CN118312813A

CN118312813A - A method and system for intelligent health diagnosis of substations based on big data algorithm

Info

Publication number: CN118312813A
Application number: CN202410420537.7A
Authority: CN
Inventors: 熊小舟; 柏杨; 胡兵; 王松; 褚红亮; 鄂驰; 邱贞宇
Original assignee: Information and Telecommunication Branch of State Grid Jiangxi Electric Power Co Ltd
Current assignee: Information and Telecommunication Branch of State Grid Jiangxi Electric Power Co Ltd
Priority date: 2024-04-09
Filing date: 2024-04-09
Publication date: 2024-07-09

Abstract

The invention discloses a large data algorithm-based intelligent health diagnosis method and system for a platform area, which are used for early warning an abnormal platform area in real time, accurately positioning abnormal factors, providing a solution for the abnormal platform area, avoiding the conventional comprehensive manual investigation time, reducing the manual consumption cost, improving the solution efficiency of an electric abnormal work order, enabling the found problems to be rapidly and effectively solved, improving the control force of an electric network management, directly eliminating defects by field personnel by using a palm machine, reducing the workload caused by inconvenient data query and untight work flow connection, improving the work efficiency, reducing the waste of electric resources and natural resources, improving the utilization rate of resources, improving the service quality of a power grid, simultaneously reducing the consumption of resources and energy resources, strengthening line loss management, avoiding the potential increase, reducing the electricity cost of industrial and commercial users, and enabling the development of the power grid to become practical.

Description

Intelligent health diagnosis method and system based on big data algorithm for transformer area

Technical Field

The invention relates to the technical field of big data algorithms, in particular to an intelligent health diagnosis method and system based on a big data algorithm platform region.

Background

Through the intelligent health diagnosis application management and control of the transformer area, a standardized analysis model is provided for a company, a transformer area line loss standardized management flow is realized, electricity stealing and electricity leakage suspected users are timely found, the comprehensive line loss rate of the company is reduced, the economic benefit of the company is improved, abnormal transformer areas can be early warned in real time, abnormal factors are accurately positioned, an abnormal transformer area solution is provided, the conventional omnibearing manual investigation time is avoided, the labor consumption cost is reduced, and the solution efficiency of an electric abnormal work order is the problem to be solved urgently in the current transformer area intelligent management process.

Disclosure of Invention

The invention aims to provide a large data algorithm-based intelligent health diagnosis method and system for a platform region, which are used for solving the defects in the background technology.

In order to achieve the above object, the present invention provides the following technical solutions: a big data algorithm-based intelligent health diagnosis method for a platform region comprises the following steps:

Analyzing a topological structure of a platform area and the running state of equipment, extracting a line loss influence factor of the platform area, and establishing a holographic representation of the platform area by combining a representation technology;

According to the power supply radius and load characteristic factors of the transformer area, cluster analysis and calculation of the line loss interval of the transformer area are carried out on the transformer area;

Carrying out differential analysis on the statistical line loss of the transformer area and the line loss interval, and realizing abnormal positioning and label display of the line loss of the transformer area through correlation analysis and an on-line monitoring technology;

based on a variance analysis technology, researching the influence of various area line loss factors on the area line loss, and constructing an area line loss abnormity diagnosis model by combining big data and artificial intelligence technology;

And (3) locating, diagnosing and treating the abnormal line loss of the transformer area, constructing a case library for diagnosing the abnormal line loss of the transformer area, and realizing intelligent diagnosis and analysis of the health of the transformer area.

Preferably, the creating a holographic representation of a region in combination with a representation technique comprises the steps of:

Based on the line loss related statistics report form data, comprehensively analyzing electricity utilization structure, developing caliber comprehensive line loss, marketing line loss and electricity utilization amount data, constructing a diagnosis model from equipment operation, communication capacity, electricity utilization information and electric energy quality direction, forming super capacity, household change relation errors, suspected gateway table wiring abnormality, reverse word passing, incomplete photovoltaic user table codes and disassembling and replacing table bottom code error abnormal labels of a platform area, counting abnormal quantity on the labels, clicking the labels to check abnormal details, cleaning and screening construction platform area operation data, extracting platform area characteristics, abnormal label technology, platform area meter, terminal and line loss management multidimensional holographic images.

Preferably, the construction of the site line loss abnormality diagnosis model includes the following steps:

The master station recalls the big data of the power-off time in the intelligent meter according to the data of the power-off time in the intelligent meter by checking the data of the power-off event of the meter, the household meter and the terminal, and accurately judges the household change relation according to the consistency of the power-off time;

Based on meter files, the line loss statistics of the areas and the characteristic data of the HPLC areas, when the line loss of the areas is abnormal, the areas are considered as suspected household change relation abnormal areas, a master station is triggered to collect voltage curves in the total tables and household tables of the areas and adjacent areas, the correlation of the big data of the voltage curves between the total tables and the household tables of each area is compared and analyzed, and the household change relation is automatically identified;

collecting current data of an examination table, and calculating the unbalance of the three-phase load current of the transformer area epsilon by using the three-phase load current:

Wherein: i _max represents the maximum one-phase load current (a), I _avp represents the average value (a) of the three-phase load currents;

Outputting a three-phase balance label when the unbalance epsilon is smaller than 25%, and outputting a one-heavy label and a two-light label when the unbalance epsilon is larger than 25% and the three-phase load is one-heavy and two-light;

dividing a station area into three parts of a distribution transformer, a circuit and a load according to the actual condition of a low-voltage station area, wherein the influence parameters of all the parts are different, and the influence parameters comprise the capacity, gear number, load rate and outgoing line number of the distribution transformer, the power supply radius, wire diameter and material of the low-voltage circuit, the unbalance degree, distribution characteristic and power factor of the load;

Constructing a line voltage loss comprehensive estimation model formed by site data of a platform area, wherein the function expression is as follows:

U _n is the nominal voltage of the line, K ₁ is an unbalance coefficient, K ₂ is a load distribution coefficient, K _3i is the single-bit voltage loss coefficient of the i-th section low-voltage line, and sigma _i is the full length ratio of the i-th section low-voltage line to the main line; i _avi is the average phase current of the low-voltage line loss of the ith section, calculates the theoretical voltage drop of the station area, compares the theoretical voltage drop with the terminal voltage of the terminal user of the station area, and automatically identifies the low-voltage abnormal station area;

The method comprises the steps of collecting external weather data, daily load of a distribution transformer, and carried electricity customer attribute and distribution transformer attribute data, taking a load value as a target input value, taking the load value as a key classification variable, analyzing importance sequences of the variables on heavy overload through mutual information concepts, screening important characteristic variables, and establishing a classification model between the variables and the distribution transformer heavy overload.

Preferably, the cluster analysis and the calculation of the line loss interval of the transformer area comprise the following steps:

Collecting data related to the line loss of the transformer area, including the transformer area monitoring line loss rate, load power factor, low-voltage line number, low-voltage line length, main line diameter, main line material, connection group, current, voltage, load curve form factor, power supply quantity, electricity sales quantity, three-phase capacity occupation ratio, household number, transformer type, distribution capacity, distributed power supply household number, access capacity and occupation ratio, charging pile number, charging station number, transformer operation year, branch box, meter box number, meter count and meter version information;

Taking the monitoring line loss rate of the transformer area as a target variable, calculating the influence of each factor on the line loss rate according to a decision tree or a random forest model and an information entropy principle, selecting factors with large significance as model input variables of the transformer area classification according to the importance ordering and significance analysis of each factor, and taking the identified transformer area classification factors as the input variables of the model to construct various transformer area clustering models;

The theoretical line loss electric quantity of the station area consists of electric quantity lost by station area lines and various electric energy meters, and the theoretical line loss rate of the station area is expressed as:

The total power supply quantity of the transformer area is total daily active total electric quantity E of the transformer area, the technical line loss rate of each transformer area is calculated by considering the correction of the three-phase unbalance and the load fluctuation influence when calculating the line loss electric quantity of the transformer area, the transformer area line loss rate is assumed to be a normal distribution random variable, and the transformer area line loss interval value of the 95% confidence interval is calculated according to the transformer area 'stable day' technical line loss statistical condition:

A big data algorithm-based intelligent health diagnosis system for a platform region comprises an analysis module, a calculation module, a positioning module, a model construction module and a case library construction module;

and an analysis module: analyzing a topological structure of a platform area and the running state of equipment, extracting a line loss influence factor of the platform area, and establishing a holographic representation of the platform area by combining a representation technology;

the calculation module: according to the power supply radius and load characteristic factors of the transformer area, cluster analysis and calculation of the line loss interval of the transformer area are carried out on the transformer area;

and a positioning module: carrying out differential analysis on the statistical line loss of the transformer area and the line loss interval, and realizing abnormal positioning and label display of the line loss of the transformer area through correlation analysis and an on-line monitoring technology;

Model construction module: based on a variance analysis technology, researching the influence of various area line loss factors on the area line loss, and constructing an area line loss abnormity diagnosis model by combining big data and artificial intelligence technology;

The case library construction module: and (3) locating, diagnosing and treating the abnormal line loss of the transformer area, constructing a case library for diagnosing the abnormal line loss of the transformer area, and realizing intelligent diagnosis and analysis of the health of the transformer area.

In the technical scheme, the invention has the technical effects and advantages that:

According to the invention, a standardized analysis model is provided for a company through the intelligent health diagnosis application management and control of the transformer area, so that a standardized management flow of the transformer area line loss is realized, suspected users of electricity theft and electricity leakage are found in time, the comprehensive line loss rate of the company is reduced, and the economic benefit of the company is improved; the system has the advantages that abnormal areas can be early warned in real time, abnormal factors can be accurately positioned, an abnormal area solution is provided, the original comprehensive manual investigation time is avoided, the manual consumption cost is reduced, the solution efficiency of an electric abnormal work order is improved, the found problems can be quickly and effectively solved, the control force of an electric network manager is improved, the defects of on-site personnel can be directly eliminated by using a palm machine, the workload caused by inconvenient data inquiry and untight connection of working procedures is reduced, the working efficiency is improved, the waste of electric resources and natural resources is reduced, the utilization rate of the resources is improved, the service quality of a power grid is improved, the consumption of resources is reduced, the line loss management is enhanced, the leakage is avoided, the potential efficiency is increased, the electric charge of commercial users is reduced, and the development achievement of the power grid is enabled to become practical in the whole society.

Drawings

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for the embodiments will be briefly described below, and it is apparent that the drawings in the following description are only some embodiments described in the present application, and other drawings may be obtained according to these drawings for a person having ordinary skill in the art.

FIG. 1 is a flow chart of the method of the present invention.

Fig. 2 is a system architecture diagram of the present invention.

Fig. 3 is a system technical architecture diagram of the present invention.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

Examples: referring to fig. 1, the intelligent health diagnosis method based on big data algorithm for a platform area according to the embodiment includes the following steps:

The method for establishing the area holographic image by combining the image technology comprises the following steps:

The construction of the platform area line loss abnormity diagnosis model comprises the following steps:

The cluster analysis and the calculation of the line loss interval of the transformer area comprise the following steps:

The flow chart of the present invention is shown in fig. 1.

As shown in fig. 2, the data architecture of the method and the system for intelligent health diagnosis of the large data algorithm-based platform area relies on the platform area basic information, the platform area power consumption information, the platform area abnormal information and the platform area label information all-service data of the system such as a power consumption information acquisition system, a marketing service application system, a GIS system, PMS2.0 and the like, and combines the technologies such as large data calculation, data mining and the like to form relevant labels and analysis data, and provides support for various marketing informatization systems through data sharing service.

The system service architecture mainly comprises a full service data center, a low-voltage area holographic portrait platform and a service application module.

The full-service data center module is used for obtaining high-quality basic data for calculation and analysis based on data of a cross-platform multi-source full-service center such as a fusion electricity information acquisition system, a marketing service application system, a GIS system, PMS2.0 and the like through screening and cleaning of multiple pairs of source data.

The low-voltage area holographic portrait platform module is used for analyzing the topological structure and the running state of equipment of the low-voltage area by relying on basic data acquired by the full-service data center module, extracting the line loss influence factors of the low-voltage area and establishing the area holographic portrait by combining portrait technology

The service application module realizes functions including the labeling of the characteristics and the abnormality of the transformer area, the holographic image of the sound transformer area, the fine classification of the low-voltage transformer area, the calculation of a reasonable line loss interval, the construction of a transformer area line loss abnormality diagnosis case library, the construction of a transformer area health intelligent physical examination function, the tracking analysis of the abnormal relevance of the transformer area line loss and the visual display of a transformer area line loss large screen.

As shown in fig. 3, the design of the technical architecture of the system of the method conforms to the 3-layer architecture of the data layer, the service layer and the presentation layer of the software development. Based on J2EE technology, corresponding service application is constructed, and the software function is fully and repeatedly utilized to meet the service requirement.

The data layer mainly stores the business system data into a database after ETL processing, and simultaneously caches part of the data into a server through a caching technology, so that the system influence speed is increased.

The business logic layer provides public service and business logic processing, comprises a business layer and a persistent layer, and adapts to the future flexibility and expansibility requirements by adopting the SOA technology based on WebService. The data interface provides a WebService mode to transmit data, so as to meet various data transmission requirements. The background generates service call log, access log, operation log and running log, and records various process information in all directions.

The main task of the presentation layer is to receive a client access application and display a user interface. And various access ways such as a mobile terminal device of a browser are supported. The front end adopts an automatic construction tool based on stream (stream), and executes construction tasks in a configuration file mode in combination with Grunt, and automatically runs set construction page tasks. The main content is based on LESS dynamic style language, and CSS is preprocessed, so that page style has dynamic property, and the method is suitable for various terminals.

Through healthy intelligent diagnosis application management and control of district, provide standardized analytical model, realize district line loss standardization management flow, in time discover steal electric, electric leakage suspicion user, reduce the company and synthesize the line loss rate, improve company's economic benefits, through healthy intelligent diagnosis application management and control of district, but unusual district of real-time early warning, accurate location abnormal factor to provide unusual district solution, avoid the all-round manual investigation time in the past, reduce the cost of labor consumption.

The solution efficiency of the power abnormal work order is improved, so that the found problems can be solved rapidly and effectively, and the control force of the power grid is improved. The on-site personnel uses the palm machine to directly eliminate the defects, so that the workload caused by inconvenient data inquiry and untight work flow connection is reduced, and the work efficiency is improved. The waste of electric power resources and natural resources is reduced, and the utilization rate of the resources is improved. And the resource consumption is reduced while the service quality of the power grid is improved.

Through improving more dimensional data service, the line loss management is enhanced, the running, the leakage and the drip are stopped, the electric charge of industrial and commercial users is reduced by the development and the efficiency of the power grid, and the development of the power grid is realized.

Based on the existing low-voltage area holographic portrait platform, application promotion in the aspects of area characteristics and anomaly analysis, line loss diagnosis analysis, health examination and the like is developed, and lean management of the area is realized. Theoretical basis is provided for checking line loss indexes of the transformer area line loss management.

1.1 Labeling of region characteristics and anomaly problems, sound region holographic representations

Based on the line loss related statistics report form data, comprehensively analyzing data such as electricity consumption structure, development caliber comprehensive line loss, marketing line loss, electricity consumption and the like, constructing a diagnosis model from the directions of equipment operation, communication capacity, electricity consumption information, electric energy quality and the like, forming abnormal labels such as super capacity, user change relation errors, suspected gateway table wiring abnormality, reverse word passing, incomplete photovoltaic user table codes, disassembly and replacement table bottom code errors and the like of a platform area, counting abnormal quantity on the labels, and checking abnormal detail by clicking the labels. The method comprises a construction area operation data cleaning and screening technology, an area characteristic extracting and abnormal label extracting technology, an area meter, a terminal, a line loss management technology and other multidimensional holographic image technologies.

1.1.1 Region operation data cleaning and screening

Based on a full-service data center, a power consumption information acquisition system, a marketing service application system, a marketing GIS system, PMS2.0 and other multi-source data preprocessing and data cleaning work related to the operation of an exhibition area, the topology of the area and line loss calculation data are synthesized, a dynamically configurable complement method and a multi-dimensional quantized rejection rule are provided, a sample selection method with the characteristic of stabilizing the hour, day and month line loss rate is designed, and data support is provided for big data mining and association analysis.

1.1.2 Region user variable relation error diagnosis model

The master station recalls the big data of the power-off time in the intelligent meter according to the data of the power-off time in the intelligent meter by checking the data of the power-off event of the household meter and the terminal, and accurately judges the household change relation according to the consistency of the power-off time;

Based on meter files, the line loss statistics of the areas and the characteristic data of the HPLC areas, when the line loss of the areas is abnormal, the areas are considered as suspected household change relation abnormal areas, the master station is triggered to collect 96-point (15 minutes of freezing primary voltage data) voltage curves in the total table and the adjacent areas and the household table, the correlation of the big data of the 96-point voltage curves between the total table and the household table of each area is compared and analyzed, and the household change relation is automatically identified.

(1) Three-phase imbalance diagnostic model

Usually collecting current data of an examination table, and calculating the unbalance degree epsilon of the three-phase load current of the transformer area by utilizing the three-phase load current:

Wherein: i _max -maximum one-phase load current (a);

i _avp -average value of three-phase load current (A).

Outputting a three-phase balance label when the unbalance epsilon is less than 25%; when the unbalance epsilon is larger than 25%, and the three-phase load is one-phase heavy (the maximum current is larger than 1.2 times of the average current, the same applies below) and two-phase light (the minimum current is smaller than 0.8 times of the average current, the same applies below), outputting a 'one-heavy' label and a 'two-light' label; the same model outputs "two-phase heavy, one-phase light" and "one-phase heavy, one-phase light, one-phase flat" labels.

1.1.3 Diagnostic model for abnormal voltage in transformer area

According to the actual condition of the low-voltage transformer area, the transformer area can be divided into three main parts of a distribution transformer, a circuit and a load, wherein the influence parameters of all the parts are different. Factors affecting the voltage loss of the low voltage region include: capacity, gear number, load rate and outgoing line number of the distribution transformer; the power supply radius, the wire diameter and the material of the low-voltage circuit; load imbalance, distribution characteristics, power factor, etc.

According to the electric energy loss calculation rule of the electric power network, a comprehensive line voltage loss estimation model formed by site data of a transformer area is provided

U _n is the line nominal voltage, V; k ₁ is the imbalance coefficient; k ₂ is a load distribution coefficient; k _3i is the unit voltage loss coefficient of the i-th section low-voltage line,% A.km; sigma _i is the full length ratio of the i-th low-voltage line to the main line; i _avi is the average phase current, A, left through the I-th section low voltage line loss. The theoretical voltage drop of the low-voltage area can be effectively calculated through the model, and compared with the terminal voltage of the terminal user of the area, the low-voltage abnormal area can be automatically identified.

1.1.4 Distribution transformer overload diagnostic model

The influence of the overload of a single distribution transformer is numerous. By collecting external weather data, daily load of distribution transformer, and carried electricity customer attribute and distribution transformer attribute data, the load value is used as a target input value, the key classification variable is adopted, the importance ordering of the heavy overload of each variable is analyzed through mutual information concept, the important characteristic variable is screened, a classification model between the variable and the distribution transformer heavy overload is established, and the final classification model output is as follows:

1.1.5 Multi-dimensional holographic image construction

Based on network parameters such as low-voltage area meters, terminals, line diameter materials, area capacity, user types and quantity, area abnormal labels and monitoring line loss rate data, an area operation index, an operation index and health index system is constructed through data mining technologies such as association analysis, cluster analysis and the like, and holographic images of main bodies of the areas, users under the areas and equipment (meters, terminals and the like) are formed; constructing a multi-dimensional area holographic image billboard function facing different objects such as an administrator, an area manager and the like; and combining the running characteristics of the transformer area and the abnormal labels, constructing a transformer area management plate function, realizing the screening of the transformer area with abnormal line loss rate and the positioning of the transformer area line loss abnormal risk points, and providing targets for line loss management.

1.2 Fine classification of low-voltage transformer area and calculation of reasonable line loss interval

Based on the electricity consumption information acquisition system, marketing business application system, PMS2.0, GIS and other system files and electricity consumption data, the topological structures of different areas are clarified by means of big data mining, association analysis and the like, the electricity consumption characteristics of the different areas are analyzed, and the intelligent subdivision of the area clustering is developed by utilizing urban and rural categories of the areas, electricity consumption load classification, and index data such as transformer capacity, power supply quantity, resident/non-resident number, power factor, three-phase imbalance, power supply radius and the like which influence the line loss of the areas, and by combining theoretical line loss calculation results of all the areas, reasonable line loss intervals of all the classified areas are calculated and used for guiding the accurate loss reduction of the areas. The method comprises the steps of cluster analysis of the area based on machine learning, construction of an area line loss reasonable interval calculation model based on machine learning, construction of an area line loss key factor knowledge base and a multi-element loss reduction strategy base, and comprehensive verification of the area line loss reasonable interval calculation model.

1.2.1 Machine learning-based region Cluster analysis

1.2.1.1 Region line loss key factor identification

First, data related to the line loss of the area is collected or calculated: the transformer area monitors information such as line loss rate, load power factor, low-voltage line number, low-voltage line length, main line diameter, main line material, connection group, current, voltage, load curve form factor, power supply quantity, electric quantity, three-phase capacity ratio, number of units, transformer model, distribution transformer capacity, distributed power supply number of units, access capacity and ratio, number of charging piles, number of charging stations, operation year of a transformer, branch boxes, number of meter boxes, meter counts, meter version and the like; and then taking the monitoring line loss rate of the transformer area as a target variable, and calculating the influence of each factor on the line loss rate according to a decision tree or a random forest model and an information entropy principle.

Finally, according to the importance ranking and the significance analysis of each factor, key factors with strong significance and larger influence are selected as model input variables of the platform region classification.

1.2.2 Table division model

And (3) taking the region classification key factors identified in the section 1 as input variables of the model, constructing a plurality of region clustering models by using K-means, KNN, hierarchical clustering and other methods, researching the applicability and robustness of different clustering methods to typical stable region clustering, screening out an optimal model suitable for the region clustering by combining the classification accuracy of the models, and finally opening the region clustering model for modeling to realize the clustering analysis of the region.

1.2.3 Area line loss reasonable interval calculation model

(1) Pressure drop method model

The theoretical line loss electric quantity of the 0.4kV low-voltage transformer area is generally composed of electric quantity lost by transformer area lines (including branch switches and other equipment) and various electric energy meters, and the theoretical line loss rate of the transformer area can be expressed as:

The total power supply quantity of the platform area is the total daily active total electric quantity E of the platform area.

According to the theoretical line loss rate composition formula of the transformer area, the line loss electric quantity of the transformer area, the loss electric quantity of the electric energy meter and the total power supply quantity of the transformer area are calculated, and the theoretical line loss rate of the transformer area can be calculated. However, since the three-phase imbalance and the load fluctuation both affect the line loss of the transformer area, the correction of the three-phase imbalance and the load fluctuation is considered when calculating the line loss electric quantity of the transformer area.

The model requires the main measurement parameters: head end voltage, tail end voltage, check meter current and check meter electric quantity.

1) Calculating the voltage drop rate of the first and the last ends of the transformer area

2) Calculating the phase angle tangent between the current and the voltage

3) Calculating the power loss and voltage loss ratioIn the method, in the process of the invention,The impedance ratio of the line of the station area can be obtained by looking up a table.

4) Calculating a three-phase imbalance correction coefficient, wherein the specific correction coefficient kb is related to the three-phase imbalance epsilon, and the three-phase imbalance correction coefficient is calculated as follows:

When the three-phase unbalance epsilon is less than 25 percent: k _b =1.17

When the three-phase unbalance epsilon is more than 25 percent:

a) Three phases are heavy in load, light in load and average in load:

b) Three-phase load is one-phase heavy, two-phase light: k _b＝1+2ε²

C) Three-phase load is heavy in two phases, and one phase is light: k _b＝1+8ε²

5. Calculating a load fluctuation correction coefficient (shape factor)

6) Calculating the power consumption of the transformer area

7) Electric energy meter loss electricity quantity calculation for transformer area

T is taken for 24 hours.

8) Calculation of theoretical line loss rate of the combined regionAccording to the calculation steps, the technical line loss rate of each station area can be calculated. Assuming that the line loss rate of the transformer area is a normal distribution random variable, calculating a reasonable interval value of the transformer area line loss of a 95% confidence interval according to the line loss statistical condition of the transformer area 'stable day' technology:

(2) Multiple linear regression model

In the process of calculating the technical line loss according to the pressure drop method, if necessary parameters are missing and cannot be calculated, a multiple linear regression model can be established, and a reasonable line loss interval of the transformer area is calculated.

A. Classification to build multiple linear regression models

1) According to the area classification model, selecting area line loss key factors as input variables, taking monitoring line loss as target variables, selecting area 'stable operation' data as samples, classifying the samples into a training set (75%) and a testing set (25%) by using a ten-fold intersection method, and respectively training multiple linear regression models for each type to obtain an area line loss linear regression calculation model.

2) And (3) using a test set, according to the trained regression model, making a histogram of the predicted measurement and the actual measurement, checking whether the error has normal distribution characteristics, verifying the rationality of the linear regression model, and simultaneously giving a scatter diagram between the predicted measurement and the actual measurement so as to intuitively check the relation between the predicted variable and the actual measurement, and performing parameter tuning on the model with larger error until the service requirement is met.

B. calculating the line loss of the transformer area according to a multiple linear regression model

And (3) for the areas which cannot be calculated by the pressure drop method, judging the category of the areas according to the clustering result, calculating and predicting reasonable line loss values of the areas according to the multiple linear regression models corresponding to the category, and solving the line loss rate qualification range with 90% confidence space.

3) Comprehensive verification of calculation model of reasonable interval of line loss of transformer area

On the basis of accurate clustering subdivision of a low-voltage area, a diversified test point area is selected from a plurality of angles such as geographic position, area capacity, area load rate and area power consumption characteristics to carry out result verification; calculating a reasonable line loss interval of the test point station area by using a reasonable line loss model of the station area, comparing and analyzing with the statistical line loss of the station area, and optimizing the model; and verifying the accuracy of the method for measuring and calculating the user loss contribution of the high-loss area through the high-loss area treatment effect.

1.2.4 Construction of district line loss key factor knowledge base and multiple loss reduction strategy base

Obtaining a line loss key factor parameter of a platform region, respectively calculating the importance change condition of the line loss key factor of the platform region under three different operating environments of high load, medium load and low load by utilizing an entropy method according to the load distribution of the platform region, and completing the construction of a knowledge base of the line loss key factor of the platform region by dynamically measuring and calculating the contribution degree weight of the line loss key factor; constructing multiple loss reduction measures aiming at line loss key factors of a low-voltage transformer area, and configuring differentiated loss reduction strategies aiming at different key factors.

1.3 Construction of case library for line loss abnormality diagnosis in transformer area

The method comprises the steps of analyzing an existing case library for abnormal line loss of the transformer area, combing and forming key problems affecting the abnormal line loss of the transformer area by combining with expert experience, wherein the key problems comprise error relation of the transformer area, three-phase imbalance, abnormal voltage of the transformer area, heavy overload of distribution transformer, suspected electricity larceny, power loss of the transformer area and the like. And a typical coping strategy is designed aiming at the key problems, an abnormality aiming at a case base is formed, and a basic unit is guided to efficiently and accurately check the reasons of the line loss abnormality of the transformer area. The method comprises the steps of collecting and carding typical cases of abnormal line loss in a platform area, designing and developing case library application and applying the case library.

1.3.1 Area line loss abnormity typical case library model

1) The method comprises the steps of researching and collecting abnormal line loss problems of various areas, combing and forming key problems affecting abnormal line loss of the areas by combining expert experience, wherein the key problems comprise abnormal change relation of the areas, three-phase imbalance, abnormal voltage of the areas, heavy overload of distribution transformer, suspected electricity larceny, power loss of the areas and the like, classifying abnormal line loss case libraries of various areas, and finally forming an abnormal line loss typical case library of files, metering, clients, equipment and other five areas; aiming at various typical case libraries, organization specialists develop line loss abnormality management seminars, design typical coping strategies and form a platform area line loss abnormality management knowledge base.

2) Construction of a Bayesian network model of a typical case base

According to the collected typical case library sample and the attribute selection method based on the gain ratio, after the original attribute set is subjected to attribute reduction, an attribute subset X= { three-phase imbalance, supercapacity, overload, voltage abnormality and current loss are obtained, wherein each attribute is used as a node variable for establishing a Bayesian network, case classification is carried out according to the reason for causing the line loss abnormality, and a Bayesian network model is respectively established for each type.

Wherein: attribute 1.., M represents the attribute subset X (three-phase imbalance, super capacity, overload etc.), cluster family 1..n represents a line loss anomaly cause category: file relationship errors, overlong power supply radius, suspected electricity larceny and the like.

1.3.2 Case library construction

Firstly, the management authority of various users to a typical case library is designed, the low-authority user only has the inquiry authority, the medium-authority user has the inquiry and modification authority, and the high-authority user has the inquiry, modification and approval authority; then, according to the collection and arrangement of the typical case library, a case library storage mode containing three parts of problems description, treatment scheme and treatment effect is designed; designing and developing an input approval process of the newly added case library; developing a function for facilitating the maintenance of a case library, wherein the case library can realize a new adding or deleting treatment scheme and can also realize the modification and deleting operation of the treatment effect; the inquiry function of the case library is developed conveniently, and single-condition inquiry and multi-condition inquiry can be realized.

1.3.3 Case library retrieval model

When the line loss of the transformer area is abnormal, extracting line loss abnormal problem description, comparing investigate into a case cases to be tested with typical cases in a case base based on the transformer area line loss abnormal case base to obtain the similarity of each domain among the cases, and realizing automatic searching and retrieval of similar cases; and simultaneously, recommending a treatment scheme of the line loss abnormality of the current area according to the case of automatic retrieval matching, and guiding the base layer to check the problem.

Case similarity calculation model

The similarity measurement is one of the cores of the case retrieval link, and the selection of a proper measurement method is helpful for quickly and accurately finding the target case. The similarity measurement method mainly uses a distance-based method, including Euclidean distance, hamming method, probability model-based measurement method, rule-based similarity measurement method and the like. The similarity between Euclidean distance metric cases is sampled herein.

Let xi and xj represent two different cases, n attributes are shared after attribute reduction, xik represents the kth attribute of case xi, and the euclidean distance is adopted:

the obtained similarity measurement formula is:

When the similarity between x _i and x _j is larger, the value of Sim (x _i,x_j) is larger, when the similarity between x _i and x _j is smaller, the value of Sim (x _i,x_j) is smaller, according to the similarity between a new case and various typical cases, the new case is pre-judged to belong to the typical cases according to the maximum similarity principle, finally, the research and judgment result is subjected to artificial examination, if the research and judgment result is correct, the case is newly added into a corresponding case family, if the judgment result is incorrect, the new case is taken as a new family in the case family, and the case is the first sample in the new family.

1.4. Healthy intelligent physical examination function construction of platform area

And carrying out daily intelligent health examination of the district based on a big data algorithm to form a diagnosis report of the intelligent health examination of the district. The diagnosis report comprises three parts of comprehensive conclusion, item indexes and diagnosis suggestions, the platform diagnosis report is displayed in a layered manner from the whole health degree of the platform, the health degree of each item index (such as equipment operation, electricity consumption, electric energy quality and the like) and specific index and label conditions, and for each key abnormality, the diagnosis report can be intelligently matched with an abnormality diagnosis case library to give specific diagnosis suggestions. Covering six-level index signboards for health intelligent examination of the area and area management and establishing a plurality of practical model tools.

1.4.1 Physical examination function of healthy intelligent area

The physical examination report mainly comprises three parts of comprehensive conclusion, item indexes and diagnosis suggestions. Wherein, the comprehensive conclusion intuitively gives the disease label and the symptom index of the area; the detailed scores of multiple indexes such as distribution network operation, platform planning, metering acquisition and the like are accurately given by the sub-index; the diagnosis proposal gives specific diagnosis and treatment advice according to the urgency of the district diseases. The emergency treatment of suspected electricity larceny, sudden power failure and other emergency treatment is carried out immediately, and the solution is carried out on the same day; chronic diseases such as unreasonable distribution load, unstable cost control and the like require work orders to be distributed and timely eliminate defects according to time limit requirements; sub-health such as the area line diameter overlength, power supply reliability low can be listed in the transformation plan, and continuous concern and early solution are realized.

1.4.2 Area management six-level index billboard

Six index signboards facing provinces, cities, counties, offices, the offices and the office managers are constructed through the health intelligent physical examination of the offices, the related index data of the offices are extracted aiming at different attention contents of management staff at all levels, and the visual and comprehensive display is carried out through the visual signboards, so that unified information inlets and working desktops for the management staff at all levels to know the operation conditions of the offices are formed. Meanwhile, a layer-by-layer analysis working mode from a visual monitoring layer, a data analysis layer and a service system layer is formed through drilling and connection of specific indexes, so that a manager can conveniently make a macroscopic decision, a foundation staff can conveniently know the health level of the managed platform area equipment in time, and further accurate operation and maintenance can be developed in a targeted mode.

1.4.3 Practical monitoring analysis model tool

① Suspected electricity larceny users and electricity larceny user feature analysis. Based on big data and calculation intelligence, the anti-electricity-theft detection analysis tool accurately researches and judges the suspected electricity-theft user, detects and analyzes the voltage, current and electric quantity change condition of the suspected electricity-theft user in real time, intuitively displays the electricity-theft behavior characteristics of the user in a multi-dimensional mode, and simultaneously combines an electricity information acquisition system and a GIS (geographic information system) to directly call a map to realize accurate positioning of the site environment and the site metering device.

② And the power failure of the station area is actively warned. The power-off event of the transformer area is utilized, the superior power-off (including planning and faults) is filtered through the topology relation of 'station-line-transformer-user', the fault research and judgment of a main line and a distribution line, meanwhile, the power-off fault of the transformer area is confirmed through the distribution transformer and the active calling detection of typical users under the transformer area, the power-off condition of the transformer area is detected and alarmed in real time, and a fault work order is pushed.

③ And (5) fault report, repair, study and judgment. And accessing a user repair work order, and carrying out study and judgment on fault types such as single-user, multi-user or station power failure by combining with upper-level scheduled maintenance and fault information such as arrearage information of marketing users, power distribution line outage, distribution transformer outage and the like through carrying out test of a user table, carrying out test of other typical users in the same station area, distribution transformer test and the like.

④ And (5) analyzing and treating three-phase unbalance of the transformer area. The phase voltage is distinguished through the access station, and the three-phase unbalanced fault is actively researched, judged and monitored for the distribution transformer which is monitored; and (3) through collecting the information of the household table under the transformer area, studying and judging the phase connected with the household table, pushing the fault work order, and combining with the GIS topological graph, assisting in developing the three-phase imbalance treatment work.

1.5. Tracking analysis of abnormal relevance of line loss of transformer area

And (3) through the correlation analysis of the daily line loss of the station area and the daily electricity consumption of the users under the station area, finding out the abnormal fluctuation condition of the daily electricity consumption of the users under the abnormal condition of the line loss of the station area, finding out the users with high correlation degree with the abnormal condition of the line loss of the station area, and realizing the tracking analysis of the abnormal users of the power consumption of the station area.

1.5.1 Correlation analysis model for line loss of transformer area and power consumption of users under transformer area

Aiming at the abnormal line loss area, firstly, daily electricity quantity information of all users under the daily line loss area and the area of the area is collected, and the correlation coefficient (r) between the daily electricity quantity information and the daily electricity quantity information is calculated, so that the correlation degree between the daily electricity quantity information and the area is achieved. Wherein the calculation formula of the correlation coefficient r is as follows:

Judging the degree of closeness of the correlation by using the correlation coefficient is generally considered as:

The value of the correlation coefficient r	Degree of correlation
		\|r\|＝0	Completely uncorrelated
0<\|r\|<＝0.3	Weak correlation
		0.3<\|r\|<＝0.5	Low degree of correlation
0.5<\|r\|<＝0.8	Significant correlation
		0.8<\|r\|<1	Highly correlated
\|r\|＝1	Complete correlation

And according to the correlation coefficient, focusing on the users with the correlation coefficient being obviously correlated to the line loss abnormal area, and performing field investigation.

1.5.2 Suspected Power theft diagnostic model

The electricity stealing is one of the main factors affecting the line loss of the transformer area, and the main method for electricity stealing comprises two methods of winding the metering device and destroying the metering accuracy of the electric energy meter. The former marketer can be basically identified by naked eyes, and the latter can establish a suspected electricity larceny diagnosis model. 1) Firstly, finding out abnormal line loss areas according to the line loss rate of the areas, and then counting an electric energy meter list with cover opening time (if the electric energy meter list can be acquired) later than installation time for electricity consumption data of a concentrator, a terminal and a client, wherein the client has suspicion of privately opening the meter; 2) And (3) counting, comparing and analyzing the same-period electricity consumption of the customers in the abnormal line loss area, and automatically identifying customers with greatly reduced daily electricity consumption, wherein the customers are important suspected electricity larceny customers.

1.6. Visual display of large-screen of line loss of transformer area

And the visual display of the line loss chart of the transformer area is realized from four levels of province companies, city companies, county companies and power supply stations. The refined subdivision results (such as high loss, high negative loss, exemption platform areas and the like) of the platform areas are distributed and displayed on the areas; the line loss rate is distributed in blocks (city, county) by thermodynamic diagram and other modes. And obtaining a distribution line topological graph and corresponding equipment file information from a marketing system and a GIS system, combing line loss display contents of the areas from four levels of province companies, city companies, county companies and power supply stations, combining the results of the fine classification of the low-voltage areas and the construction of the health intelligent experience functions of the reasonable line loss areas and the areas, and displaying the fine results (such as high loss, high negative loss, exemption areas and the like) of the areas in a thermodynamic diagram mode.

1.6.1 Theoretical line loss and reasonable line loss interval display

According to the selected management units, the platform numbers, platform responsible persons, the platform types and time options, the theoretical line loss calculated value and the line loss reasonable interval value of the platform of the management unit of the current next level are displayed, and meanwhile, the platform operation characteristic indexes such as the total active/reactive power of the platform, the three-phase balance degree, the voltage drop and the like are displayed.

1.6.2 Theoretical line loss calculation rate display

And displaying theoretical line loss calculation rate display of the current next-level management unit station area according to the selected management unit and time (month or day) options, and displaying station area indexes such as the total number of station areas, the total number of calculated station areas, the number of station areas with failed voltage acquisition and the like.

1.6.3 Zone type distribution display

The distribution of the region types is presented in the form of a thermodynamic diagram based on the selected management units, region type (e.g., high loss, high negative loss, exempt region, etc.) options.

The system architecture of the method mainly comprises a full-service data center, a low-voltage area holographic portrait platform and a service application module.

The technical architecture design of the system of the method follows the 3-layer architecture of a data layer, a business layer and a presentation layer of software development. Based on J2EE technology, corresponding service application is constructed, and the software function is fully and repeatedly utilized to meet the service requirement.

In the deployment architecture of the method system, a tag library system is deployed in an information intranet of a provincial company, and 2 database servers and 1 application server are adopted.

The database server directly adopts dual-machine hot standby to realize high availability of the database. Basic data of the database is extracted from a marketing system, a power consumption acquisition system, a full-service data center and other systems or platforms through the ETL.

The application server deploys related applications of the tag library and accesses the database in a JDBC mode.

The data architecture based on the big data algorithm platform area intelligent health diagnosis method and system relies on the platform area basic information, the platform area power consumption information, the platform area abnormal information and the platform area label information full-service data of the system platform area basic information, the platform area power consumption information, the platform area abnormal information such as a marketing service application system, a GIS system and a PMS2.0, and forms relevant labels and analysis data by combining the big data calculation, data mining and other technologies, and provides support for various marketing informatization systems through data sharing service.

The security architecture of the method system conforms to the national network security standard and requirement, and has a security management mechanism according to the file requirements of national grid company application software general security requirement, national grid company information system security protection general scheme, national grid company information system management method, implementation guidance opinion about information security level protection construction and the like, so that the information storage security, information transmission and processing security are guaranteed, and the system can normally operate, is not accessed by unauthorized and is not damaged by attack. The security policy follows the following principle:

(1) The information security overall strategy of the national power grid company is followed, and the information security requirement of the national power grid company is met;

(2) The safety protection strength reaches the safety protection standard of the information intranet of the national power grid limited company;

(3) The operation safety is paid attention to, and the safety risk diffusion is avoided;

(4) And the safety management and the safety protection measures are repeated.

Key technology of the method and the system is as follows:

1 cache technology

The caching technology is a technology for locally storing frequently accessed information, and can effectively solve the problem of impact on a database caused by massive concurrent data access. Mainly comprises the following applications:

1) The WEB server caches the application, so that the request pressure on a back-end application server is effectively reduced, and the page access response efficiency is improved;

2) The database caching application reduces the query pressure to the database;

3) The application program caches the application, so that the I/O pressure on the file system is reduced;

The client browser caches the application, reduces access pressure to the website, and reduces the transmission amount on the WAN link and the Web server.

2 Load balancing technique

The load balancing technology is a performance optimization technology, ensures that a large amount of concurrent access or data traffic is shared to a plurality of node devices for processing respectively, reduces the waiting response time of users, and is mainly applied to Web servers, FTP servers, enterprise key application servers and the like.

3 Area line loss calculation and area cluster analysis

(1) Area line loss calculation

A. line loss calculation by mesa technique ("pressure drop method")

Calculating theoretical line loss of the transformer area according to a voltage drop method, and calculating a reasonable interval value of the transformer area line loss of a 95% confidence interval according to the statistical condition of the transformer area 'stable day' technical line loss on the assumption that the transformer area line loss rate is a normal distribution random variable:

B. area cluster analysis

(1) KNN algorithm

KNN is classified by measuring the distance between different eigenvalues. The idea is as follows: if a sample belongs to a class for the majority of the k most similar (i.e., nearest neighbor) samples in the feature space, then the sample also belongs to that class. K is typically an integer no greater than 20. In the KNN algorithm, the selected neighbors are all objects that have been correctly classified. The method only determines the category to which the sample to be classified belongs according to the category of one or more samples which are nearest to each other in the classification decision.

The KNN algorithm implementation steps:

1) Calculating the distance between the test data and each training data;

2) Sorting according to the increasing relation of the distance;

3) Selecting K points with the smallest distance;

The most common distance representation between two points or between multiple points, also known as euclidean metric, is defined in euclidean space as the distance between points x= (x ₁,...,x_n) and y= (y ₁,...,y_n) is:

4) Determining the occurrence frequency of categories of the first K points;

5) And returning the category with highest occurrence frequency in the first K points as the prediction classification of the test data.

(2) Analysis of key influence factors of line loss of transformer area

In the variable screening, a correlation coefficient method, an information gain method, or the like can be used.

Correlation coefficient method

The Pearson product difference correlation coefficient (commonly known as the simple correlation coefficient) measures the degree of linear correlation between two quantitative variables. Simple correlation coefficients are affected by other factors and reflect often immaterial relationships. A simple correlation coefficient between two variables is large does not mean causality, since it is entirely possible to make a high correlation due to the intermediation of the other variable.

Information gain

In machine learning, one concept that is not bypassed is entropy (Entropy), information entropy. Information entropy is often used as a quantitative indicator of the information content of a system and thus can be further used as a target for system equation optimization or as a criterion for parameter selection. Entropy is used as a criterion for sample optimal attribute partitioning in the decision tree generation process.

Entropy definition:

The parent claude shannon of the information theory gives three important properties of information entropy: 1) Monotonicity, the higher the occurrence probability, the lower the amount of information carried by the event; 2) The non-negativity, information entropy can be regarded as a breadth measure, and the non-negativity is a reasonable necessity; 3) The cumulative, i.e., the measure of total uncertainty that multiple random events occur simultaneously, is a representation of the sum of the measures of uncertainty of the events, which is also a broad measure.

The larger the entropy is, the larger the uncertainty of the random variable is, and when the variable can take a certain value, the probability distribution of each value is taken to be averaged, and the entropy value is larger. Wherein the decision tree (ID 3 algorithm and C4.5 algorithm) is a split of the selected attributes based on the entropy of the information. When the fine classification of the areas is carried out, key factors influencing the line loss can be selected according to the information entropy.

C. Reasonable line loss interval of transformer area ('regression model')

Calculating a reasonable line loss interval of the station area, wherein the reasonable line loss interval is mainly calculated according to the following steps:

1) Carrying out KNN clustering on each subarea data after data preprocessing, wherein the clustering number is set to be 2-15;

2) Selecting the cluster number with the best cluster quality according to the size of the contour value, regarding the cluster number as an optimal cluster result, analyzing the optimal cluster result to obtain each cluster center, and verifying whether the clusters are reasonable or not;

3) According to the optimal clustering result, respectively carrying out linear regression on each class to obtain a linear regression expression, analyzing the regression result, and judging whether the linear regression model is reasonable or not;

4) Making a histogram of the predicted measurement and the actual measurement, checking whether the error has normal distribution characteristics, verifying the rationality of the linear regression model, and simultaneously giving a scatter diagram between the predicted measurement and the actual measurement so as to intuitively check the relation between the predicted variable and the actual measurement;

5) And obtaining a line loss rate qualification range with 90% confidence space.

The above formulas are all formulas with dimensions removed and numerical values calculated, the formulas are formulas with a large amount of data collected for software simulation to obtain the latest real situation, and preset parameters in the formulas are set by those skilled in the art according to the actual situation.

It should be understood that the term "and/or" is merely an association relationship describing the associated object, and means that three relationships may exist, for example, a and/or B may mean: there are three cases, a alone, a and B together, and B alone, wherein a, B may be singular or plural. In addition, the character "/" herein generally indicates that the associated object is an "or" relationship, but may also indicate an "and/or" relationship, and may be understood by referring to the context.

It should be understood that, in various embodiments of the present application, the sequence numbers of the foregoing processes do not mean the order of execution, and the order of execution of the processes should be determined by the functions and internal logic thereof, and should not constitute any limitation on the implementation process of the embodiments of the present application.

Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application. It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described systems, apparatuses and units may refer to corresponding procedures in the foregoing method embodiments, and are not repeated herein.

The foregoing is merely illustrative of the present application, and the present application is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A big data algorithm-based intelligent health diagnosis method for a platform region is characterized by comprising the following steps of: the diagnostic method comprises the steps of:

2. The intelligent health diagnosis method based on big data algorithm platform area according to claim 1, which is characterized in that: the method for establishing the area holographic image by combining the image technology comprises the following steps:

3. The intelligent health diagnosis method based on big data algorithm platform area according to claim 2, which is characterized in that: the construction of the platform area line loss abnormity diagnosis model comprises the following steps:

4. The intelligent health diagnosis method based on big data algorithm platform area according to claim 3, wherein the method is characterized in that: the cluster analysis and the calculation of the line loss interval of the transformer area comprise the following steps:

5. A big data algorithm-based intelligent health diagnosis system for implementing the diagnosis method of any one of claims 1 to 4, characterized in that: the system comprises an analysis module, a calculation module, a positioning module, a model construction module and a case library construction module;