorten the power outage duration. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a distribution network area profiling method based on multi-source data fusion.

[0005] To solve the above problems, the technical solution adopted by the present invention is: A distribution network area profiling method based on multi-source data fusion, the method comprising: Acquire emergency repair data and weather data, and perform data integration on the emergency repair data and the weather data to obtain initial data; Processing the initial data to obtain application data that satisfies image modeling; Building a station area portrait model based on the application data; The distribution network substation status is rated based on the substation portrait model.

[0006] As an implementation mode of the invention, processing the initial data includes: Data dimension reduction: Delete irrelevant attribute entries in the initial data, and then use the class attribute subset selection to find the minimum attribute set in the data; Data quantity reduction: reduce the amount of data in each attribute set and classify similar data; Data exchange: Normalize the data to be classified so that each data is within the preset range.

[0007] As an implementation method of the invention, the data quantity reduction step includes: dividing the data set into several mutually exclusive sample sets, using different field characteristics to select data sets at different levels, and then integrating these mutually exclusive sample sets in different proportions and categories to select sample sets that reflect the original data.

[0008] As an implementation mode of the invention, the data quantity reduction step further includes: using a clustering algorithm to divide the sample set into clusters, then performing random sampling selection within the clusters, and finally completing quantity reduction.

[0009] As an embodiment of the invention, the data normalization method includes: z-score normalization, min-max normalization and decimal calibration normalization.

[0010] As an implementation mode of the invention, constructing a station area portrait model based on the application data includes: Get sample data after data exchange; According to the requirements of the substation profiling, define tags: obtain the equipment number, voltage level, power supply station, and commissioning time based on the basic attribute information of the sample data; obtain the historical fault cause and fault location information based on the fault repair work order of the sample data; The fault cause information of the substation is analyzed to obtain the situation and rules before the substation fault, and a fault model of the substation is constructed, thereby constructing the feature labels of the substation portrait.

[0011] As an implementation mode of the invention, the method further includes: predicting the substation through the established label system, and making early warning predictions on possible faults in the substation based on known data information under the same environment.

[0012] As an implementation mode of the invention, the rating of the distribution network substation status based on the substation portrait model includes: The equipment status rating indicators are divided into multiple levels, each level is strictly classified, and the corresponding level of the relevant scoring system is determined for each indicator to obtain a reasonable equipment fuzzy rating matrix; In addition to using factor analysis to perform decision analysis to determine the matrix weights, the normalization method can also be used to establish the total indicator weights. Finally, these indicator weights are used to obtain the overall equipment rating vector and the comprehensive rating value results.

[0013] The beneficial effects of adopting the above technical solution are: The distribution network substation portrait method based on multi-source data fusion provided by the present invention obtains emergency repair data and weather data, and integrates the emergency repair data and the weather data to obtain initial data; processes the initial data to obtain application data that meets image modeling; constructs a substation portrait model based on the application data; and rates the distribution network substation status based on the substation portrait model. The present invention is based on the distribution network emergency repair organizational structure, analyzes each link in the emergency repair process based on the problem of low emergency repair efficiency, uses big data preprocessing technology to process the substation emergency repair work order and ledger data, and establishes a substation portrait label system; deeply integrates the distribution network big data system, considers the characteristic indicators under different attributes, integrates from the multi-dimensional and multi-objective levels, and selects appropriate rating indicators for this purpose, formulates a set of distribution network substation clusters with extensiveness, and constructs a set of distribution network substation status rating system. By predicting the number of emergency repair work orders, it can help strengthen the monitoring of distribution network substations, and at the same time, it can also warn of substation failure problems in advance and minimize losses in advance. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a schematic diagram of the distribution station area status rating index system.

[0015] Figure 2 It is a flow chart of fuzzy comprehensive evaluation method. DETAILED DESCRIPTION

[0016] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention is clearly and completely described below in conjunction with specific embodiments.

[0017] The embodiment of the present invention provides a distribution network area profiling method based on multi-source data fusion, the method comprising: Step 1: Acquire emergency repair data and weather data, and perform data integration on the emergency repair data and the weather data to obtain initial data.

[0018] The emergency repair data required for the area portrait comes from the records of the emergency repair department, and the weather data comes from the meteorological department. On the surface, they are unrelated. These data sets have different dimensions, large differences in size, and even no connection, so data integration is needed for processing. The first step is to clean up the data structure style and find common attributes that can unite these data together. These data come from different departments, record different things, and have special formats. In order to facilitate research, these formats must be standardized. The standard format has a key attribute called Key. It is the only key to find all information. Regularly integrate the determined data sets, maintain the original meaning, and determine whether this attribute allows invalid values ​​or zero values, etc.; the new problem that comes with it is that the redundancy problem associated with data integration must be dealt with. When multiple data sets are integrated together, there will inevitably be integration redundancy. The correlation analysis principle provides a solution. By judging the correlation of data sets, the higher the correlation, the greater the redundancy, and the lower the correlation, the less redundancy. Finally, integration will also cause data conflict problems. Data sets from different sources have various standards, which may result in multiple different collection or storage standards for one thing.

[0019] Step 2: Process the initial data to obtain application data that meets the requirements of image modeling.

[0020] Processing the initial data includes: Step S201: Data specification The metering and storage capabilities of distribution networks are constantly improving, and data is becoming increasingly large and detailed, which brings opportunities for research and analysis, but also creates certain troubles. The purpose of data reduction is to make data analysis more accurate while reducing the time and effort caused by the increasing amount of big data. It includes data dimension reduction and data quantity reduction. (1) Data dimension reduction: Delete irrelevant attribute entries in the initial data, and then use class attribute subset selection to find the minimum attribute set in the data; (2) Data quantity reduction: reduce the amount of data in each attribute set and classify similar data; The steps include: dividing the data set into several mutually exclusive sample sets, using different field characteristics to select data sets at different levels, and then integrating these mutually exclusive sample sets in different proportions and categories to select sample sets that reflect the original data.

[0021] Alternatively, a clustering algorithm is used to divide the sample set into clusters, and then random sampling is performed within the clusters to finally complete the quantity reduction.

[0022] Step S202, data exchange: normalize the data to be classified so that each data is within a preset range.

[0023] In the data preprocessing stage, character data such as weather conditions are transformed and included in the calculation to enhance the accuracy of the results. The first is data normalization. The size and width of each data attribute value in the data set are different, and the data mining algorithm needs to control the original data attributes to be used within a certain range. By using data normalization, the original data value conversion can be controlled within a more reasonable range. These common methods of numerical normalization include z-score normalization, min-max normalization, and decimal calibration normalization. The second is data discretization. Some data have relatively concentrated attribute information, which needs to be processed through discretization to make the final data analysis more reasonable and convincing. According to different discretization methods, one is to subdivide the data attributes, and the other is to merge the data attributes. Finally, feature construction, data feature construction has a strong significance for data processing. Data construction takes basic data information as the research basis and creates data features that are helpful for analysis results, making these data more valuable.

[0024] Step three: construct a station area portrait model based on the application data.

[0025] This step includes: Step S301, obtain original sample data. The data used in this paper are the historical repair work order data of the repair command platform and the weather forecast system data. By collecting the data on the causes of the faults in the substation area and analyzing the correlation between the fault factors, the fault characteristics of the substation area are obtained; Step S302: Analyze historical sample data and establish definition tags according to the requirements of the substation image. Clean and simplify the substation equipment data. From the basic attribute information of the substation, you can get the equipment number, voltage level, power supply station, time of commissioning, etc. From the fault repair work order, you can get the historical fault cause, fault location, etc. Step S303: Analyze the association rules of the fault causes of the substation area and construct the characteristic labels of the substation area portrait. By analyzing the fault causes and other information of the substation area, the situation and rules before the fault of the substation area are obtained, and the fault model of the substation area is constructed; Step S304: predict the substation area through the established label system, and make early warning predictions on possible faults in the substation area based on known data information under the same environment to improve emergency repair efficiency.

[0026] Step 4: Rating the status of the distribution network substation based on the substation portrait model.

[0027] The steps include: Step S401: Determine the electronic device status evaluation index Considering the basic attributes, usage characteristics and historical overview of power equipment in the distribution network, the following power equipment status evaluation indicators are selected and summarized. Considering the distribution network substation status rating indicator system under multi-source data sources, Figure 1 shown.

[0028] Ⅰ) Meteorological indicators A: The meteorological department releases detailed meteorological data for each region, which is of great help in collecting meteorological information in the area where the station is located. The predicted meteorological data is used to evaluate the safety level of the environment in the area where the station is located. These include: ① temperature; ② humidity; ③ weather; ④ wind force. Based on the original data of the meteorological forecast department, strengthening the use of meteorological data helps to understand the working environment status of the station in detail, which greatly helps to warn the safety status of the equipment in advance.

[0029] II) Power equipment attribute index B: The basic information of power equipment in the distribution network substation also has a vital impact on the substation status, including basic attribute information of power equipment such as transformers and cables. The source data include: ① transformer service life; ② cable service life; ③ circuit breaker service life; ④ protection method; ⑤ number of users in the substation; ⑥ substation load level.

[0030] III) Historical fault index C: Its source data includes ① number of transformer failures; ② number of line failures; ③ number of overvoltages; ④ number of overloads; ⑤ number of improper operating modes; ⑥ number of equipment defects; ⑦ number of equipment aging; ⑧ number of customer false alarms.

[0031] IV) Real-time status indicator D of power equipment: Starting from the real-time data of the distribution network substation, based on real-time information monitoring, paying close attention to the equipment status information can greatly strengthen the monitoring of the substation operation status. The source data include: ① daily maximum voltage; ② daily minimum voltage; ③ daily maximum current; ④ daily minimum current; ⑤ voltage qualification rate; ⑥ daily maximum load rate; ⑦ daily maximum imbalance rate.

[0032] Step S402: Divide the equipment status rating indicators into multiple levels, strictly classify each level, determine the relevant scoring system of the corresponding level for each indicator, and obtain a reasonable equipment fuzzy rating matrix; Step S403: In addition to using factor analysis to perform decision analysis to determine the matrix weights, the normalization method can also be used to determine the total indicator weights. Finally, these indicator weights are used to obtain the overall equipment rating vector and the comprehensive rating value result. The evaluation process is as follows: Figure 2 shown.

[0033] The distribution network substation portrait method based on multi-source data fusion provided by the present invention takes the distribution network emergency repair organizational structure as the background, analyzes each link in the emergency repair process based on the problem of low emergency repair efficiency, uses big data preprocessing technology to process the substation emergency repair work order and ledger data, and establishes a substation portrait label system; deeply integrates the distribution network big data system, considers the characteristic indicators under different attributes, integrates from the multi-dimensional and multi-objective levels, and selects appropriate rating indicators for this purpose, formulates a set of distribution network substation clusters with extensiveness, and constructs a set of distribution network substation status rating system. By predicting the number of emergency repair work orders, it can help strengthen the monitoring of distribution network substations, and at the same time, it can also warn of substation failure problems in advance and minimize losses in advance.

Claims

1. A distribution network area profiling method based on multi-source data fusion, characterized in that: The method comprises: Acquire emergency repair data and weather data, and perform data integration on the emergency repair data and the weather data to obtain initial data; Processing the initial data to obtain application data that satisfies image modeling; Building a station area portrait model based on the application data; The distribution network substation status is rated based on the substation portrait model.

2. A distribution network area profiling method based on multi-source data fusion according to claim 1, characterized in that: Processing the initial data includes: Data dimension reduction: Delete irrelevant attribute entries in the initial data, and then use the class attribute subset selection to find the minimum attribute set in the data; Data quantity reduction: reduce the amount of data in each attribute set and classify similar data; Data exchange: Normalize the data to be classified so that each data is within the preset range.

3. A distribution network area profiling method based on multi-source data fusion according to claim 2, characterized in that: The data quantity reduction step includes: dividing the data set into several mutually exclusive sample sets, selecting data sets of different levels by using different field characteristics, and then integrating these mutually exclusive sample sets in different proportions and categories to select sample sets that reflect the original data.

4. A distribution network area profiling method based on multi-source data fusion according to claim 2, characterized in that: The data quantity reduction step also includes: using a clustering algorithm to cluster the sample set, then performing random sampling selection within the cluster, and finally completing the quantity reduction.

5. A distribution network area profiling method based on multi-source data fusion according to claim 2, characterized in that: The data normalization methods include: z-score normalization, min-max normalization and decimal calibration normalization.

6. A distribution network area profiling method based on multi-source data fusion according to claim 1, characterized in that: Building a station area portrait model based on the application data includes: Get sample data after data exchange; According to the requirements of the substation profiling, define tags: obtain the equipment number, voltage level, power supply station, and commissioning time based on the basic attribute information of the sample data; obtain the historical fault cause and fault location information based on the fault repair work order of the sample data; The fault cause information of the substation is analyzed to obtain the situation and rules before the substation fault, and a fault model of the substation is constructed, thereby constructing the feature labels of the substation portrait.

7. A distribution network area profiling method based on multi-source data fusion according to claim 6, characterized in that: The method also includes: predicting the substation through the established label system, and making early warning predictions on possible faults in the substation based on known data information under the same environment.

8. The distribution network area profiling method based on multi-source data fusion according to claim 1 is characterized in that: The rating of the distribution network substation status based on the substation portrait model includes: The equipment status rating indicators are divided into multiple levels, each level is strictly classified, and the corresponding level of the relevant scoring system is determined for each indicator to obtain a reasonable equipment fuzzy rating matrix; In addition to using factor analysis to perform decision analysis to determine the matrix weights, the normalization method can also be used to establish the total indicator weights. Finally, these indicator weights are used to obtain the overall equipment rating vector and the comprehensive rating value results.

Images