CN116341788A - A Power Fingerprint Precise Governance Method for Distribution Network Line Loss Analysis - Google Patents

Abstract

The invention relates to an accurate electric fingerprint management method for analyzing the line loss of a power distribution network, which comprises the following steps: s1, data acquisition; the acquired data comprise operation information data, and the required data are acquired according to a data application flow; s2, data processing: when the correlation between the line loss rate and the electricity consumption of the hanging user is analyzed, the sudden change data is processed; carrying out modal decomposition on the data signals of the line loss and the electric quantity, and taking out user data in advance; filling data of the missing electricity consumption data by using a Lagrangian interpolation method; s3, extracting and analyzing 'electric fingerprint'; s4, modeling analysis; s5, grading the suspicion degree of the abnormal user as a foothold point, and establishing a method strategy model.

Description

A Power Fingerprint Precise Governance Method for Distribution Network Line Loss Analysis

technical field

The invention belongs to the technical field of intelligent management of power grid line loss analysis, and in particular relates to a precise power fingerprint management method for distribution network line loss analysis.

Background technique

Most of the grid loss occurs in the medium and low voltage distribution network. The management of the distribution network line and the line loss of the station area can not only reduce the power loss, but also detect the abnormality of the metering device, the imbalance of the three-phase load, and theft of electricity in time. Abnormal behavior of electricity consumption. Therefore, line loss management is directly related to the economic benefits of enterprises and the implementation of national energy-saving policies, and how to reduce line loss has become the key work of power supply enterprises and the key research object of electric power workers.

At present, the intelligent technical means of line loss analysis used by power grid companies mainly realizes line loss diagnosis and locates abnormal line loss problems through functional modules such as the simultaneous line loss system and the power consumption information collection system, such as station area physical examination and loss reduction closed-loop management. It still has its limitations. It lacks a set of relationship model solutions that focus on line loss and power from the perspective of big data, and digs out the accurate management method of line loss to identify the problems of line loss and power data, so as to improve the efficiency of line loss management.

Contents of the invention

In order to solve the problems in the background technology, effectively reduce the manual workload of line loss management, and improve the efficiency of line loss management, the present invention provides a precise power fingerprint management method for distribution network line loss analysis, which realizes the location of line loss abnormal points and Suspicion diagnosis and early warning for possible abnormalities have the advantages of reducing the manual workload of line loss control and improving the efficiency of line loss control.

In order to achieve the above objectives, the technical solution adopted by the present invention is: a method for accurately managing power fingerprints for distribution network line loss analysis, including the following steps:

S1. Data collection; the collected data includes operation information data, and the required data is collected according to the data application process;

S2. Data processing: when analyzing the correlation between the line loss rate and the power consumption of the downlink users, for the sudden change data processing; for the data signal of the line loss and power consumption, the modal decomposition is performed, and the user data is taken out in advance; for missing The electricity consumption data is filled with Lagrangian interpolation method;

S3. "Power fingerprint" extraction and analysis:

3.1) Bring fingerprint recognition into the generation of power curve data, convert the continuous data of each level object including the station area into a map texture, and fit the map texture to form a real distribution network line loss change state Atlas, that is, the "power fingerprint" of line loss, constructs line loss and electric quantity fingerprints respectively;

3.2) Sampling and selecting the "power fingerprint" of the station area with abnormal line loss, analyzing the characteristics of the extracted fingerprint from the perspective of central tendency, degree of dispersion, and descriptive statistics of distribution characteristics, and intuitively presenting the influence relationship between power and line loss;

S4. Modeling analysis:

4.1) Fingerprint correlation analysis model: By carrying out the correlation analysis between the fluctuation of the line loss fingerprint in the station area and the fluctuation of the fingerprint of the connected user's electricity, locate the suspected user who affects the abnormal fluctuation of the line loss in the station area;

4.2) The analysis model of fingerprint fluctuation variation: through the analysis of the inflection point of the change of the fingerprint fluctuation of the line loss in the station area and the fingerprint fluctuation of the user's electricity under the connection, locate the suspected user who affects the abnormal fluctuation of the line loss in the station area;

4.3) Empirical mode decomposition model of fingerprint signal: by analyzing the time-frequency signal analysis of the fingerprint fluctuation of the line loss in the station area and the fingerprint fluctuation of the electricity quantity of the downlink user, locate the suspected user who affects the abnormal fluctuation of the line loss in the station area;

S5. Based on the grading of the suspicion degree of abnormal users as a foothold, a method strategy model is established.

Further, in step S1, the operation information data includes the file information of 10 kV distribution network line, station area, high and low voltage users, and daily frozen power and line loss;

In step S2, the data of users whose daily average power consumption is 0 and daily average power consumption is less than 1kW·h are taken out in advance;

In step S3, the continuous data of each hierarchical object includes data of lines, stations, and users.

Further, the modeling process of step 4.1): ①data preprocessing, including deduplication of user electricity consumption data, eliminating users with zero daily electricity consumption, and deleting useless fields; ②extracting fingerprints, including station area lines The fingerprint curve map of the loss rate and the electricity consumption of the connected users; ③Calculate the correlation coefficient, and use Pearson and other correlation coefficient methods to quantitatively analyze the correlation degree between each user’s power fingerprint and the line loss rate fingerprint of the station area; ④Locate the strongly related users, Using the results of the correlation coefficient between the power consumption of each user and the line loss of the station area, screen and locate the abnormal suspect users whose power fingerprint and line loss fingerprint have a strong correlation in the same direction.

Further, the modeling process of step 4.2): ①Data preprocessing, including deduplication of user electricity consumption data, elimination of invalid data, and splicing of wide data tables; ②Lock the date of the inflection point of change, and select the most obvious change in the line loss curve of the station area ③Calculate the k value of the inflection point, and use the defined "K value = user power change amount/station area loss power change" model rule to calculate the K of the station area line loss power change rate caused by each user's power change speed ④ Locate abnormal users, use the K value results of each user's electricity consumption and station area line loss to screen and locate abnormal suspected users with large relative deviations from station area line loss anomalies.

Further, the modeling process of step 4.3): ①Data preprocessing, including deduplication of user electricity consumption data, Lagrangian interpolation for missing data, calculation of daily average electricity consumption, and wide table splicing; ②Calculation of user electricity consumption Correlation coefficient r between power consumption and station area line loss rate, and sort the results based on daily average power consumption and correlation coefficient r; ③ Select the top 5% users as the primary screening users, and use the EMD algorithm to analyze the user power consumption and Decompose the power consumption in the station area for signal mode decomposition, extract high-frequency components (IMF) respectively, and complete signal spectrum fitting; ④ mark abnormal users, use the empirical mode decomposition results of fingerprint signals to screen and locate suspected users with abnormal line loss.

Still further, the step S5 includes:

Method Strategy model construction: The three major distribution network line loss "power fingerprint" abnormal user identification algorithm models are "fingerprint correlation analysis model", "fingerprint fluctuation variation analysis model", and "fingerprint signal empirical mode decomposition model". Locate the scope of suspected users that have a strong correlation with line loss abnormalities, and divide the suspicion degree of abnormal users into gradient labels; according to the division results of users under each analysis model, combined with the user's own operating data during normal and abnormal online loss periods, Carry out power trend analysis, abnormal event analysis, and equipment operation status analysis to assist in the verification of specific abnormal power consumption behaviors and occurrence periods;

Application process: ① Import the file information of 10 kV distribution network lines, station areas, high and low voltage users, and data related to power consumption and line loss into the three major line loss abnormal user identification algorithm models; ② Output model results: based on the three models Respectively output the scope of suspected users that have a strong correlation with line loss anomalies; ③ match and fuse the three major model result labels to complete the comprehensive rating of the degree of suspicion of abnormal users, and differentiate the suspected users; ④ for different categories of suspected users, Execute by category until the loss reduction target is achieved.

Furthermore, the modeling analysis in step S4 is based on the correlation analysis of line loss and power consumption for abnormal user positioning business implementation logic and early data processing, frame the scope of machine learning, deep learning big data related algorithms, and use Python training to affect the effect between algorithms , Efficiency and stability for comparative analysis, and finally based on the power and line loss spectrum fingerprints, complete the "Fingerprint Correlation Analysis Model", "Fingerprint Fluctuation Variation Analysis Model", and "Fingerprint Signal Empirical Mode Decomposition Model" The training and construction of the algorithm model of line loss abnormal user identification.

The technical effect of the present invention lies in: the precise power fingerprint management method of the distribution network line loss analysis of the present invention realizes the location of the abnormal point of the line loss and conducts suspicious diagnosis and early warning on whether the abnormality may occur, and has the ability to reduce the manual work of line loss management The advantages of improving the efficiency of line loss management.

Description of drawings

Fig. 1 is the fingerprint spectrum graph of line loss rate A day in Taiwan area in the specific embodiment of the present invention;

Fig. 2 is a graph of the fingerprint spectrum graph of electricity supply and sales in Taiwan District A day in a specific embodiment of the present invention;

Fig. 3 is a graph showing the correlation analysis between power consumption and line loss of user a under the station area in a specific embodiment of the present invention;

Fig. 4 is the inflection point fingerprint graph of line loss change in Taiwan area A day in the specific embodiment of the present invention;

Fig. 5 is a high-frequency component diagram of the fingerprint signal of user b under the platform area in a specific embodiment of the present invention;

Fig. 6 is a flow chart of the application of the "power fingerprint" precise management method for distribution network line loss analysis in a specific embodiment of the present invention;

Fig. 7 is a schematic diagram of a correlation modeling algorithm in a specific embodiment of the present invention.

Detailed ways

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some, not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

A precise power fingerprint management method for distribution network line loss analysis, comprising the following steps:

S1. Data collection: The collected data includes the file information of 10kV distribution network lines, station areas, high and low voltage users, daily frozen power, line loss and other operational information data, and collects the required data according to the data application process;

S2. Data processing: When analyzing the correlation between the line loss rate and the power consumption of the downlink users, it is also valuable to analyze sudden changes such as sudden increase and decrease, without the need to process abnormal values; for the data of line loss and power consumption The signal is modal decomposed, and user data such as "zero power consumption" (the average daily power consumption is 0) and "small power consumption" (the average daily power consumption is less than 1kW h) are taken out in advance to avoid affecting the accuracy of the algorithm and Applicability; use Lagrangian interpolation method to fill in missing electricity consumption data;

S3. "Power fingerprint" extraction and analysis:

3.1) Bring fingerprint recognition into power curve data generation, convert the continuous data of various levels of objects such as lines, stations, users, etc. into map textures, and fit the map textures to form real distribution network line losses The map of the changing state, that is, the "power fingerprint" of the line loss, respectively constructs the line loss and electric power fingerprints;

3.2) Sampling and selecting the "power fingerprint" of the abnormal line loss station area, analyzing the characteristics of the extracted fingerprint from the perspective of descriptive statistics such as central tendency, degree of dispersion, and distribution characteristics, and intuitively presenting the influence relationship between power and line loss; analysis method It is also applicable to the analysis of other 10 kV distribution network lines and station areas;

S4. Modeling analysis:

4.1) Fingerprint correlation analysis model: By carrying out the correlation analysis between the fluctuation of the line loss fingerprint in the station area and the fluctuation of the power fingerprint of the connected user, locate the suspected user who affects the abnormal fluctuation of the line loss in the station area.

Modeling process: ① Data preprocessing, including deduplication of user power consumption data, eliminating users with zero daily power consumption, deleting useless fields, etc.; Electricity fingerprint curve; ③Calculate the correlation coefficient, and use Pearson and other correlation coefficient methods to quantitatively analyze the degree of correlation between each user’s electricity fingerprint and the line loss rate fingerprint of the station area; ④Locate strongly related users, and use the electricity consumption of each user Based on the results of the correlation coefficient with the line loss in the station area, screen and locate abnormal suspects whose power fingerprints and line loss fingerprints are strongly correlated with changes in the same direction;

4.2) Analysis model of fingerprint fluctuation variation: By analyzing the inflection point of the change of fingerprint fluctuation of line loss in the station area and the fingerprint fluctuation of the electricity quantity of the connected users, the suspected user who affects the abnormal fluctuation of line loss in the station area is located.

Modeling process: ①Data preprocessing, including deduplication of user electricity consumption data, elimination of invalid data, data wide table splicing, etc.; ②Lock the date of the inflection point of change, and select the most obvious mutation point of the line loss curve in the station area; ③Calculate the inflection point K value, using the defined model rule of "K value = change of user's power quantity/change of power consumption in the station area" to calculate the K value of the change speed of line loss power in the station area caused by the change speed of each user's power; ④ locate abnormal users, Use the K value results of each user's electricity consumption and the line loss of the station area to screen and locate the abnormal suspect users with a large relative deviation from the abnormal line loss of the station area;

4.3) Empirical mode decomposition model of fingerprint signal: By analyzing the time-frequency signal analysis of the fingerprint fluctuation of the line loss in the station area and the fingerprint fluctuation of the electricity quantity of the downlink user, the suspected user who affects the abnormal fluctuation of the line loss in the station area is located.

Modeling process: ① Data preprocessing, including deduplication of user electricity consumption data, Lagrangian interpolation for missing data, daily average electricity consumption calculation, wide table splicing, etc.; ② Calculation of user electricity consumption and station area lines The correlation coefficient r of the loss rate, and the results are sorted based on the daily average power consumption and the correlation coefficient r; ③ The top 5% users are selected as the primary screening users, and the signal simulation of the user power consumption and the power consumption of the station area is carried out through the EMD algorithm. modal decomposition, extract high-frequency components (IMF) respectively, and complete signal spectrum fitting; ④ mark abnormal users, and use fingerprint signal empirical mode decomposition results to screen and locate suspected users with abnormal line loss;

S5. Establishing a precise governance method strategy model is to provide a set of differentiated governance strategies (models) and the application process of model results from the perspective of how big data modeling and analysis results are implemented and applied based on the classification of abnormal user suspicion levels:

Method Strategy model construction: using ("fingerprint correlation analysis model", "fingerprint fluctuation analysis model", "fingerprint signal empirical mode decomposition model") three major distribution network line loss "power fingerprint" abnormal user identification algorithm models , respectively locate the range of suspected users that have a strong correlation with the line loss anomaly, and perform gradient label division for the degree of suspicion of abnormal users, and the division rules are determined by each model rule. Based on the user's division results under each analysis model, combined with the user's own operating data during normal and abnormal periods of online loss, analysis of power variation trend analysis, abnormal event analysis, and equipment operating status analysis are carried out to assist in verifying specific power consumption. Abnormal behavior and occurrence time, assisting in loss reduction management planning.

Application process: ① Import the file information of 10 kV distribution network lines, station areas, high and low voltage users, power consumption, line loss and other related data into ("Fingerprint Correlation Analysis Model", "Fingerprint Fluctuation Variation Analysis Model" , "Fingerprint Signal Empirical Mode Decomposition Model") three major line loss abnormal user identification algorithm models; ② output model results: based on the three models, respectively output the range of suspected users that have a strong correlation with line loss anomalies; ③ matching and fusion of the three major The model result label completes the comprehensive rating of the degree of suspicion of abnormal users, and differentiates the suspected users (such as dividing into "key attention, general attention, and non-key attention"); According to the precise management strategy,) classification (policy) implementation, until the loss reduction target is achieved.

The present invention relates to a power fingerprint precision management method for distribution network line loss analysis. This method aims at problems such as low efficiency of manual investigation, large amount of basic data, and difficulty in diagnosis of abnormal root causes in the process of line loss management by grassroots personnel, relying on multi-algorithm fusion Advanced big data modeling and analysis technology, focusing on in-depth analysis and research on the relationship between line loss and power time series fluctuations, and constructing a set of distribution network line loss analysis "power fingerprint" big data analysis model and precise management methods to help locate users suspected of abnormal line loss And analysis, providing differentiated and precise management strategies, thereby reducing the manual workload of line loss management and improving the efficiency of line loss management.

Based on the specific embodiment of the present invention, it is only a preferred specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any equivalent replacement or change of the technical solution of the invention and its inventive concept shall fall within the protection scope of the present invention.

Specifically, a precise power fingerprint management method for distribution network line loss analysis includes the following steps:

S1. Data collection: the data required for collection includes file information of 10kV distribution network lines, station areas, high and low voltage users, daily frozen power, line loss and other operational information, and the required data is collected according to the data application process;

S2. Data processing: When analyzing the correlation between the line loss rate and the power consumption of the downlink users, it is also valuable to analyze sudden changes such as sudden increase and decrease, without the need to process abnormal values; for the data of line loss and power consumption The signal is modal decomposed, and user data such as "zero power consumption" (the average daily power consumption is 0) and "small power consumption" (the average daily power consumption is less than 1kW h) are taken out in advance to avoid affecting the accuracy of the algorithm and Applicability; use Lagrangian interpolation method to fill in missing electricity consumption data;

S3. "Power fingerprint" extraction and analysis:

3.1) Bring fingerprint recognition into power curve data generation, convert the continuous data of various levels of objects such as lines, stations, users, etc. into map textures, and fit the map textures to form real distribution network line losses The map of the changing state, that is, the "power fingerprint" of the line loss, respectively constructs the line loss and electric power fingerprints;

3.2) Sampling and selecting the "power fingerprint" of the abnormal line loss station area, analyzing the characteristics of the extracted fingerprint from the perspective of descriptive statistics such as central tendency, degree of dispersion, and distribution characteristics, and intuitively presenting the influence relationship between power and line loss; specific implementation Refer to Figures 1 and 2 for the results; this step is also applicable to the analysis of other 10 kV distribution network lines and station areas;

S4. Modeling analysis: Based on the correlation analysis of line loss and power consumption, the business implementation logic and early data processing of abnormal user location are framed, and the scope of machine learning and deep learning big data related algorithms is framed. The effect, efficiency and The advantages and disadvantages of the stability are compared and analyzed, and finally based on the fingerprint of the electric quantity and the line loss map, the three major line loss anomalies of the "fingerprint map correlation analysis model", "fingerprint fluctuation variation analysis model", and "fingerprint signal empirical mode decomposition model" are completed Training and construction of user identification algorithm model:

4.1) Fingerprint correlation analysis model: By carrying out the correlation analysis between the fluctuation of the line loss fingerprint in the station area and the fluctuation of the power fingerprint of the connected user, locate the suspected user who affects the abnormal fluctuation of the line loss in the station area. Modeling process: ① Data preprocessing, including deduplication of user electricity consumption data, elimination of users with zero daily electricity consumption (8 in total), and deletion of useless fields; ② Fingerprint extraction, including line loss rate and Connect the fingerprint curve of the user's power consumption; ③ calculate the correlation coefficient, and use Pearson and other correlation coefficient methods to quantitatively analyze the degree of correlation between each user's power fingerprint and the line loss rate fingerprint of the station area; ④ locate strongly related users, use each The results of the correlation coefficient between the user's electricity consumption and the line loss of the station area are used to screen and locate abnormal suspected users whose electricity fingerprints and line loss fingerprints are strongly correlated with changes in the same direction; refer to Figure 3;

4.2) Analysis model of fingerprint fluctuation variation: By analyzing the inflection point of the change of fingerprint fluctuation of line loss in the station area and the fingerprint fluctuation of the electricity quantity of the connected users, the suspected user who affects the abnormal fluctuation of line loss in the station area is located. Modeling process: ①Data preprocessing, including deduplication of user electricity consumption data, elimination of invalid data, data wide table splicing, etc.; ②Lock the date of the inflection point of change, and select the most obvious mutation point of the line loss curve in the station area; ③Calculate the inflection point K value, use the defined model rules to calculate the K value of the change speed of power loss in the station area caused by the power change speed of each user; ④ locate abnormal users, use the K value results of the power consumption of each user and the line loss of the station area , screening and locating abnormal suspected users with a large relative deviation from line loss anomalies in the station area; refer to Figure 4;

4.3) Empirical mode decomposition model of fingerprint signal: By analyzing the time-frequency signal analysis of the fingerprint fluctuation of the line loss in the station area and the fingerprint fluctuation of the electricity quantity of the downlink user, the suspected user who affects the abnormal fluctuation of the line loss in the station area is located. Modeling process: ① Data preprocessing, including deduplication of user electricity consumption data, Lagrangian interpolation for missing data, daily average electricity consumption calculation, wide table splicing, etc.; ② Calculation of user electricity consumption and station area lines The correlation coefficient r of the loss rate, and the results are sorted based on the daily average power consumption and the correlation coefficient r; ③ The top 5% users are selected as the primary screening users, and the signal simulation of the user power consumption and the power consumption of the station area is carried out through the EMD algorithm. Mode decomposition, extracting high-frequency components (IMF) respectively, and completing signal spectrum fitting; ④ mark abnormal users, and use fingerprint signal empirical mode decomposition results to screen and locate suspected users with abnormal line loss; refer to Figure 5;

S5. Establish a precise governance method strategy model: take the abnormal user suspicion level as the foothold, and establish a method strategy model;

Method Strategy model construction: The three major distribution network line loss "power fingerprint" abnormal user identification algorithm models are "fingerprint correlation analysis model", "fingerprint fluctuation variation analysis model", and "fingerprint signal empirical mode decomposition model". Locate the range of suspected users that have a strong correlation with line loss anomalies, and divide the suspected degree of abnormal users into gradient labels.

Match and fuse the result labels of the three major models to complete the comprehensive rating of the degree of suspicion of abnormal users; refer to the following table:

Specific application process: ① Import the file information of 10 kV distribution network lines, station areas, high and low voltage users, power consumption, line loss and other related data into the "Fingerprint Correlation Analysis Model" and "Fingerprint Fluctuation Variation Analysis Model" , "Fingerprint Signal Empirical Mode Decomposition Model" three major line loss abnormal user identification algorithm models; ②Output model results: based on the three models, respectively output the range of suspected users that have a strong correlation with line loss abnormalities; ③Matching and fusion of the three major models The result label completes the comprehensive rating of the degree of suspicion of abnormal users, and differentiates the suspected users (specifically divided into "key attention, general attention, and non-key attention"); precise governance strategy,) implement classified policies until the loss reduction target is achieved; refer to Figure 6.

The above step S4 modeling algorithm, in order to find the user with strong correlation with the line loss fluctuation in the station area (take the electricity consumption of user Wang as an example), use the Pearson correlation coefficient and Spearman correlation analysis method to calculate each Correlation between user electricity and station area line loss rate, and a correlation coefficient threshold is given. If the absolute value of the correlation coefficient is greater than the given threshold, it is defined as the abnormal correlation between the user electricity and station area line loss;

The correlation coefficient is the measure of the degree of linear correlation between the studied variables, denoted by the letter r. Correlation tables and correlograms reflect the relationship between two variables and their direction of correlation, but cannot exactly indicate the degree of correlation between two variables. The correlation coefficient is a statistical indicator used to reflect the closeness of the correlation between variables. The correlation coefficient is calculated according to the product difference method, based on the deviation between the two variables and their respective average values, and the degree of correlation between the two variables is reflected by multiplying the two deviations. The formula is as follows (where Cov(X,Y) is X Covariance with Y, Var[X] is variance of X, Var[Y] is variance of Y).

The absolute value of r is between 0 and 1. The closer |r| is to 1, the stronger the degree of correlation between the two variables, conversely, the closer |r| is to 0, the weaker the degree of correlation between the two variables; the correlation coefficient r result classification table:

The above steps are based on user suspicion classification of fingerprint correlation analysis:

Based on the fingerprint correlation analysis model, the correlation coefficient analysis and calculation between the line/station area line loss and the power consumption of the connected users is realized, and the user suspicion is differentiated according to the correlation coefficient value. The details are as follows:

The |r| value is above 0.9 as "very strong correlation", the |r| value is between 0.7-0.9 as "strong correlation", and the |r| value is between 0.4-0.7 as "moderate correlation". The value of |r| between 0.2-0.4 is "weak correlation", and the value of |r| below 0.2 is "very weak correlation".

Pay attention to users whose labels are classified as "extremely relevant", "strongly relevant", and "moderately relevant", and check whether such users have abnormal power consumption.

The above steps are based on user suspicion classification based on fingerprint fluctuation analysis:

Based on the analysis model of fingerprint fluctuation variation, the calculation of the relationship between the line/station area line loss and the power change speed of the downlink user is realized, that is, the K value calculation of the line/station area line loss power change speed caused by the change of power consumption. According to the K value, the two classifications of user suspicion are carried out, as follows:

When the K value is positive, it means that the user has an abnormal addition to the line/station area line loss, and further diagnoses and analyzes whether there is abnormal power consumption behavior as an abnormal suspected user. When the K value is larger, it means that the relative deviation degree of the user's power consumption to the abnormal line loss is larger.

The above steps are based on user suspicion classification of fingerprint signal empirical mode decomposition:

Based on the fingerprint signal empirical mode decomposition model, the high-frequency signal analysis of line loss/station area line loss fingerprint fluctuation and downlink user power fingerprint fluctuation is realized. Two classifications of user suspicions are performed according to the coincidence degree of signal waveforms, as follows:

When the coincidence degree of the high-frequency component of the user's power consumption and the high-frequency component of the power consumption of the line/station area is relatively high, the user as an abnormal suspect is further diagnosed and analyzed for abnormal power consumption behavior. The higher the coincidence degree of the two waveforms, the higher the user's suspicion of abnormality.

It should be noted that the term "comprising", "comprising" or any other variation thereof is intended to cover a non-exclusive inclusion such that a process, method, article or apparatus comprising a set of elements includes not only those elements, but also Also includes other elements not expressly listed, or which are inherent in the process, method, article, or device.

Although the embodiments of the present invention have been shown and described, those skilled in the art can understand that various changes, modifications and substitutions can be made to these embodiments without departing from the principle and spirit of the present invention. and variants.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and scheme of the present invention shall be included in the scope of the present invention. within the scope of protection.

Claims (7)

1. A power fingerprint precision management method for distribution network line loss analysis, characterized in that, comprising the following steps: S1. Data collection; the collected data includes operation information data, and the required data is collected according to the data application process; S2. Data processing: when analyzing the correlation between the line loss rate and the power consumption of the downlink users, for the sudden change data processing; for the data signal of the line loss and power consumption, the modal decomposition is performed, and the user data is taken out in advance; for missing The electricity consumption data is filled with Lagrangian interpolation method; S3. "Power fingerprint" extraction and analysis: 3.1) Bring fingerprint recognition into the generation of power curve data, convert the continuous data of each level object including the station area into a map texture, and fit the map texture to form a real distribution network line loss change state Atlas, that is, the "power fingerprint" of line loss, constructs line loss and electric quantity fingerprints respectively; 3.2) Sampling and selecting the "power fingerprint" of the station area with abnormal line loss, analyzing the characteristics of the extracted fingerprint from the perspective of central tendency, degree of dispersion, and descriptive statistics of distribution characteristics, and intuitively presenting the influence relationship between power and line loss; S4. Modeling analysis: 4.1) Fingerprint correlation analysis model: By carrying out the correlation analysis between the fluctuation of the line loss fingerprint in the station area and the fluctuation of the fingerprint of the connected user's electricity, locate the suspected user who affects the abnormal fluctuation of the line loss in the station area; 4.2) The analysis model of fingerprint fluctuation variation: through the analysis of the inflection point of the change of the fingerprint fluctuation of the line loss in the station area and the fingerprint fluctuation of the user's electricity under the connection, locate the suspected user who affects the abnormal fluctuation of the line loss in the station area; 4.3) Empirical mode decomposition model of fingerprint signal: by analyzing the time-frequency signal analysis of the fingerprint fluctuation of the line loss in the station area and the fingerprint fluctuation of the electricity quantity of the downlink user, locate the suspected user who affects the abnormal fluctuation of the line loss in the station area; S5. Based on the grading of the suspicion degree of abnormal users as a foothold, a method strategy model is established. 2. A method for precise management of power fingerprints for distribution network line loss analysis according to claim 1, characterized in that, in step S1, the operation information data includes 10 kV distribution network lines, station areas, high and low voltage users File information and daily frozen power, line loss; In step S2, the data of users whose daily average power consumption is 0 and daily average power consumption is less than 1kW·h are taken out in advance; In step S3, the continuous data of each hierarchical object includes data of lines, stations, and users. 3. According to claim 1 or 2, a power fingerprint precision management method for distribution network line loss analysis, characterized in that, the modeling process of step 4.1): ① data preprocessing, including deduplication of user electricity consumption data , Eliminate users with zero daily electricity consumption, and delete useless fields; ② Extract fingerprints, including the fingerprint curves of the line loss rate in the station area and the electricity consumption of connected users; ③ Calculate the correlation coefficient, using Pearson and other correlation coefficients The method quantitatively analyzes the degree of correlation between each user's power fingerprint and the line loss rate fingerprint of the station area; ④ locates strongly correlated users, and uses the correlation coefficient results of each user's power consumption and the line loss of the station area to screen and locate the power fingerprint and line loss fingerprint Anomaly suspect users who are strongly related to the direction change. 4. According to claim 1 or 2, a power fingerprint precision management method for distribution network line loss analysis, characterized in that, the modeling process of step 4.2): ① data preprocessing, including deduplication of user electricity consumption data , Eliminate invalid data, data wide table splicing; ②Lock the change inflection point date, select the most obvious mutation point of the line loss curve change in the station area; ③Calculate the k value of the inflection point, use the defined "K value = user power change amount / station area The K value of the change rate of power loss in the station area caused by the change rate of the power consumption of each user is calculated according to the model rule of "consumption power change"; Locate abnormal suspected users with large relative deviation of line loss anomalies in the station area. 5. According to claim 1 or 2, a power fingerprint precision management method for distribution network line loss analysis, characterized in that, the modeling process of step 4.3): ① data preprocessing, including deduplication of user electricity consumption data , Lagrangian interpolation for missing data, calculation of daily average power consumption, and wide table splicing; ② Calculate the correlation coefficient r between user power consumption and the line loss rate of the station area, and based on the daily average power consumption and correlation coefficient r Sorting the results; ③Select the top 5% users as the primary screening users, and use the EMD algorithm to decompose the signal mode of the user's power consumption and the power consumption of the station area, extract the high-frequency components (IMF) respectively, and complete the signal spectrum fitting ; ④ mark abnormal users, and use the empirical mode decomposition results of fingerprint signals to screen and locate suspected users with abnormal line loss. 6. A method for accurately managing electric power fingerprints for distribution network line loss analysis according to claim 2, wherein said step S5 comprises: Method Strategy model construction: The three major distribution network line loss "power fingerprint" abnormal user identification algorithm models are "fingerprint correlation analysis model", "fingerprint fluctuation variation analysis model", and "fingerprint signal empirical mode decomposition model". Locate the scope of suspected users that have a strong correlation with line loss abnormalities, and divide the suspicion degree of abnormal users into gradient labels; according to the division results of users under each analysis model, combined with the user's own operating data during normal and abnormal online loss periods, Carry out power trend analysis, abnormal event analysis, and equipment operation status analysis to assist in the verification of specific abnormal power consumption behaviors and occurrence periods; Application process: ① Import the file information of 10 kV distribution network lines, station areas, high and low voltage users, and data related to power consumption and line loss into the three major line loss abnormal user identification algorithm models; ② Output model results: based on the three models Respectively output the scope of suspected users that have a strong correlation with line loss anomalies; ③ match and fuse the three major model result labels to complete the comprehensive rating of the degree of suspicion of abnormal users, and differentiate the suspected users; ④ for different categories of suspected users, Execute by category until the loss reduction target is achieved. 7. The precise power fingerprint management method for distribution network line loss analysis according to claim 1, characterized in that the modeling analysis in step S4 is based on the business implementation logic and early stage of abnormal user location based on line loss and power correlation analysis Data processing, frame the scope of machine learning and deep learning big data related algorithms, conduct comparative analysis of the effect, efficiency and stability of the algorithms through Python training, and finally complete the "fingerprint correlation" based on the power and line loss fingerprints The training and construction of the three major line loss abnormal user identification algorithm models: analysis model", "fingerprint fluctuation variation analysis model", and "fingerprint signal empirical mode decomposition model".

Images