A method and system for detecting abnormal power users in a power grid

Abstract

The invention discloses a method and system for detecting abnormal power users of a power grid, which relates to the technical field of electric load detection, and comprises: obtaining power data and meteorological factor data of at least one transmission line within a set time period, and daily power consumption data of multiple power users under the transmission line; calculating the daily line loss rate of the transmission line; inputting a historical line loss rate time series sequence and a meteorological factor data time series sequence before a test day into a line loss rate prediction model based on CNN-BiLSTM, and outputting a predicted value of the line loss rate on the test day; locating a line loss abnormal line and a line loss abnormal mutation day according to a difference between the predicted value of the line loss rate on the test day and an actual value of the line loss rate; calculating a correlation coefficient according to the line loss value of the line loss abnormal line on the line loss mutation abnormal day and the day before and the daily power consumption of multiple users under the line loss abnormal line, performing K-means clustering on all power users according to the correlation coefficient, and quickly identifying abnormal power users according to the clustering result.

Description

Technical Field

[0001] The present invention relates to the technical field of electric load detection, and in particular to a method and system for detecting abnormal power users in a power grid. Background Art

[0002] Due to the existence of resistance, inductance and other factors in the transmission line, electric energy will be lost during the transmission process. This loss is called line loss. The percentage of line loss to output power is called line loss rate. This line loss rate is an important indicator to measure the healthy economic development of the distribution network. Line loss includes technical line loss and non-technical line loss. Among them, technical line loss is an inevitable and necessary loss in the transmission process. When the equipment status and the grid operation mode remain unchanged, the technical line loss is in a relatively constant state. Therefore, the change of line loss rate mainly depends on non-technical line loss. Non-technical loss (NTL) refers to the part of the grid transmission and distribution loss that cannot be explained by technology. Abnormal power consumption behavior of users, including power theft, is the main cause of NTL. Abnormal power consumption behavior such as power theft not only affects the technical indicators of the line loss rate of the power grid, causing power grid enterprises to suffer power loss and economic benefit loss, but also poses serious safety hazards.

[0003] At present, the detection of abnormal electricity users based on abnormal line loss is mainly carried out through manual meter reading. Relying on the staff's many years of professional experience, the user's historical electricity consumption data and the substation line loss rate data are manually compared and analyzed. On the one hand, the detection relies on the staff's relevant work experience, and subjective judgment cannot guarantee the accuracy of the detection results. On the other hand, it requires analysis and comparison of a large amount of user electricity consumption data, which will consume a lot of manpower and energy. Summary of the invention

[0004] In order to address the deficiencies of the above-mentioned prior art, the present invention provides a method and system for detecting abnormal electricity users in a power grid, which considers the correlation between the line loss rate of the electricity user's transmission line and the user's daily electricity consumption, introduces the line loss rate into the detection of abnormal electricity users, predicts the line loss rate of the next day based on the historical line loss rate of each transmission line, locates the line loss abnormal line and the time when the abnormality occurs by comparing the predicted value and the actual value of the line loss rate, and then measures and identifies the abnormal electricity users under the abnormal line based on the correlation between the actual line loss change of the transmission line and the change of the user's electricity consumption, thereby achieving accurate identification from the abnormal transmission line to the abnormal electricity users under the line, reducing the workload of data comparison and analysis, reducing the calculation complexity while ensuring the recognition accuracy, and improving the detection efficiency.

[0005] In a first aspect, the present invention provides a method for detecting abnormal electricity users in a power grid.

[0006] A method for detecting abnormal power users in a power grid, comprising:

[0007] Obtaining the electricity consumption data and meteorological factor data of at least one transmission line within a set time period, as well as the daily electricity consumption data of multiple electricity users under the transmission line;

[0008] Based on the power consumption data of the transmission line, the daily line loss rate of the transmission line is calculated to form a time series of line loss rate;

[0009] The historical line loss rate time series and meteorological factor data time series before the test day are input into the line loss rate prediction model based on CNN-BiLSTM, and the predicted value of the line loss rate on the test day is output;

[0010] According to the difference between the predicted value of the line loss rate on the day to be tested and the actual value of the line loss rate, the line loss abnormality and the day of abnormal line loss mutation are located by comparing with the set value;

[0011] According to the line loss values ​​of the abnormal line on the day of abnormal line loss mutation and the day before, and the daily power consumption of multiple users on the abnormal line loss line, the correlation coefficient is calculated, and K-means clustering is performed on all electricity users according to the correlation coefficient. The abnormal electricity users are identified based on the clustering results.

[0012] In a further technical solution, the power data includes active output power and active input power with a time granularity of one day;

[0013] For each day, the line loss rate of the transmission line with a time granularity of day is calculated based on the active output power and active input power with a time granularity of day, and this line loss rate is the actual value of the line loss rate.

[0014] In a further technical solution, the acquisition of the line loss rate prediction value includes:

[0015] The discrete wavelet transform is used to decompose the historical line loss rate time series, and the low-frequency component time series and the high-frequency component time series are obtained by the decomposition.

[0016] The high-frequency component time series is decomposed twice using variational mode decomposition, and multiple sub-high-frequency component time series are obtained by the secondary decomposition;

[0017] The time series of multiple modal components after primary and secondary decomposition and the time series of meteorological factor data are input into the line loss rate prediction model based on CNN-BiLSTM, and the predicted values ​​of multiple modal components are output;

[0018] The predicted values ​​of multiple modal components are reconstructed and superimposed to obtain the predicted value of line loss rate on the test day.

[0019] A further technical solution is to screen the meteorological factors, including:

[0020] Obtaining the line loss rate of the transmission line and multiple meteorological factor data to be screened at the location of the transmission line; the meteorological factors to be screened include the average temperature, average wind speed, maximum temperature, minimum temperature, maximum wind speed, and minimum wind speed at a time granularity of the day;

[0021] The maximum mutual information correlation calculation is performed on each meteorological factor data to be screened and each modal component after the primary and secondary decomposition of the line loss rate. The calculated results are sorted to screen out the most relevant meteorological factors.

[0022] According to a further technical solution, the most relevant meteorological factors screened out include the average temperature, average wind speed, maximum temperature and minimum temperature at a time granularity of the day.

[0023] In a further technical solution, the correlation coefficient is calculated as follows:

[0024] Determine the difference in daily electricity consumption of multiple users on the line with abnormal line loss based on the daily electricity consumption on the day of abnormal line loss mutation and the day before;

[0025] Determine the line loss difference based on the line loss value of the abnormal line loss line on the day of abnormal line loss mutation and the day before;

[0026] Based on the ratio of the difference in daily power consumption of users to the difference in line loss, the correlation coefficient is obtained.

[0027] In a further technical solution, the identification of abnormal electricity users includes:

[0028] The calculated correlation coefficients of all users are used as the sample data set to be classified, and the number of clusters is set to two categories: abnormal and normal. K-means clustering is used to divide users into two categories according to the size of the correlation coefficient to identify abnormal electricity users and normal electricity users.

[0029] In a second aspect, the present disclosure provides a system for detecting abnormal electricity users in a power grid.

[0030] A power grid abnormal power user detection system, comprising:

[0031] A data acquisition module is used to acquire the power consumption data and meteorological factor data of at least one power transmission line within a set time period, as well as the daily power consumption data of multiple power users under the power transmission line;

[0032] A line loss rate calculation module is used to calculate the daily line loss rate of the transmission line based on the power consumption data of the transmission line to form a line loss rate time series;

[0033] The line loss rate prediction module is used to input the historical line loss rate time series and the meteorological factor data time series before the test day into the line loss rate prediction model based on CNN-BiLSTM, and output the predicted value of the line loss rate on the test day;

[0034] The abnormal line loss line detection module is used to compare and determine the difference between the predicted line loss rate and the actual line loss rate on the day to be tested with the set value, and locate the abnormal line loss line and the day of abnormal line loss mutation;

[0035] The abnormal electricity user identification module is used to calculate the correlation coefficient based on the line loss values ​​of the abnormal line loss on the day of abnormal line loss mutation and the day before, and the daily power consumption of multiple users under the abnormal line loss line, and perform K-means clustering on all electricity users according to the correlation coefficient, and identify abnormal electricity users based on the clustering results.

[0036] In a third aspect, the present disclosure further provides an electronic device, comprising a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the steps of the method described in the first aspect are completed.

[0037] In a fourth aspect, the present disclosure further provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, complete the steps of the method described in the first aspect.

[0038] One or more of the above technical solutions have the following beneficial effects:

[0039] 1. The present invention provides a method and system for detecting abnormal power users in a power grid. The method considers the correlation between the line loss rate of the power user's transmission line and the user's daily power consumption, introduces the line loss rate into the detection of abnormal power users, predicts the line loss rate of the next day based on the historical line loss rate of each transmission line, locates the line loss abnormal line and the time when the abnormality occurs by comparing the predicted value and the actual value of the line loss rate, and then measures and identifies the abnormal power users under the abnormal line based on the correlation between the actual line loss change of the transmission line and the change of the user's power consumption, thereby achieving accurate identification from the abnormal transmission line to the abnormal power users under the line, reducing the workload of data comparison and analysis, reducing the calculation complexity while ensuring the recognition accuracy, and improving the detection efficiency.

[0040] 2. Considering the problem that line loss data is nonlinear and highly volatile, which makes it difficult to accurately predict line loss, the present invention performs primary and secondary decomposition on historical line loss data to obtain multiple modal components, and calculates the mutual information correlation coefficient between each external factor and each modal component, from which the influencing factors with a greater impact on each modal component are screened to achieve the purpose of feature dimensionality reduction. Finally, each component and the corresponding influencing factor are input into CNN-BiLSTM, and the feature extraction capability of CNN is used to extract features from the data after secondary decomposition. The extracted features are input into BiLSTM to output the prediction results of each component. Finally, the predicted values ​​of each component are reconstructed and superimposed and then output, so as to obtain a more accurate line loss prediction value, thereby ensuring the accuracy of subsequent abnormality identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0042] Figure 1 The present invention is an overall flow chart of the method for detecting abnormal power users in a power grid according to an embodiment of the present invention. DETAILED DESCRIPTION

[0043] It should be noted that the following detailed descriptions are exemplary only, are intended to describe specific embodiments, are intended to provide further explanation of the present invention, and are not intended to limit exemplary embodiments according to the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those of ordinary skill in the art to which the present invention belongs. In addition, it should also be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0044] Embodiment 1

[0045] The power grid is developed and interconnected by many sub-grids. Usually, the power grid can be divided into transmission network, secondary transmission network, high-voltage distribution network and low-voltage distribution network. Among them, the transmission network is composed of trunk transmission lines with a voltage of more than 220kV, which connects large power plants, large-capacity users and adjacent sub-grids; the voltage of the secondary transmission network is generally 110~220kV, which is connected to the transmission network above and the high-voltage distribution network below, usually connecting regional power plants and large users; the distribution network is a network that supplies power to medium and small users. The 10~35kV is called the high-voltage distribution network, and the 1kV or less is called the low-voltage distribution network. With respect to multiple transmission lines in the power grid, each transmission line is connected to multiple electricity users. Considering the huge amount of electricity consumption data of electricity users, it is difficult to perform abnormality detection directly based on the electricity consumption data of all electricity users. To this end, this embodiment proposes a method for detecting abnormal electricity users in a power grid, which introduces line loss of transmission lines. By first determining the abnormal line based on the line loss, and then determining the abnormal electricity users under the abnormal line based on the correlation between the line loss and the electricity consumption of the users, more efficient and accurate abnormal electricity user detection is achieved, avoiding complex and cumbersome data calculation and processing.

[0046] This embodiment provides a method for detecting abnormal power users in a power grid. Figure 1 As shown, the specific steps include:

[0047] Step S1, obtaining the power consumption data and meteorological factor data of at least one power transmission line within a set time period, as well as the daily power consumption data of multiple power users under the power transmission line;

[0048] Step S2: based on the power consumption data of the transmission line, calculate the daily line loss rate of the transmission line to form a line loss rate time series;

[0049] Step S3, inputting the historical line loss rate time series sequence and the meteorological factor data time series sequence before the day to be measured into the line loss rate prediction model based on CNN-BiLSTM, and outputting the line loss rate prediction value of the day to be measured;

[0050] Step S4, comparing the difference between the predicted value of the line loss rate on the day to be measured and the actual value of the line loss rate with the set value, and locating the line with abnormal line loss and the day with abnormal line loss mutation;

[0051] Step S5, calculate the correlation coefficient based on the line loss values ​​of the abnormal line loss on the day of abnormal line loss mutation and the day before, and the daily power consumption of multiple users on the abnormal line loss line, perform K-means clustering on all power users based on the correlation coefficient, and identify abnormal power users based on the clustering results.

[0052] The following content introduces the abnormal power user detection method of the power grid proposed in this embodiment in more detail.

[0053] In step S1, the power data and meteorological factor data of at least one transmission line within a set time period, as well as the daily power consumption data of multiple power users under the transmission line are obtained. Among them, the power data refers to the data sampled by the electric energy metering device in the transmission line at fixed intervals, including the active output power and active input power with a time granularity of one day; the meteorological data refers to the meteorological data of the location of the transmission line, which refers to the data obtained from the meteorological information system, including the average temperature, average wind speed, maximum temperature and minimum temperature with a time granularity of one day.

[0054] In step S2, based on the power consumption data of the transmission line, the daily line loss rate of the transmission line is calculated to form a line loss rate time series. In this embodiment, the line loss rate is calculated based on various data such as voltage, current, load and other data during the operation of the transmission line. Specifically, for each day, the line loss rate with a time granularity of one day is calculated based on the active output power and active input power with a time granularity of one day. The line loss rate is the actual value of the line loss rate, and the line loss rate time series is constructed in this way, laying the foundation for subsequent data processing. Among them, the calculation formula of the line loss rate is:

[0055] ;

[0056] In the above formula, is the line loss rate with a time granularity of day, , They respectively represent the active output power and active input power with a time granularity of day.

[0057] In step S3, the historical line loss rate time series sequence and the meteorological factor data time series sequence before the day to be measured are input into the line loss rate prediction model based on CNN-BiLSTM, and the line loss rate prediction value of the day to be measured is output.

[0058] Taking into account that the power consumption curve and line loss rate curve of abnormal power users such as power theft users have a certain correlation, therefore, this embodiment firstly performs a preliminary judgment on a large scale and a large area based on the line loss rate of the transmission line. The line loss of the transmission line mainly includes technical line loss and non-technical line loss. The technical line loss of a normally operating transmission line is in a relatively constant state, and its fluctuation mainly comes from non-technical line loss. Under normal circumstances, the non-technical line loss is affected by external weather and other factors and has a certain fluctuation, but the overall trend is still regular. When there are certain abnormal behaviors such as power theft, the line loss of the transmission line will change to a large extent, so that the line loss abnormality line can be preliminarily judged.

[0059] In this embodiment, the set time period is obtained as t The power data of at least one transmission line in the day is then calculated. t Time series of line loss rate of the day , with the t Day is the day to be tested, and the historical line loss time series is constructed , using this sequence to combine t The line loss rate prediction model based on CNN-BiLSTM is used to obtain the line loss rate prediction value of the test day. In the above manner, a prediction is made based on the historical line loss time series of normally operating transmission lines to obtain a predicted line loss rate for the next day. The predicted value is an estimated value that conforms to the normal law of line loss. The predicted value is compared with the actual value. When there is a large difference and the difference exceeds the set value, the transmission line is considered to be abnormal.

[0060] Furthermore, in order to improve the prediction accuracy of the line loss rate, this embodiment also proposes a method suitable for predicting the line loss rate, which specifically includes:

[0061] Step S3.1, using discrete wavelet transform to decompose the historical line loss rate time series, and obtain a low-frequency component time series and a high-frequency component time series.

[0062] Step S3.2: perform secondary decomposition on the high-frequency component time series sequence using variational mode decomposition, and obtain multiple secondary high-frequency component time series sequences through secondary decomposition.

[0063] Specifically, considering that the line loss data of the distribution network has certain randomness and nonlinearity, it may contain multiple fluctuation patterns at a time point. For this reason, this embodiment uses discrete wavelet transform to decompose the line loss rate data to obtain high-frequency, low-frequency and other information of the data. The above information is input into the neural network, which enables the neural network to learn the features of each dimension after decomposition in a targeted manner and improve the sophistication of feature extraction.

[0064] Furthermore, considering that the original line loss data still has high complexity and irregularity in its high-frequency components after being decomposed by discrete wavelet transform, if the decomposed components are directly predicted, the prediction results may have large errors. Variational mode decomposition is an adaptive, fully non-recursive variational mode signal processing method that can effectively process non-stationary and nonlinear signals, and quickly extract several inherent modal components of the signal from time series signals. It has strong robustness to noise. Therefore, in order to further extract the implicit information in the high-frequency components, this embodiment performs variational mode decomposition on the high-frequency components after wavelet decomposition again to weaken their nonlinear characteristics, and obtain sub-high-frequency components of multiple modes after decomposition.

[0065] Step S3.3: Input the multiple modal component time series after primary and secondary decomposition and the meteorological factor data time series into the line loss rate prediction model based on CNN-BiLSTM, and output the predicted values ​​of multiple modal components.

[0066] Specifically, since the conductors of high-voltage transmission lines are directly exposed to the external environment, external meteorological factors such as temperature and wind speed will have a certain impact on the resistance of the conductors, thereby causing disturbances in the line loss rate. Therefore, in this embodiment, meteorological data is also used as a factor affecting the transmission line and input into the prediction model to improve the accuracy of the prediction. Among them, the screening and determination of the most important meteorological factors include:

[0067] Obtain the line loss rate of the transmission line and multiple meteorological factor data to be screened at the location of the transmission line; the meteorological factors to be screened include the average temperature, average wind speed, maximum temperature, minimum temperature, maximum wind speed, minimum wind speed, etc. at a time granularity of the day;

[0068] The maximum mutual information correlation calculation is performed on each meteorological factor data to be screened and each modal component after the primary and secondary decomposition of the line loss rate. The calculated results are sorted to screen out the most relevant meteorological factors, including: the average temperature, average wind speed, maximum temperature and minimum temperature with a time granularity of the day.

[0069] In addition, the convolutional bidirectional long short-term memory neural network CNN-BiLSTM is used for prediction. The convolutional network has good feature extraction capabilities, can capture local patterns and trends in time series data, and better understand the local features of time series data; BiLSTM can capture long-term dependencies in sequence data. Through the bidirectional network, the past and future hidden layer states can be recursively fed back, and historical data can be trained in both positive and negative directions to learn more effective information. CNN-BiLSTM can be used to extract and learn features in time series data at different levels, better use time series information for prediction, and overcome their respective limitations to improve the generalization ability of the model.

[0070] In this embodiment, the decomposed data is input into a one-dimensional convolutional layer for preliminary feature extraction, and the extracted features are subjected to maximum pooling to retain important features and achieve the purpose of dimensionality reduction, and a padding layer is passed to ensure that the length of the data does not change and that important feature information is not lost: then, the output features are subjected to a one-dimensional convolutional layer for secondary feature purification again, and dimensionality reduction is performed through maximum pooling, and the data from which features are extracted through the above two convolutional layers is input into a BiLSTM layer, and the network weights are updated through forward propagation combined with gradient calculation; finally, the output result of the BiLSTM is output after passing through a fully connected layer to obtain the final prediction result.

[0071] Step S3.4, reconstruct and superimpose the predicted values ​​of multiple modal components to obtain the predicted value of line loss rate on the test day, that is, obtain the first t Line loss rate prediction value for the day to be tested .

[0072] In step S4, the difference between the predicted line loss rate and the actual line loss rate on the day to be measured is compared with the set value to locate the line with abnormal line loss and the day with abnormal line loss mutation.

[0073] Specifically, for each transmission line, the line loss rate prediction value of the line is and actual value of line loss rate Make a difference and compare the difference with the set value. When the difference exceeds the set value, it is considered that the fluctuation of the predicted day is large and there is an abnormal fluctuation. At this time, the line loss abnormal line and the line loss abnormal mutation day are located.

[0074] In step S5, the correlation coefficient is calculated based on the line loss values ​​of the abnormal line loss on the day of abnormal line loss mutation and the day before, and the daily power consumption of multiple users on the abnormal line loss line, and K-means clustering is performed on all electricity users based on the correlation coefficient, and abnormal electricity users are identified based on the clustering results.

[0075] Specifically, the daily electricity consumption curve of the electricity stealing user and the line loss rate curve show a certain regular relationship. When the daily electricity consumption curve of the user and the line loss rate curve show a strong correlation, the user is suspected of stealing electricity. In order to reflect the degree of change in line loss caused by the change in the daily electricity consumption of the user, the ratio of the change value of the daily electricity consumption of the user to the change value of the line loss electricity is used as the correlation coefficient, and the strong correlation judgment of the line loss caused by the user is made based on the correlation coefficient. Among them, according to step S4, it is determined that the line loss rate has a sudden change from the t-1 day to the t day, and the correlation coefficient is calculated as:

[0076] According to the daily power consumption of multiple users on the abnormal line loss line on the day of abnormal line loss mutation and the day before, the difference in daily power consumption of users is determined, that is, ;

[0077] According to the line loss values ​​of the abnormal line loss line on the day of abnormal line loss mutation and the day before, the line loss difference is determined, that is, ;

[0078] According to the ratio of the difference in daily power consumption of users to the difference in line loss rate, as the correlation coefficient, the calculation formula is: .

[0079] In the above formula, is the difference in daily power consumption of users two days before and after the line loss rate mutation point. For the t Daily electricity consumption by users, For the t -1 day's electricity consumption of the user; It is the difference of line loss power two days before and after the line loss rate mutation point. The line loss power is the line loss value. The line loss value is related to the line loss rate. The line loss value is actually the difference between the active output power and the active input power with a time granularity of one day. For the t The line loss value of the day, For the t -1 day line loss value. Correlation coefficient X The larger the value is, the greater the correlation between the user's daily electricity consumption change and the line loss electricity change, that is, the user has a greater impact on the line loss. The user is highly suspected of stealing electricity and is judged to be an abnormal electricity user.

[0080] By locking in abnormal electricity users in the above manner, the scope of abnormal users can be further narrowed on the basis of conventional measurement of abnormalities based on the correlation between users' daily electricity consumption and line loss rate, with a high accuracy rate.

[0081] Furthermore, considering that the correlation coefficient threshold is manually selected, the judgment is made based on the threshold, and the electricity users with a value greater than the threshold are selected as abnormal users. The process of selecting the threshold is highly subjective and may affect the detection efficiency and detection accuracy. For example, if the threshold is set too high, abnormal users may be missed; and if the threshold is set too low, the selection is inaccurate, and the workload of subsequent analysis is increased, resulting in low work efficiency. For this reason, the K-means clustering algorithm is also introduced in this embodiment, and the calculated correlation coefficients of all users are used as sample data sets to be classified, and the number of clusters is set to abnormal and normal. K-means clustering is used to classify users into two categories according to the size of the correlation coefficient, so as to identify abnormal electricity users and normal electricity users.

[0082] Specifically, set the number of clusters k =2, initialization k The calculated correlation coefficients of all users are used as the sample data set to be classified. k clusters; sample data points for calculating correlation coefficients arrive k The Euclidean distance of the centroids, The data points are classified into the cluster with the closest centroid to it until all the data points in the data set are classified. For each cluster that has been classified, the mean of all the data points in the cluster is calculated, and the obtained value becomes the new centroid. k The clusters obtained k A new centroid; based on k A new centroid is generated, and the sample data classification steps are repeated to update the data in the cluster; after each round of classification, the centroid update step is repeated in the cluster until the centroid position does not change, at which point clustering ends. The K-means clustering method is used to cluster the X values. After clustering by the clustering algorithm, the threshold can be selected without manual judgment, which can effectively reduce labor costs and improve detection efficiency.

[0083] Embodiment 2

[0084] This embodiment provides a system for detecting abnormal power users in a power grid, including:

[0085] A data acquisition module is used to acquire the power consumption data and meteorological factor data of at least one power transmission line within a set time period, as well as the daily power consumption data of multiple power users under the power transmission line;

[0086] A line loss rate calculation module is used to calculate the daily line loss rate of the transmission line based on the power consumption data of the transmission line to form a line loss rate time series;

[0087] The line loss rate prediction module is used to input the historical line loss rate time series and the meteorological factor data time series before the test day into the line loss rate prediction model based on CNN-BiLSTM, and output the predicted value of the line loss rate on the test day;

[0088] The abnormal line loss line detection module is used to compare and determine the difference between the predicted line loss rate and the actual line loss rate on the day to be tested with the set value, and locate the abnormal line loss line and the day of abnormal line loss mutation;

[0089] The abnormal electricity user identification module is used to calculate the correlation coefficient based on the line loss values ​​of the abnormal line loss on the day of abnormal line loss mutation and the day before, and the daily power consumption of multiple users under the abnormal line loss line, and perform K-means clustering on all electricity users according to the correlation coefficient, and identify abnormal electricity users based on the clustering results.

[0090] Embodiment 3

[0091] This embodiment provides an electronic device, including a memory and a processor, and computer instructions stored in the memory and running on the processor. When the computer instructions are run by the processor, the steps in the above-mentioned method for detecting abnormal power users in the power grid are completed.

[0092] Embodiment 4

[0093] This embodiment further provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the steps in the method for detecting abnormal power users in a power grid as described above are completed.

[0094] The steps involved in the above embodiments 2 to 4 correspond to the method embodiment 1, and the specific implementation methods can refer to the relevant description part of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0095] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0096] The above description is only a preferred embodiment of the present invention. Although the specific implementation mode of the present invention is described in conjunction with the accompanying drawings, it is not a limitation of the protection scope of the present invention. Those skilled in the art should understand that on the basis of the technical solution of the present invention, various modifications or variations that can be made by those skilled in the art without creative work are still within the protection scope of the present invention.

Claims

1. A method for detecting abnormal power users in a power grid, characterized in that: include: Obtaining the electricity consumption data and meteorological factor data of at least one transmission line within a set time period, as well as the daily electricity consumption data of multiple electricity users under the transmission line; Based on the power consumption data of the transmission line, the daily line loss rate of the transmission line is calculated to form a time series of line loss rate; The historical line loss rate time series and the meteorological factor data time series before the day to be measured are input into the line loss rate prediction model based on CNN-BiLSTM, and the line loss rate prediction value of the day to be measured is output; the acquisition of the line loss rate prediction value includes: using discrete wavelet transform to perform a primary decomposition on the historical line loss rate time series, and the primary decomposition obtains a low-frequency component time series sequence and a high-frequency component time series sequence; using variational mode decomposition to perform a secondary decomposition on the high-frequency component time series sequence, and the secondary decomposition obtains multiple sub-high-frequency component time series sequences; the multiple modal component time series sequences after the primary and secondary decompositions and the meteorological factor data time series sequence are input into the line loss rate prediction model based on CNN-BiLSTM, and the prediction values ​​of multiple modal components are output; the prediction values ​​of multiple modal components are reconstructed and superimposed to obtain the line loss rate prediction value of the day to be measured; According to the difference between the predicted value of the line loss rate on the day to be tested and the actual value of the line loss rate, a comparison is made with the set value to determine the location of the abnormal line loss line and the abnormal line loss mutation day; according to the line loss value of the abnormal line loss line on the abnormal line loss mutation day and the day before and the daily power consumption of multiple users under the abnormal line loss line, a correlation coefficient is calculated, and K-means clustering is performed on all power users according to the correlation coefficient, and the abnormal power users are identified according to the clustering result; the correlation coefficient is: according to the daily power consumption of multiple users under the abnormal line loss line on the abnormal line loss mutation day and the day before, the difference in daily power consumption of users is determined; according to the line loss value of the abnormal line loss line on the abnormal line loss mutation day and the day before, the line loss difference is determined; the correlation coefficient is obtained based on the ratio of the difference in daily power consumption of users to the difference in line loss.

2. A method for detecting abnormal power users in a power grid as claimed in claim 1, characterized in that: The power data includes active output power and active input power with a time granularity of one day; For each day, the line loss rate of the transmission line with a time granularity of day is calculated based on the active output power and active input power with a time granularity of day, and this line loss rate is the actual value of the line loss rate.

3. A method for detecting abnormal power users in a power grid as claimed in claim 1, characterized in that: The screening of meteorological factors includes: Obtaining the line loss rate of the transmission line and multiple meteorological factor data to be screened at the location of the transmission line; the meteorological factors to be screened include the average temperature, average wind speed, maximum temperature, minimum temperature, maximum wind speed, and minimum wind speed at a time granularity of the day; The maximum mutual information correlation calculation is performed on each meteorological factor data to be screened and each modal component after the primary and secondary decomposition of the line loss rate. The calculated results are sorted to screen out the most relevant meteorological factors.

4. A method for detecting abnormal power users in a power grid as claimed in claim 3, characterized in that: The most relevant meteorological factors screened out include the average temperature, average wind speed, maximum temperature and minimum temperature with a time granularity of day.

5. A method for detecting abnormal power users in a power grid as claimed in claim 1, characterized in that: The identification of abnormal electricity users includes: The calculated correlation coefficients of all users are used as the sample data set to be classified, and the number of clusters is set to two categories: abnormal and normal. K-means clustering is used to divide users into two categories according to the size of the correlation coefficient to identify abnormal electricity users and normal electricity users.

6. A power grid abnormal power user detection system, characterized in that: include: A data acquisition module is used to acquire the power consumption data and meteorological factor data of at least one power transmission line within a set time period, as well as the daily power consumption data of multiple power users under the power transmission line; A line loss rate calculation module is used to calculate the daily line loss rate of the transmission line based on the power consumption data of the transmission line to form a line loss rate time series; The line loss rate prediction module is used to input the historical line loss rate time series sequence and the meteorological factor data time series sequence before the test day into the line loss rate prediction model based on CNN-BiLSTM, and output the line loss rate prediction value of the test day; the acquisition of the line loss rate prediction value includes: using discrete wavelet transform to perform a primary decomposition on the historical line loss rate time series sequence, and the primary decomposition obtains a low-frequency component time series sequence and a high-frequency component time series sequence; using variational mode decomposition to perform a secondary decomposition on the high-frequency component time series sequence, and the secondary decomposition obtains multiple sub-high-frequency component time series sequences; the multiple modal component time series sequences after the primary and secondary decompositions and the meteorological factor data time series sequence are input into the line loss rate prediction model based on CNN-BiLSTM, and the prediction values ​​of multiple modal components are output; the prediction values ​​of multiple modal components are reconstructed and superimposed to obtain the line loss rate prediction value of the test day; The abnormal line loss line detection module is used to compare and determine the difference between the predicted line loss rate and the actual line loss rate on the day to be tested with the set value, and locate the abnormal line loss line and the day of abnormal line loss mutation; The abnormal electricity user identification module is used to calculate the correlation coefficient based on the line loss value of the abnormal line on the day of abnormal line loss mutation and the day before and the daily electricity consumption of multiple users under the abnormal line loss line, and perform K-means clustering on all electricity users according to the correlation coefficient, and identify the abnormal electricity users according to the clustering result; the correlation coefficient is: according to the daily electricity consumption of multiple users under the abnormal line loss line on the day of abnormal line loss mutation and the day before, determine the user's daily electricity consumption difference; according to the line loss value of the abnormal line loss line on the day of abnormal line loss mutation and the day before, determine the line loss difference; the correlation coefficient is obtained based on the ratio of the user's daily electricity consumption difference to the line loss difference.

7. A power grid abnormal power user detection system as claimed in claim 6, characterized in that: The power data includes active output power and active input power with a time granularity of one day; For each day, the line loss rate of the transmission line with a time granularity of day is calculated based on the active output power and active input power with a time granularity of day, and this line loss rate is the actual value of the line loss rate.

8. A power grid abnormal electricity user detection system as claimed in claim 6, characterized in that: The screening of meteorological factors includes: Obtaining the line loss rate of the transmission line and multiple meteorological factor data to be screened at the location of the transmission line; the meteorological factors to be screened include the average temperature, average wind speed, maximum temperature, minimum temperature, maximum wind speed, and minimum wind speed at a time granularity of the day; The maximum mutual information correlation calculation is performed on each meteorological factor data to be screened and each modal component after the primary and secondary decomposition of the line loss rate. The calculated results are sorted to screen out the most relevant meteorological factors.

9. An electronic device, characterized in that: The method comprises a memory and a processor, and computer instructions stored in the memory and executed on the processor. When the computer instructions are executed by the processor, the steps of a method for detecting abnormal power users in a power grid as claimed in any one of claims 1 to 5 are completed.

10. A computer-readable storage medium, characterized in that: Used to store computer instructions, which, when executed by a processor, complete the steps of a method for detecting abnormal power users in a power grid as described in any one of claims 1-5.

Images