A method for identifying electricity stealing users of 10 kilovolt non-economic operation line

Abstract

This invention relates to the field of electricity theft identification technology in the power industry, specifically a method for identifying electricity theft users on 10 kV non-economically operating lines. The method first collects daily line loss rate and daily electricity consumption data from users, performing preprocessing including missing value repair and outlier removal using box plots. Next, it calculates the Pearson correlation coefficient between the line loss rate and user electricity consumption, constructs an adaptive anomaly judgment threshold range based on the statistical characteristics of the correlation coefficient, verifies the significance of the correlation coefficient through hypothesis testing, and finally combines the threshold range with the test results to perform a joint judgment to identify suspected electricity theft users. This invention identifies the hidden correlation between line loss rate and user electricity consumption from a statistical correlation perspective, overcoming the limitations of traditional methods that rely on sudden changes in line loss rate. It can accurately capture behaviors such as diversion and slow electricity theft, achieving high detection accuracy and strong engineering practicality, providing technical support for electricity theft prevention work.

Description

Technical Field

[0001] This invention relates to the field of electricity theft identification technology in the power industry. Background Technology

[0002] With the deepening application of big data analysis and intelligent metering systems in the power industry, integrated power consumption and synchronous line loss systems have been widely deployed in power supply companies. This system can synchronously collect and statistically analyze the line loss rate at the line level and the power consumption at the user level on a daily basis, providing a data foundation for carrying out line loss analysis and abnormal user identification based on historical operating data.

[0003] Traditional line loss management and anti-theft methods mainly rely on manual experience, identifying days with sudden changes in line loss rate and comparing them with abnormal fluctuations in user electricity consumption. This method is only effective in identifying sudden and drastic power consumption anomalies and has significant limitations in practical applications: First, for high-loss lines with monthly line loss rates that remain in a high range [6%, 10%] for a long time, or lines in the non-economic operating range [3%, 6%), their daily line loss rate curves are usually stable and lack significant abrupt changes, making it difficult to capture anomalies through the sudden change day method; Second, covert power theft behaviors such as diversion theft and slow power theft often do not cause drastic fluctuations in the daily line loss rate, making traditional methods unable to effectively identify such abnormal users.

[0004] From the perspective of distribution network operation mechanism analysis, under engineering approximation conditions, line losses mainly consist of conductor resistance losses and transformer losses. Line current distribution is directly related to downstream user load changes, and theoretically, there is an inherent correlation between line loss rate and downstream user electricity consumption. Under normal metering conditions, changes in the load of a single user have a relatively small impact on the overall line loss rate, and this correlation is weak. However, when a user engages in electricity theft, current diversion, or metering anomalies, a deviation occurs between their actual load and metered load, significantly disrupting the original balance between line load and line loss. This results in an abnormal statistical correlation between the user's metered electricity consumption and the line loss rate. Based on this operation mechanism, there is an urgent need for a method to accurately identify users engaging in abnormal electricity theft on uneconomical operating lines from a statistical correlation perspective. Summary of the Invention

[0005] To overcome the problems of existing electricity theft identification methods relying on sudden changes in line loss rate and failing to effectively identify concealed electricity theft in 10 kV uneconomical operating lines, this invention provides a method for identifying electricity theft users on 10 kV uneconomical operating lines, comprising the following steps:

[0006] Collect operational data for the target 10 kV non-economical operating line, including the line's daily line loss rate and the user's daily electricity consumption;

[0007] The linear correlation coefficient is obtained by measuring the degree of linear correlation between the daily line loss rate and the daily electricity consumption of users.

[0008] An adaptive threshold model is constructed and solved based on the linear correlation coefficient to obtain the adaptive anomaly detection threshold range;

[0009] The statistical significance test was performed based on the linear correlation coefficient, and the test results were obtained.

[0010] A joint decision is made based on the abnormal decision threshold range and the test results. The decision result is the identification result of the electricity theft user on the target 10 kV non-economic operation line.

[0011] Preferably, it further includes:

[0012] Missing values ​​are repaired in the collected operational data. When the number of missing days is less than a preset threshold, the mean interpolation method of adjacent time windows is used for repair. When the missing ratio exceeds the preset threshold or shows long-term missing characteristics, the data is directly removed.

[0013] Outlier removal was performed on the collected daily line loss rate data. Let the sample sequence of the collected daily line loss rate data be:

[0014] ;

[0015] in, For the first Daily line loss rate;

[0016] The daily line loss rate sample series was sorted in ascending order of numerical value, and the first quartile of its three core statistics was calculated. Third and quartiles and interquartile range ;

[0017] Outlier criteria are:

[0018] or ;

[0019] Remove samples that meet the above criteria.

[0020] Preferably, the linear correlation between the daily line loss rate and the daily electricity consumption of users is measured, specifically as follows:

[0021] The Pearson correlation coefficient is used to measure:

[0022] ;

[0023] in, For the first The linear correlation coefficient of each user For the first Daily line loss rate, For the first The user Daily electricity consumption This represents the average of the daily line loss rate data samples. This represents the average daily electricity consumption data sample for users.

[0024] Preferably, the adaptive threshold model is constructed and solved based on the linear correlation coefficient as follows:

[0025] Calculate the mean of the linear correlation coefficient. with standard deviation :

[0026] ;

[0027] ;

[0028] in, For the first The linear correlation coefficient of each user Total number of users

[0029] Define the adaptive anomaly detection threshold range as follows: The calculation formula is:

[0030] ;

[0031] ;

[0032] in, This is the sensitivity coefficient.

[0033] Preferably, the statistical significance test based on the linear correlation coefficient is specifically performed as follows:

[0034] For users Let its null hypothesis be The alternative hypothesis is ,exist , The test statistic is calculated as follows:

[0035] ;

[0036] in, For the sample size, under the null hypothesis Under the conditions established, Obeying the degree of freedom The t-distribution, To construct the significance level of the test, when If so, then reject the null hypothesis. .

[0037] Preferably, the joint decision based on the anomaly judgment threshold range and the test results is as follows: when the linear correlation coefficient is higher than the anomaly judgment threshold range for line users, and the linear correlation coefficient passes the statistical significance test, the user is judged as a suspected abnormal user, that is, a suspected electricity theft user.

[0038] The beneficial effects of this invention are as follows:

[0039] Improving the accuracy and adaptability of anomaly diagnosis: This invention integrates data cleaning, Pearson correlation analysis, and adaptive threshold technology to identify hidden correlations between line loss rate and user electricity consumption from a statistical correlation perspective. It overcomes the limitations of traditional methods that rely on sudden changes in line loss rate. It is especially suitable for non-economical operating lines and high-loss lines with stable line loss curves and no obvious sudden changes. It can accurately capture hidden electricity theft behaviors such as diversion theft and slow electricity theft, thereby improving the accuracy of electricity theft diagnosis.

[0040] Enhance the ability to mine the value of power data: Use box plots to identify non-parametric outliers, effectively eliminating noise points and outliers caused by non-operational factors such as meter reading failures, system maintenance, and data recalculation, while retaining the true operational characteristics of the data; through correlation analysis and statistical hypothesis testing, deeply reveal the intrinsic relationship between user electricity consumption behavior and line loss changes, providing a reliable basis for in-depth power data analysis and decision support, and fully releasing the application potential of the data.

[0041] Achieving threshold adaptation and intelligent decision-making: Based on the statistical distribution of the linear correlation coefficient of users on the line itself, a dynamic adaptive decision threshold is constructed, which overcomes the applicability limitations of fixed thresholds in different line scenarios; through hypothesis testing and dual decision-making mechanisms, the false alarm rate is effectively controlled while improving detection sensitivity, thereby enhancing the robustness and engineering practicality of the method in different operating environments.

[0042] Modular design and good system compatibility: The method described in this invention can be encapsulated as an independent computing module, which can be flexibly embedded into existing integrated power and synchronous line loss systems, line loss management systems or power big data platforms without changing the original system architecture. It has good scalability and compatibility; it supports batch processing and real-time analysis, providing efficient tool support for power supply companies' refined line loss management and anti-theft work.

[0043] Empirical verification demonstrates high engineering feasibility: This invention uses actual 10 kV line operation data as the analysis object, successfully identifying multiple abnormal users who divert electricity for theft. The results were confirmed by on-site verification, proving that this method is not only theoretically rigorous, but also highly feasible and effective in engineering applications, and can be directly applied to the actual work of power supply companies in preventing electricity theft. Attached Figure Description

[0044] Figure 1 This is a schematic diagram of the overall process of an embodiment of the present invention. Detailed Implementation

[0045] Example 1:

[0046] Embodiment 1 of the present invention provides a method for identifying electricity theft users on 10 kV non-economical operating lines, such as... Figure 1 As shown, it includes the following steps:

[0047] S1. Collect operational data of the target 10 kV non-economical operating line.

[0048] Through an integrated power consumption and synchronous line loss system deployed in the power industry, operational data of the target 10 kV non-economical operating lines are collected, including:

[0049] Daily line loss rate: The line loss rate recorded on a daily basis within the statistical period;

[0050] Daily electricity consumption per user: Records of daily electricity consumption of all users downstream of this line within the statistical period.

[0051] S2. Preprocess the running data to obtain preprocessed running data.

[0052] A preprocessing method combining missing value repair and outlier removal is used to remove noisy data caused by non-operational mechanisms. Specifically:

[0053] S2-1, Missing Value Repair:

[0054] For missing data caused by communication interruptions, equipment failures, or other reasons during the collection of daily line loss rate data and daily electricity consumption data by the integrated power consumption and synchronous line loss system, the following processing strategy is adopted.

[0055] When the number of missing days is less than a preset threshold (e.g., consecutive missing days ≤ 3 days), imputation using the mean of adjacent time windows is used. When the missing proportion exceeds a preset threshold (e.g., consecutive missing days > 3 days) or exhibits long-term missing characteristics, the data for that period is considered statistically unrepresentative and is directly removed. This approach ensures sample continuity while avoiding systematic bias introduced by over-imputation.

[0056] S2-2, Outlier Removal:

[0057] To overcome the shortcomings of traditional thresholding and mean-standard deviation methods, which are highly sensitive to data distribution patterns and susceptible to extreme values, this invention employs a nonparametric outlier identification method based on boxplots to clean daily line loss rate data. This method does not rely on the assumption that the data must conform to a specific distribution, thus exhibiting strong robustness to non-normal distributions and outliers that may exist in actual line loss rate data.

[0058] Suppose the daily line loss rate data sample sequence of the lines to be removed is as follows:

[0059] ;

[0060] in, For the first Daily line loss rate, This represents the total number of days within the statistical period.

[0061] The daily line loss rate sample series of the lines to be removed are sorted in ascending order of numerical value, and the first quartile of their three core statistics is calculated. Third and quartiles and interquartile range This describes the distribution structure of the data.

[0062] Outlier criteria are:

[0063] or ;

[0064] Samples that meet the above criteria will be considered statistical outliers generated by non-operational mechanisms and will be removed in subsequent analyses.

[0065] After completing missing value repair and outlier removal, the preprocessed runtime data is obtained:

[0066] The preprocessed daily line loss rate data sample sequence is as follows:

[0067] ;

[0068] The preprocessed user daily electricity consumption data sample sequence is as follows:

[0069] ;

[0070] in, For the first The user Daily electricity consumption.

[0071] S3. Measure the degree of linear correlation between the preprocessed running data.

[0072] The Pearson correlation coefficient is used to measure the degree of linear correlation between the pre-processed daily line loss rate and the daily electricity consumption of users:

[0073] ;

[0074] in, For the first The linear correlation coefficient for each user is used to measure the degree of linear correlation. This represents the average of the daily line loss rate data samples. This represents the average daily electricity consumption data sample for most users. Under normal metering conditions, this is the average daily electricity consumption data for most users. The value is relatively small; if the user exhibits abnormal behavior, its The value will deviate significantly from the overall distribution of the line.

[0075] S4. Construct and solve an adaptive threshold model based on the degree of linear correlation to obtain the adaptive anomaly judgment threshold range.

[0076] To adapt to the differences in load structure, number of users and operating status of different lines, this embodiment constructs an adaptive threshold model based on the statistical characteristics of the line itself, avoiding the use of fixed thresholds.

[0077] Suppose a certain 10 kV line corresponds to the integrated power consumption and synchronous line loss system. The set of linear correlation coefficients for each user is as follows:

[0078] ;

[0079] Calculate its mean with standard deviation :

[0080] ;

[0081] ;

[0082] Define the adaptive anomaly detection threshold range as follows: The calculation formula is:

[0083] ;

[0084] ;

[0085] in, This is a sensitivity coefficient, typically set to 2-3 in engineering practice. When a user's linear correlation coefficient exceeds this range, it is considered a suspected anomaly. It should be noted that the aforementioned adaptive anomaly judgment threshold range is used to characterize the overall statistical boundary of the distribution of line users' linear correlation coefficients. In practical engineering applications, when a user's linear correlation coefficient is significantly higher than the concentrated distribution range of line users, it indicates that the user has clearly deviated from the overall statistical behavior pattern of the line. To describe the degree of this deviation, the concentrated range of the line's linear correlation coefficients can be represented as... ,in This is used to balance the sensitivity of anomaly detection with the overall distribution stability. In this embodiment, the sensitivity coefficient is set to 2.47.

[0086] S5. Perform a statistical significance test based on the degree of linear correlation to obtain the test results.

[0087] To avoid spurious correlations caused by random fluctuations under limited sample size, this invention introduces a t-test based on the Pearson correlation coefficient to verify the statistical significance of the correlation results. Only when the linear correlation coefficient passes the significance test is it used in subsequent abnormal user determination, thereby further improving the reliability of the identification results.

[0088] For users Let its null hypothesis be The alternative hypothesis is ,exist , The test statistic is calculated as follows:

[0089] ;

[0090] in, The sample size. Under the null hypothesis... Under the conditions established, Obeying the degree of freedom The t-distribution, To construct the significance level of the test, when If so, then reject the null hypothesis. The results suggest that there is a significant statistical correlation between the user's electricity consumption and the line loss rate; thus, the test results were obtained.

[0091] It should be noted that the Pearson correlation coefficient does not strictly follow a normal distribution under finite sample conditions. However, in engineering applications with sufficient sample size, cleaned outliers, and a true correlation coefficient close to zero, the correlation coefficient, as an aggregate of multiple sample statistics, can be approximated as a symmetrical distribution under the central limit theorem. Furthermore, by constructing a t-test statistic, it can satisfy the t-distribution assumption in an approximate engineering sense.

[0092] S6. Based on the abnormal judgment threshold range and the test results, a joint judgment is executed to obtain the identification results of electricity theft users on the target 10 kV non-economic operation line.

[0093] To avoid misjudgment based on a single indicator, this invention employs a joint decision mechanism. When the linear correlation coefficient is significantly higher than the anomaly judgment threshold range for line users, and this linear correlation coefficient passes the statistical significance test, the user can be considered to exhibit stable abnormal statistical correlation characteristics. The user was identified as a suspected abnormal user. A suspected abnormal user is a suspected electricity theft user, thus obtaining the electricity theft user identification result.

[0094] The formula for determining whether the linear correlation coefficient exceeds the anomaly detection threshold range is:

[0095] ;

[0096] The formula for determining whether the linear correlation coefficient passes the statistical significance test is:

[0097] .

[0098] This joint judgment mechanism can effectively suppress misjudgments caused by random fluctuations and improve the robustness of the identification results.

[0099] Example 2:

[0100] Embodiment 1 of this invention provides an application example of a method for identifying electricity theft users on 10 kV non-economically operating lines. The specific analysis object is the operating data of the Chunhua Substation 103 Kaidan line from January to February 2022, recorded in the integrated electricity consumption and synchronous line loss system. This period involves one 10 kV line, generating 65 daily line loss rate sample data points generated by the integrated electricity consumption and synchronous line loss system. After data quality preprocessing, 6 invalid and abnormal data points, accounting for approximately 9.2% of the total original sample, were removed due to meter reading failures, temporary system interference, etc. A total of 59 continuous and valid daily samples were obtained, covering 10 users on the line. The Pearson correlation coefficient between each user's electricity consumption and daily line loss rate was calculated, and a significance test was performed (taking...). Further calculations Statistics and The values ​​and results are shown in the table below (sorted by the absolute value of the linear correlation coefficient):

[0101]

[0102] Further calculation of the statistical parameters of the linear correlation coefficient for all users yields the mean. Standard deviation Take the sensitivity coefficient. Then the upper limit of the adaptive decision threshold The linear correlation coefficients for user lines A, B, and C are 0.605, 0.550, and 0.493, respectively. All values ​​are greater than 4, under the condition of 57 degrees of freedom. The value is much less than 0.01, indicating a significant correlation; the linear correlation coefficients of the remaining users are generally below 0.3, indicating insufficient statistical significance. In the example, the linear correlation coefficients of most normal users are concentrated below 0.2, distributed around the overall mean of the line. The three users are all significantly higher than the concentrated distribution range of the line's users, that is, significantly exceeding... The statistical scope it describes indicates that it has deviated from the conventional statistical behavior pattern of line users.

[0103] On-site verification confirmed that all three households were engaging in electricity theft by diverting power, thus verifying that the method of this invention has high accuracy, reliability, and engineering feasibility in 10 kV non-economical operating lines.

[0104] This invention has been described through embodiments. Those skilled in the art will understand that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, under the teachings of this invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, this invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of this invention.

Claims

1. A method for identifying electricity theft users on 10 kV non-economically operating lines, characterized in that, Includes the following steps: Collect operational data for the target 10 kV non-economical operating line, including the line's daily line loss rate and the user's daily electricity consumption; The linear correlation coefficient is obtained by measuring the degree of linear correlation between the daily line loss rate and the daily electricity consumption of users. An adaptive threshold model is constructed and solved based on the linear correlation coefficient to obtain the adaptive anomaly detection threshold range; The statistical significance test was performed based on the linear correlation coefficient, and the test results were obtained. A joint decision is made based on the abnormal decision threshold range and the test results. The decision result is the identification result of the electricity theft user on the target 10 kV non-economic operation line.

2. The method for identifying electricity theft users on 10 kV non-economically operating lines according to claim 1, characterized in that, Also includes: Missing values ​​are repaired in the collected operational data. When the number of missing days is less than a preset threshold, the mean interpolation method of adjacent time windows is used for repair. When the missing percentage exceeds the preset threshold or exhibits long-term missing characteristics, it is directly removed; Outlier removal was performed on the collected daily line loss rate data. Let the sample sequence of the collected daily line loss rate data be: ； in, For the first Daily line loss rate; The daily line loss rate sample series was sorted in ascending order of numerical value, and the first quartile of its three core statistics was calculated. Third and quartiles and interquartile range ; Outlier criteria are: or ; Remove samples that meet the above criteria.

3. The method for identifying electricity theft users on 10 kV non-economically operating lines according to claim 1, characterized in that, The measure of the linear correlation between the daily line loss rate and the daily electricity consumption of users is as follows: The Pearson correlation coefficient is used to measure: ； in, For the first The linear correlation coefficient of each user For the first Daily line loss rate, For the first The user Daily electricity consumption This represents the average of the daily line loss rate data samples. This represents the average daily electricity consumption data sample for users.

4. The method for identifying electricity theft users on 10 kV non-economically operating lines according to claim 1, characterized in that, The specific steps for constructing and solving the adaptive threshold model based on the linear correlation coefficient are as follows: Calculate the mean of the linear correlation coefficient. with standard deviation : ； ； in, For the first The linear correlation coefficient of each user Total number of users Define the adaptive anomaly detection threshold range as follows: The calculation formula is: ； ； in, This is the sensitivity coefficient.

5. The method for identifying electricity theft users on 10 kV non-economically operating lines according to claim 1, characterized in that, The statistical significance test based on the linear correlation coefficient is specifically as follows: For users Let its null hypothesis be The alternative hypothesis is ,exist , The test statistic is calculated as follows: ； in, For the sample size, under the null hypothesis Under the conditions established, Obeying the degree of freedom The t-distribution, To construct the significance level of the test, when If so, then reject the null hypothesis. .

6. The method for identifying electricity theft users on 10 kV non-economically operating lines according to claim 1, characterized in that, The joint decision based on the anomaly judgment threshold range and the test results is as follows: when the linear correlation coefficient is higher than the anomaly judgment threshold range for line users, and the linear correlation coefficient passes the statistical significance test, the user is judged as a suspected abnormal user, that is, a suspected electricity theft user.

Images