A method and system for checking the relationship between household transformers in substations based on ensemble learning LSCP algorithm

Abstract

A method and system for checking the relationship between household transformers in a substation area based on an integrated learning LSCP algorithm, comprising: step 1, collecting substation operation data within N days of multiple substations; step 2, calculating the abnormal scores of users in the substation area based on multiple base anomaly detectors and substation operation data to build a household transformer relationship checking model; step 3, the household transformer relationship checking model screens users with normal household transformer relationships in the substation area and calculates the substation line loss rate according to their operation data to determine the optimal input parameters of the model; step 4, inputting the optimal input parameters into the household transformer relationship checking model to obtain users with incorrect household transformer relationships in the substation area. The present invention can effectively identify incorrect users in household transformer relationship files, greatly improves both the recall rate and the precision rate, saves the cost of manual investigation, realizes a comprehensive judgment method based on integrated learning, and compared with a single anomaly detection model, the integrated method combines multiple anomaly detection models, and the recognition result is more stable and reliable.

Description

Technical Field

[0001] The invention belongs to the field of electric power distribution, relates to household transformer relationship identification, and specifically is a method for checking household transformer relationship in a transformer substation based on an integrated learning LSCP algorithm. Background Art

[0002] The transformation and upgrading of distribution lines has caused frequent changes in the household-to-substation relationship in the substation. If the household-to-substation information of the distribution system is not updated in a timely manner or errors are entered during manual updating, errors will occur in the household-to-substation relationship in the system, affecting the operation and maintenance of the power grid, as well as causing inconsistencies in the metering data between the substation master meter and the user meter in the marketing system, abnormal line losses in the substation, and affecting the lean management of the substation.

[0003] In order to solve the problem of abnormal line loss in the substation caused by incorrect household-to-substation relationships, it is necessary to conduct household-to-substation relationship correction in the substation, identify users with incorrect household-to-substation relationships and correct them. Traditional engineering methods include instantaneous power outage method and pulse current method, which require manual on-site verification, have low accuracy and efficiency, and do not meet the intelligent development goals of the distribution network. Therefore, it is urgent to propose a new household-to-substation relationship correction method that meets the requirements of accuracy and intelligence and helps the digital transformation of the distribution network.

[0004] Prior art document 1 (CN114021430 A) proposes a method for identifying household-to-substation relationships based on the fusion of electricity and voltage information. According to the conservation of total meter electricity in the substation and user electricity, a parameterized model of electricity consumption in the substation is constructed; the global optimal solution of the parameterized model of electricity consumption in the substation is established by using random gradient iteration and learning rate self-adjustment methods, and the affiliation between users and substations is preliminarily identified; a multiple judgment method based on a sliding time window is adopted to divide the historical data of the substation into multiple time windows with different start dates, and multiple solutions are performed, and the results of multiple solutions are taken as a union to obtain the household-to-substation affiliation, that is, the preliminary household-to-substation identification result; the voltage data of users with normal household-to-substation relationships constitute training samples; a class of support vector machine methods is used to learn the voltage characteristics of users with normal household-to-substation relationships in the substation, and a household-to-substation relationship identification model is constructed. Household-to-substation identification is performed on all users in the current files of the substation, and users with incorrect household-to-substation relationship files are obtained, so as to realize the correct identification of household-to-substation relationships in the substation. The shortcoming of the prior art document 1 is that when processing voltage data, only a recognition method based on a type of support vector machine is used, resulting in a low accuracy rate in identifying household change relationships. Summary of the invention

[0005] In order to solve the deficiencies in the prior art, the purpose of the present invention is to provide a method and system for checking the relationship between household transformers in the substation area based on the ensemble learning LSCP (Local Selective Combination in Parallel Outlier Ensembles) algorithm, using the local selective combination parallel ensemble anomaly detection algorithm LSCP, and taking 4 classic anomaly detection algorithms as the base anomaly detectors to accurately detect the differences in the voltage curves of users in the substation area, and realize the correction of the relationship between household transformers in the substation area. At the same time, the key input parameters in the LSCP algorithm are determined by calculating the line loss rate, thereby improving the practicality of the algorithm.

[0006] The present invention adopts the following technical solution.

[0007] A method for checking the relationship between household transformers in a transformer area based on an integrated learning LSCP algorithm comprises the following steps:

[0008] Step 1, collecting operation data of multiple substations within N days;

[0009] Step 2, calculating the anomaly scores of users in the area based on multiple base anomaly detectors and area operation data to build a household change relationship correction model;

[0010] Step 3, the household transformer relationship correction model selects users with normal household transformer relationship in the transformer area and calculates the transformer area line loss rate according to the operation data of users with normal household transformer relationship to determine the best input parameters of the model;

[0011] Step 4, inputting optimal input parameters into the household change relationship correction model to obtain the household change relationship error users in the station area.

[0012] The substation operation data includes: daily power consumption and voltage readings of all users in the substation, and daily power consumption and voltage readings of power supply transformers in the substation.

[0013] The household change relationship correction model includes a plurality of base anomaly detectors, wherein the base anomaly detectors are used to determine the household change relationship of users in the station area.

[0014] The construction process of the household change relationship investigation model includes:

[0015] S1, input the voltage readings of all users in the substation area, and use multiple base anomaly detectors to calculate the anomaly score of each user's voltage reading;

[0016] S2. The user with the highest anomaly score is determined as a user with incorrect household change relationship, and the remaining users are determined as users with normal household change relationship.

[0017] S1 also includes training base anomaly detectors: the voltage readings of all users in the substation area are input into R base anomaly detectors to calculate R local anomaly scores of the voltage readings of the users, and a local anomaly score matrix is ​​established. The correlation between the data in the matrix and the maximum value of the local anomaly score is calculated respectively, and the x base anomaly detectors with the largest correlation are screened out.

[0018] The construction process also includes: using the selected x base anomaly detectors to calculate the number of users x in the ith area i The average of the x anomaly scores is taken as the anomaly score of the user.

[0019] The anomaly scores of all users in the station area constitute an anomaly score set S. The first Y data with high scores in the anomaly score set S are regarded as the number of users with incorrect household-to-household relationship in the station area, where: Y = n × c, where n is the total number of users in the station area, and c is the proportion of abnormal data. The number Y of users with incorrect household-to-household relationship in the station area is controlled by the abnormal data proportion c.

[0020] The selected base anomaly detectors include Isolation Forest, OC-SVM, COPOD, and outlier factor detection algorithm LOF;

[0021] The Pearson correlation coefficient is used to calculate the correlation between the data in the matrix and the maximum value of the local anomaly score.

[0022] Step 3 also includes:

[0023] Step 301, traverse and take values ​​of the abnormal data ratio every 1 / n in the interval (0, 0.1] to obtain multiple abnormal data ratio values, where n is the total number of users in the station area;

[0024] Step 302, inputting the abnormal data ratio value into the household change relationship correction model to obtain the users with normal household change relationship in the station area and the users with wrong household change relationship in the station area;

[0025] Step 303, the sum of daily power consumption of users with normal household transformer relationships in the transformer area is selected to calculate the transformer area line loss rate, and a minimum line loss rate threshold is set;

[0026] Step 304, determine whether the line loss rate of the station area is lower than the minimum line loss threshold. If not, it is considered that there are missed users with wrong household change relationships and step 305 is executed. If yes, the next abnormal data ratio value is obtained and step 302 is executed repeatedly;

[0027] Step 305, determine whether the line loss rate of the substation area is the minimum line loss rate of the substation area. If not, take the next abnormal data ratio value and loop through step 302. If yes, the abnormal data ratio value corresponding to the line loss rate of the substation area is the optimal input parameter.

[0028] The calculation method of the line loss rate of the substation area is as follows:

[0029]

[0030] The value range of the minimum line loss threshold is [-2%, 5%].

[0031] After obtaining the users with incorrect household change relationships in the area, the users with incorrect household change relationships are manually verified and the system household change relationship files are updated.

[0032] A method for checking the relationship between household changes in a substation area based on an integrated learning LSCP algorithm A system for checking the relationship between household changes in a substation area based on an integrated learning LSCP algorithm comprises:

[0033] Data collection module, used to collect substation operation data: daily power consumption and voltage readings of all users in the substation, daily power consumption and voltage readings of power supply transformers in the substation;

[0034] Model building module: used to build a household-to-household relationship correction model based on multiple anomaly detection algorithms and substation operation data;

[0035] Data calculation module: used to calculate the local anomaly score of the voltage reading of users in the substation area, the anomaly score, the substation area line loss rate and the correlation between the data in the matrix and the maximum value of the local anomaly score;

[0036] Algorithm screening module: used to screen out base anomaly detectors according to the calculation results of the data calculation module;

[0037] Output module: used to obtain users with household transformer relationship errors in the substation area based on the output results of the model.

[0038] The beneficial effect of the present invention is that, compared with the prior art, the present invention combines multiple anomaly detection models to achieve detection of data from different angles, strengthens the learning of data features to obtain better performance and reliability, strengthens the stability and accuracy of the correction results, and applies the anomaly detection method with an integrated framework to the identification and correction of household-to-electricity transformer relationships. The present invention can use the actual area operation data of existing power consumption information systems, marketing systems, etc. as a basis, and use a parallel integrated anomaly detection algorithm based on local selection combination to perform household-to-electricity transformer relationship correction, calculate the users with wrong household-to-electricity transformer relationships in a certain area, and clarify the direction of on-site verification at the system level.

[0039] The beneficial effects of the present invention also include:

[0040] (1) The present invention optimizes and determines the key input parameters of the algorithm from the perspective of line loss rate, thereby avoiding the influence of artificially set parameters on the household change relationship correction results, and also increasing the reliability and accuracy of the results, further improving the practicality of the household change relationship correction algorithm proposed in this paper.

[0041] (2) Based on the method of the present invention, the number of abnormal data, that is, the number of users with incorrect household-to-household relationship, is controlled by controlling the proportion of abnormal data for a large number of substations. This allows the user list with incorrect household-to-household relationship to be accurately determined. Based on the household-to-household inspection by actual engineering maintenance personnel, the specific incorrect users can be identified, thereby effectively reducing the workload of on-site inspections, improving the efficiency of on-site inspections, and saving inspection costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A schematic flow chart of a method for checking the relationship between household transformers in a transformer area based on an integrated learning LSCP algorithm in the present invention;

[0043] Figure 2 It is the overall framework of the integrated anomaly detection algorithm LSCP in the present invention;

[0044] Figure 3 The present invention is a flowchart for determining the optimal input parameters of the household change relationship correction model. DETAILED DESCRIPTION

[0045] The present application is further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present application.

[0046] Figure 1 FIG. 1 is a flow chart of a method for checking the relationship between household transformers in a transformer area based on an integrated learning LSCP algorithm in the present invention. Figure 1 The following steps are shown:

[0047] Step 1, collecting the substation operation data of multiple substations within N days; the substation operation data includes: daily power consumption and voltage readings of all users in the substation, and daily power consumption and voltage readings of power supply transformers in the substation.

[0048] Step 2, calculating the anomaly scores of users in the area based on multiple base anomaly detectors and area operation data to build a household change relationship correction model;

[0049] Household Change Relationship Correction Model The household change relationship correction model includes multiple base anomaly detectors, where the base anomaly detectors are used to determine the household change relationship of users in the station area.

[0050] The construction process includes:

[0051] S1, input the voltage readings of all users in the substation area, and use multiple base anomaly detectors to calculate the anomaly score of each user's voltage reading;

[0052] S2. The user with the highest anomaly score is determined as a user with incorrect household change relationship, and the remaining users are determined as users with normal household change relationship.

[0053] Specifically, the base anomaly detector is trained: the voltage readings of all users in the substation area are respectively input into R base anomaly detectors to calculate the R local anomaly scores of the voltage readings of the users, and a local anomaly score matrix is ​​established. The correlation between the data in the matrix and the maximum value of the local anomaly score is calculated respectively, and the x base anomaly detectors with the largest correlation are screened out.

[0054] Using the selected x base anomaly detectors, the number of users x in the ith area is calculated. i The x anomaly scores of the user are calculated, and the average of the x anomaly scores is taken as the anomaly score of the user. The selected base anomaly detectors include isolation forest, OC-SVM, COPOD and outlier factor detection algorithm LOF; the Pearson correlation coefficient is used to calculate the correlation between the data in the matrix and the maximum value of the local anomaly score.

[0055] The anomaly scores of all users in the station area constitute an anomaly score set S. The first Y data with high scores in the anomaly score set S are regarded as the number of users with incorrect household-to-household relationship in the station area, where: Y = n × c, where n is the total number of users in the station area, and c is the proportion of abnormal data. The number Y of users with incorrect household-to-household relationship in the station area is controlled by the abnormal data proportion c.

[0056] Step 3: The household transformer relationship correction model selects users with normal household transformer relationship in the transformer area and calculates the transformer area line loss rate based on the operation data of users with normal household transformer relationship to determine the optimal input parameters of the model, specifically including:

[0057] Step 301, traverse and take values ​​of the abnormal data ratio every 1 / n in the interval (0, 0.1] to obtain multiple abnormal data ratio values, where n is the total number of users in the station area;

[0058] Step 302, inputting the abnormal data ratio value into the household change relationship correction model to obtain the users with normal household change relationship in the station area and the users with wrong household change relationship in the station area;

[0059] Step 303, the sum of daily power consumption of users with normal household transformer relationships in the transformer area is selected to calculate the transformer area line loss rate, and a minimum line loss rate threshold is set;

[0060] Step 304, determine whether the line loss rate of the station area is lower than the minimum line loss threshold. If not, it is considered that there are missed users with wrong household change relationships and step 305 is executed. If yes, the next abnormal data ratio value is obtained and step 202 is executed repeatedly;

[0061] Step 305, determine whether the area line loss rate is the minimum area line loss rate, if not, then take the next abnormal data ratio value and loop through step 202, if yes, then the abnormal data ratio value corresponding to the area line loss rate is the optimal input parameter. The area line loss rate is calculated as follows:

[0062]

[0063] The value range of the minimum line loss threshold is [-2%, 5%].

[0064] Step 4, inputting optimal input parameters into the household change relationship correction model to obtain the household change relationship error users in the station area.

[0065] After obtaining the users with incorrect household change relationships in the area, the users with incorrect household change relationships are manually verified and the system household change relationship files are updated.

[0066] Figure 2 It is the overall framework of the integrated anomaly detection algorithm LSCP. LSCP is a completely unsupervised anomaly detection algorithm that integrates multiple anomaly detection algorithms in parallel. The characteristics of the LSCP algorithm are: ① Multiple base anomaly detection models are calculated independently and judged in parallel; ② According to the performance of the base anomaly detection model, the combined base model is dynamically selected; ③ Subsequences of different dimensions of randomly sampled data are used to construct feature sample subspaces, emphasizing local anomalies of the data; ④ Supports the combination of heterogeneous base anomaly detection models, and also supports the combination of homogeneous base anomaly detection models. Combining heterogeneous models can make the model diverse, and combining homogeneous models can also achieve diversity by setting different hyperparameters. The specific process of LSCP algorithm for abnormal data detection includes the following steps.

[0067] Input the dataset to be tested X∈R n×d The data set to be detected is the voltage readings of all users in the substation area, and the base anomaly detector is C = {C 1 ,C 2 ,…C R}, train R base anomaly detectors and obtain the anomaly score matrix O(X)∈R n×R .

[0068] O(X)=[C 1 (X),C 2 (X),…,C R (X)] (1)

[0069] For each data x i ∈Xx i ∈R 1×d , perform random sampling, and construct The characteristic subsequence x of dimension i ′, and obtain a set of feature subspaces of the data set X.

[0070] In the feature subspace, calculate x i With x i ′, and the Euclidean distance with x is obtained by using the K nearest neighbor method (KNN). i′The k nearest neighbor samples. Perform t groups of random sampling to obtain t groups of feature subspaces. i Among the neighboring samples in the feature subspace of , samples that appear more than t / 2 times are taken as x i The local space ψ i .

[0071]

[0072] Each data x i The cardinality b of the local space is different, that is, the number of samples contained therein is different.

[0073] Get data x i After the local space is obtained, the base anomaly detector is used to obtain x i The local anomaly score O(ψ i )∈R b ×R Assume x i The local space of contains 10 data samples. The anomaly score of the local space is calculated by R base anomaly detectors to obtain a 10×R dimensional local anomaly score matrix O(ψ i ).

[0074] The maximum value of the scores of the R base anomaly detectors is used as the pseudo target target of the data sample, then the data x i The pseudo target is target(x i ), which is calculated as follows:

[0075] Target(x i )=f(O(ψ i )) max (3)

[0076] Calculate the local anomaly score matrix O(ψ i ) and data x i The pseudo target target(ψ i ), select the local anomaly score matrix O(ψ i ) are the x local anomaly scores that are most similar to the pseudo target, and the base anomaly detector that calculates these x local anomaly scores is used as the base anomaly detector of the household change relationship correction model.

[0077] Using the selected x base anomaly detectors, calculate the data x i The anomaly score is obtained by i )∈R x×1 , and take the average of these x anomaly scores as x iThe anomaly score s is

[0078] s=f avg (O(x i )) (4)

[0079] The above method is used to calculate the anomaly scores of all data in the dataset X to be detected, and an anomaly score set S (S∈R n×1 ), by inputting the abnormal data ratio c, the number of abnormal data in the data set is controlled to be Y, where Y is the number of users with incorrect household-to-household relationship in the station area, and the first Y data with high scores in the abnormal score S are regarded as abnormal data.

[0080] Y=n×c (5)

[0081] Where n is the number of samples in the data set, that is, the total number of users in the area, and c is the proportion of abnormal data.

[0082] In order to improve the reliability and stability of the integrated model, this paper combines the LSCP algorithm with a heterogeneous base anomaly detection model and uses four classic anomaly detection algorithms as base anomaly detectors, namely Isolation Forest, OC-SVM (One-Class Support Vector Machine), COPOD (Copula-Based Outlier Detection), and LOF (Local Outlier Factor). This enables detection of data anomalies from different angles and strengthens the learning of data difference features.

[0083] Isolation forest is an algorithm for detecting anomalies based on the idea of ​​partitioning. It partitions the data set. The fewer times a sample is partitioned, the easier it is to be isolated and the higher the degree of abnormality of the sample. OC-SVM is an algorithm for detecting abnormal data proposed for unbalanced samples. It maps the original data to a high-dimensional space through a kernel function, making a clear difference between normal data and abnormal data, and constructing a hyperplane to separate the two. COPOD is an algorithm for detecting anomalies based on statistics. It calculates the tail probability of each point in the data for multidimensional data and the diversity of data distribution, and calculates the skewness of the distribution to correct the tail probability of the data, so as to estimate the degree of abnormality of the data. LOF is based on the idea of ​​density. It detects anomalies by comparing the density of each point with its neighboring points. In addition, the LOF algorithm calculates the density through the Kth neighborhood of the point, which will not cause abnormal misjudgment due to different data density dispersion.

[0084] In order to avoid the impact of users' random electricity consumption behavior on the household transformer relationship correction and to amplify the difference between the voltage curves of users with normal household transformer relationships and those with errors, this application selects two days of voltage data of users in the transformer area as input.

[0085] There are differences in voltage data between users in different areas, showing different trends and characteristics, and the proportion of users with incorrect household-to-substation relationships in the area is very small. The household-to-substation relationship correction problem in the area can be defined as abnormal data detection of unbalanced samples. In order to ensure the accuracy and reliability of the results, this paper adopts the LSCP anomaly detection algorithm with an integrated framework for household-to-substation relationship correction.

[0086] The principle of the method for checking the relationship between household transformers in the transformer area based on the LSCP algorithm is as follows: input the voltage data of the users in the transformer area, use the LSCP algorithm to accurately detect the abnormal trend of the voltage data, judge the abnormal degree of the voltage data, and obtain the abnormal score of the voltage data of each user. The individual user with the highest data abnormality score is determined as a user with an incorrect household transformer relationship, and the output is 1. The remaining users are determined as users with normal household transformer relationships, and the output is 0. When using the LSCP algorithm to check the relationship between household transformers, the adjustable parameters are as shown in Table 1:

[0087] Table 1 LSCP algorithm parameter description

[0088]

[0089] Adjusting the proportion of abnormal data can change the number of users with incorrect household change relationships in the output; adjusting the number of KNN neighbor samples, the number of random sampling times, the lowest dimension of random sampling, the highest dimension of random sampling, and the threshold of the number of feature space samples can change the local space of the data.

[0090] The abnormal data ratio controls the number of abnormal data outputs, that is, the number of users with incorrect household-to-household relationship changes, and is a key input parameter of the household-to-household relationship correction algorithm. When correcting the household-to-household relationship change in a substation, the proportion of users with incorrect household-to-household relationship change in the substation is unknown. For a large number of substations, if the abnormal data ratio is set to a fixed value, it will affect the accuracy and efficiency of the calculation results. This paper attempts to optimize and determine the key input parameters of the algorithm from the perspective of line loss rate to avoid the impact of artificially set parameters on the household-to-household relationship correction results, while also increasing the reliability and accuracy of the results, and further improving the practicality of the household-to-household relationship correction algorithm proposed in this paper.

[0091] Theoretically, the relationship between the power consumption of the substation and the user is:

[0092]

[0093] Where P is the total power consumption of the area, p(i) is the power consumption of user i, and n is the number of users in the area.

[0094] In fact, due to the online loss of the area memory:

[0095]

[0096] According to the regulations of the power supply company, the line loss rate of the substation normally fluctuates within a certain range. If there are users with incorrect household-to-substation relationships in the substation system file, the total power consumption of the substation in the marketing metering system will be less than the sum of the power consumption of the users, as shown in formula (8).

[0097]

[0098] Therefore, this embodiment uses the line loss rate of the transformer area to determine the optimal input parameters of the household transformer relationship correction model, and the determination process is as shown in the attached figure. Figure 3 As shown:

[0099] Step 301, traverse and take values ​​of the abnormal data ratio every 1 / n in the interval (0, 0.1] to obtain multiple abnormal data ratio values, where n is the total number of users in the station area;

[0100] Step 302, inputting the abnormal data ratio value into the household change relationship correction model to obtain the users with normal household change relationship in the station area and the users with wrong household change relationship in the station area;

[0101] Step 303, the sum of daily power consumption of users with normal household transformer relationships in the transformer area is selected to calculate the transformer area line loss rate, and a minimum line loss rate threshold is set;

[0102] Step 304, determine whether the line loss rate of the station area is lower than the minimum line loss threshold. If not, it is considered that there are missed users with wrong household change relationships and step 305 is executed. If yes, the next abnormal data ratio value is obtained and step 202 is executed repeatedly;

[0103] Step 305, determine whether the area line loss rate is the minimum area line loss rate, if not, then take the next abnormal data ratio value and loop through step 202, if yes, then the abnormal data ratio value corresponding to the area line loss rate is the optimal input parameter. The area line loss rate is calculated as follows:

[0104]

[0105] The value range of the minimum line loss threshold is [-2%, 5%].

[0106] Input different abnormal data ratio values, use LSCP algorithm to calculate, get the number of users with wrong household change relationship, and determine the users with normal household change relationship in the area. In this embodiment, the abnormal data ratio is traversed and valued every 0.01 in the interval (0,0.1], because the number of users in the area in the simulation scenario of this specific embodiment is about 100, and when the interval is 0.01, the number of users with wrong household change relationship is outputted by approximately 1 each time. If 0.005 is taken, the amount of calculation will increase; if 0.02 is taken, the parameter changes too much, and it is difficult to find the optimal parameter.

[0107] Considering the existence of line loss during normal operation of the substation, a minimum line loss rate threshold is set. The minimum line loss rate threshold can be set based on the system's historical line loss records or actual substation characteristics.

[0108] If the line loss rate calculated based on different input parameters is lower than the minimum line loss threshold line loss, it is considered that there are missed users with incorrect household change relationships. When the input abnormal data ratio value is increased, the number of users with incorrect household change relationships increases, and users with normal household change relationships may be misjudged as incorrect users, which will further reduce the sum of electricity consumption of users in the substation area and increase the line loss rate in the substation area. Therefore, based on the minimum line loss threshold, the input value corresponding to the minimum line loss rate is selected as the optimal parameter.

[0109] The specific implementation steps of the method for checking the relationship between household transformers in the transformer area based on the integrated learning LSCP algorithm of this embodiment can be summarized as follows:

[0110] Collect the operation data of the substation area from the power consumption system, including voltage data and daily total power consumption data. The abnormal data ratio is traversed and taken every 0.01 in the interval (0, 0.1], and the LSCP algorithm is used to check the household-to-substation relationship to obtain users with normal household-to-substation relationship, calculate the line loss rate of the substation area, determine the value that meets the conditions, and obtain the optimal input parameters. Input the optimal parameters, use the LSCP algorithm to check the household-to-substation relationship, and obtain users with incorrect household-to-substation relationship in the substation area. The staff verifies the users with incorrect household-to-substation relationship on site and updates the system household-to-substation relationship files in a timely manner.

[0111] This embodiment also provides a system for checking the relationship between household changes in a substation area based on an LSCP algorithm based on the above-mentioned method for checking the relationship between household changes in a substation area, which specifically includes:

[0112] Data collection module, used to collect substation operation data: daily power consumption and voltage readings of all users in the substation, daily power consumption and voltage readings of power supply transformers in the substation;

[0113] Model building module, used to build a household-to-substation relationship correction model based on multiple anomaly detection algorithms and substation operation data;

[0114] A data calculation module is used to calculate the local anomaly score of the voltage readings of users in the substation area, the anomaly score, the substation area line loss rate and the correlation between the maximum value of the data in the matrix and the local anomaly score;

[0115] An algorithm screening module, used for screening out-of-base anomaly detectors according to calculation results of the data calculation module;

[0116] The output module is used to obtain the users with household transformer relationship errors in the substation area according to the output results of the model.

[0117] In order to verify the universality and reliability of the proposed method, in a simulated scenario, two days of voltage data were used to check the relationship between household transformers in six substations. The test results are shown in Table 2.

[0118] Table 2 Test results under simulation scenario

[0119]

[0120] In Table 2, the recall rate of area 5 is 80% and the precision rate is 98.71%, and one user with incorrect household-to-household relationship is missed. The recall rate and precision rate of the other five areas are both 100%. However, for area 5, if more than 2 days of voltage data are used for calculation, the recall rate and precision rate can still reach 100%. In practical applications, for higher accuracy and reliability, when the voltage data of the area is sufficient, more than 2 days of voltage data can be used for calculation. Therefore, the present invention has high accuracy and applicability in the identification and correction of household-to-household relationship in the area, and can be well adapted to the practical application of the correction of household-to-household relationship in a large number of areas.

[0121] The beneficial effect of the present invention is that, compared with the prior art, the present invention combines multiple anomaly detection models to achieve detection of data from different angles, strengthens the learning of data features to obtain better performance and reliability, strengthens the stability and accuracy of the correction results, and applies the anomaly detection method with an integrated framework to the identification and correction of household-to-electricity transformer relationships. The present invention can use the actual area operation data of existing power consumption information systems, marketing systems, etc. as a basis, and use a parallel integrated anomaly detection algorithm based on local selection combination to perform household-to-electricity transformer relationship correction, calculate the users with wrong household-to-electricity transformer relationships in a certain area, and clarify the direction of on-site verification at the system level.

[0122] The beneficial effects of the present invention also include:

[0123] (1) The present invention optimizes and determines the key input parameters of the algorithm from the perspective of line loss rate, thereby avoiding the influence of artificially set parameters on the household change relationship correction results, and also increasing the reliability and accuracy of the results, further improving the practicality of the household change relationship correction algorithm proposed in this paper.

[0124] (2) Based on the method of the present invention, the number of abnormal data, that is, the number of users with household-to-household relationship errors, can be accurately determined by controlling the ratio of abnormal data for a large number of substations. Then, based on the household-to-household investigation by actual engineering maintenance personnel, the specific erroneous users can be verified, effectively reducing the workload of on-site verification, improving the efficiency of on-site verification, and saving verification costs. The applicant of the present invention has made a detailed explanation and description of the implementation examples of the present invention in conjunction with the drawings of the specification, but those skilled in the art should understand that the above implementation examples are only preferred implementation schemes of the present invention, and the detailed description is only to help readers better understand the spirit of the present invention, but not to limit the scope of protection of the present invention. On the contrary, any improvement or modification based on the spirit of the present invention should fall within the scope of protection of the present invention.

Claims

1. A method for checking the relationship between household transformers in a transformer area based on an integrated learning LSCP algorithm, characterized in that: The following steps are involved: Step 1, collecting operation data of multiple substations within N days; Step 2, calculating the anomaly scores of users in the area based on multiple base anomaly detectors and area operation data to build a household change relationship correction model; Step 3: The household transformer relationship correction model selects users with normal household transformer relationship in the transformer area and calculates the transformer area line loss rate based on the operation data of users with normal household transformer relationship to determine the optimal input parameters of the model, including: Step 301, traverse and take values ​​of the abnormal data ratio every 1 / n in the interval (0, 0.1] to obtain multiple abnormal data ratio values, where n is the total number of users in the station area; Step 302, inputting the abnormal data ratio value into the household change relationship correction model to obtain the users with normal household change relationship in the station area and the users with wrong household change relationship in the station area; Step 303, the sum of daily power consumption of users with normal household transformer relationships in the transformer area is selected to calculate the transformer area line loss rate, and a minimum line loss rate threshold is set; Step 304, determine whether the line loss rate of the station area is lower than the minimum line loss rate threshold. If not, it is considered that there are missed users with wrong household change relationships and step 305 is executed. If yes, the next abnormal data ratio value is obtained and step 302 is executed repeatedly; Step 305, determine whether the area line loss rate is the minimum area line loss rate, if not, then take the next abnormal data ratio value and loop through step 302, if yes, then the abnormal data ratio value corresponding to the area line loss rate is the best input parameter, the area line loss rate is calculated as follows: Step 4, inputting optimal input parameters into the household change relationship correction model to obtain the household change relationship error users in the station area.

2. A method for checking the relationship between household transformers in a transformer area based on an integrated learning LSCP algorithm according to claim 1, characterized in that: The substation operation data includes: daily power consumption and voltage readings of all users in the substation, and daily power consumption and voltage readings of power supply transformers in the substation.

3. According to the method for checking the relationship between household transformers in a transformer area based on an integrated learning LSCP algorithm as described in claim 1, it is characterized in that: The household change relationship correction model includes a plurality of base anomaly detectors, wherein the base anomaly detectors are used to determine the household change relationship of users in a station area.

4. A method for checking the relationship between household transformers in a transformer area based on an integrated learning LSCP algorithm according to claim 3, characterized in that: The construction process of the household change relationship investigation model includes: S1, input the voltage readings of all users in the substation area, and use multiple base anomaly detectors to calculate the anomaly score of each user's voltage reading; S2. The user with the highest anomaly score is determined as a user with incorrect household change relationship, and the remaining users are determined as users with normal household change relationship.

5. A method for checking the relationship between household transformers in a transformer area based on an integrated learning LSCP algorithm according to claim 4, characterized in that: The S1 also includes a training base anomaly detector: the voltage readings of all users in the substation area are respectively input into R base anomaly detectors to calculate R local anomaly scores of the voltage readings of the users, and a local anomaly score matrix is ​​established, and the correlation between the data in the matrix and the maximum value of the local anomaly score is respectively calculated, and x base anomaly detectors with the largest correlation are screened out.

6. A method for checking the relationship between household transformers in a transformer area based on an integrated learning LSCP algorithm according to claim 5, characterized in that: The construction process also includes: using the selected x base anomaly detectors to calculate the number of users in the i-th station area The average of the x anomaly scores is taken as the anomaly score of the user.

7. A method for checking the relationship between household transformers in a transformer area based on an integrated learning LSCP algorithm according to claim 4, characterized in that: The anomaly scores of all users in the area constitute the anomaly score set , the first Y data with high scores in the abnormal score set S are regarded as the data of users with wrong household transformer relationships in the station area, where: , where n is the total number of users in the station area, c is the abnormal data ratio, and the number of users with incorrect household-to-household relationship errors in the station area, Y, is controlled by the abnormal data ratio c.

8. A method for checking the relationship between household transformers in a transformer area based on an integrated learning LSCP algorithm according to claim 5, characterized in that: The x-based anomaly detectors with the greatest correlation screened out include Isolation Forest, OC-SVM, COPOD, and the outlier factor detection algorithm LOF; The Pearson correlation coefficient is used to calculate the correlation between the data in the matrix and the maximum value of the local anomaly score.

9. The method for checking the relationship between household transformers in a transformer area based on an integrated learning LSCP algorithm according to claim 1, characterized in that: The value range of the minimum line loss rate threshold is [-2%, 5%].

10. The method for checking the relationship between household transformers in a transformer area based on an integrated learning LSCP algorithm according to claim 1, characterized in that: After obtaining the users with incorrect household change relationships in the area, the users with incorrect household change relationships are manually verified and the system household change relationship files are updated.

11. A method for checking the relationship between household changes in a substation area based on an integrated learning LSCP algorithm according to any one of claims 1 to 9, a system for checking the relationship between household changes in a substation area based on an integrated learning LSCP algorithm, characterized in that: include: Data collection module, used to collect substation operation data: daily power consumption and voltage readings of all users in the substation, daily power consumption and voltage readings of power supply transformers in the substation; Model building module: used to build a household-to-household relationship correction model based on multiple anomaly detection algorithms and substation operation data; Data calculation module: used to calculate the local anomaly score, anomaly score, line loss rate of the area and the correlation between the maximum value of the local anomaly score and the data in the matrix; wherein, by inputting the voltage readings of all users in the area and using multiple base anomaly detectors to calculate the anomaly score of each user's voltage reading, the voltage readings of all users in the area are respectively input into R base anomaly detectors to calculate R local anomaly scores of the user's voltage readings, and a local anomaly score matrix is ​​established, and the correlation between the data in the matrix and the maximum value of the local anomaly score is respectively calculated, and x base anomaly detectors with the largest correlation are selected; the correlation between the data in the matrix and the maximum value of the local anomaly score is calculated using the Pearson correlation coefficient; the calculation method of the line loss rate of the area is as follows: ； Algorithm screening module: used to screen out base anomaly detectors according to the calculation results of the data calculation module; Output module: used to obtain users with household transformer relationship errors in the substation area based on the output results of the model.

Images