Method and system for processing electrical energy monitoring data

Abstract

The invention belongs to the technical field of data processing, and discloses a method and a system for processing electrical energy monitoring data. Comprising the steps of collecting power data of different users; constructing a user space association network, and determining an adjacency relation set of each user; historical data of each user is obtained, a corresponding electricity utilization reference model is constructed, and trend component data and periodic component data are extracted; extracting random component data from the power data of each user; identifying reverse fluctuation modes among different users in the same adjacency relation set, and obtaining suspected user pairs; comprehensively calculating the power utilization inverse correlation degree of each group of suspected user pairs; extracting a power utilization feature set from the power data of each group of suspected user pairs, and identifying electricity stealing users in combination with the power utilization inverse correlation degree; according to the invention, the electricity stealing behavior in the resident area can be accurately identified, the accuracy and timeliness of electricity stealing detection are improved, and the safety stability and resource utilization efficiency of a power distribution network are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and more particularly to a method and system for processing electric power monitoring data. Background Art

[0002] Patent publication number CN115049410A discloses a method, device, electronic device, and computer-readable storage medium for identifying electricity theft. The method includes: determining the user corresponding to the collected information and the electricity usage scenario to which the user belongs based on the user identifier of abnormal collected information, and determining all collected information of the user in the electricity theft interval from the cache as the target information set, determining the electricity theft analysis model corresponding to the electricity usage scenario, and then analyzing the user's electricity usage in the electricity theft interval based on the electricity theft analysis model and the feature value of each collected information in the target information set to obtain the user's electricity theft analysis value. Then, based on the electricity theft analysis value, it is determined whether the user's electricity usage constitutes electricity theft, thereby realizing real-time monitoring and identification. In addition, electricity usage analysis is performed using an electricity theft analysis model specific to the electricity usage scenario to achieve precise analysis, thereby improving the accuracy of electricity theft identification and enhancing the identification effect.

[0003] However, although the above technology can identify electricity theft, the identification process mainly relies on the behavior analysis of a single user and the model matching of preset feature quantities, and lacks in-depth exploration of the relationship between electricity usage behaviors among users; moreover, the above method is highly dependent on preset weighting factors and threshold settings, and has limited adaptability and generalization capabilities for complex electricity usage scenarios, resulting in a high misjudgment rate or delayed recognition, thereby limiting the robustness and scalability of the identification system.

[0004] In view of this, the present invention proposes a method and system for processing electric energy monitoring data to solve the above problems. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art and achieve the above-mentioned objectives, the present invention provides the following technical solutions: a method for processing electric energy monitoring data, comprising: Collect electricity data from different users in residential areas; Obtain user location information and grid topology, build a user spatial association network, and determine the adjacency set of each user; Obtain historical data for each user and build a corresponding electricity usage benchmark model. Extract trend component data and periodic component data from each electricity usage benchmark model. Based on the trend component data and periodic component data, the time series of each user's electricity data is decomposed to extract the random component data; Based on random component data, sliding window analysis is applied to identify reverse fluctuation patterns between different users in the same adjacency set and obtain suspected user pairs; Perform deep correlation analysis on the random component data of each group of suspected user pairs to obtain the correlation coefficient and time-lagged correlation coefficient, and comprehensively obtain the electricity consumption anti-correlation degree; The electricity usage feature set is extracted from the power data of each group of suspected user pairs, and combined with the electricity usage anti-correlation, the electricity theft users are identified from the suspected user pairs.

[0006] Furthermore, the power data includes power consumption data for the day, the power consumption data includes power consumption and power factor, and the power consumption data is acquired according to a preset sampling frequency; The method for constructing a user space association network includes: The electrical connection relationship between all distribution unit nodes and user nodes is obtained from the power grid topology to form a basic network framework; the user location information of each user is sequentially mapped to the corresponding user node in the basic network framework to build a user space association network; The steps for determining the adjacency set for each user include: Step S101: randomly select a user who is not marked as a selected user and mark it as the current user; Step S102: Mark the user node of the current user as the current node, obtain the user nodes adjacent to the current node from the user space association network, and mark them as spatial adjacent nodes; Step S103: Mark the power distribution unit node connected to the current node in the user space association network as the current unit, and mark all user nodes connected to the current unit as electrical adjacent nodes; Step S104: construct an adjacency relationship set corresponding to the current user based on all spatially adjacent nodes and electrically adjacent nodes, and mark the current user as a selected user; Step S105: looping steps S101 to S104 until all users are marked as selected users, and then the loop ends, and the adjacency relationship set of each user is determined.

[0007] Furthermore, the method for constructing the electricity consumption benchmark model includes: Historical data refers to electricity consumption data collected at historical moments. All historical data of each user are taken as a set of data sets, and the data sets correspond to users one by one. According to the sampling frequency, the time of the day is divided into time periods to obtain Periods, is an integer greater than 1; obtain the collection time corresponding to each set of historical data, and according to the collection time, take the historical data corresponding to the same date in each set of data as a set of same-day sets, and the same-day sets correspond to the dates one by one; according to the collection time, obtain the time period corresponding to each set of historical data; according to the same-day sets corresponding to each set of data and the time period corresponding to each set of historical data, construct the electricity consumption benchmark model corresponding to each user; wherein, the electricity consumption benchmark model is A matrix, each element represents a set of historical data, is the number of sets on the same day; each row in the matrix represents a date, and the dates corresponding to each row are arranged in ascending order of time, and each column represents a time period; Methods for extracting trend component data from the electricity consumption benchmark model include: The cycle of the electricity consumption benchmark model is set to one week, and the electricity consumption at the same cycle position in the electricity consumption benchmark model is taken as a group of same-week sets, and the same-week sets correspond to the cycle positions one-to-one; the electricity consumption in the corresponding time period in each group of same-week sets is taken as a group of simultaneous sets, and the simultaneous sets correspond to the time period one-to-one; the electricity consumption in each group of simultaneous sets is averaged to obtain the trend component data corresponding to each group of simultaneous sets.

[0008] Furthermore, the method of extracting periodic component data from the electricity consumption benchmark model includes: Obtain seasonal division standards. Based on the seasonal division standards, divide all months of the year into four seasons, each consisting of three months. Obtain the season corresponding to each electricity consumption in the electricity consumption benchmark model based on the collection time corresponding to each set of historical data. Treat the electricity consumption corresponding to the same season in the electricity consumption benchmark model as a set of seasonal sets, with the seasonal sets corresponding to the seasons one-to-one. Treat the electricity consumption corresponding to the same time period in each set of seasonal sets as a set of seasonal time sets, with the seasonal time sets corresponding to the time periods one-to-one. Perform mean processing on the electricity consumption in each set of seasonal time sets to obtain the seasonal component data corresponding to each set of seasonal time sets. Monday to Friday are defined as working days, and Saturday and Sunday are defined as rest days. Based on the collection time corresponding to each set of historical data, each electricity consumption in the electricity consumption benchmark model is divided into work and rest days, that is, divided into work days or rest days. Based on the division results, all electricity consumption in the electricity consumption benchmark model is divided into two groups of work and rest sets. The electricity consumption with the same corresponding time period in each group of work and rest sets is regarded as a group of working hour sets, and the working hour sets correspond to the time periods one by one. The electricity consumption in each group of working hour sets is averaged to obtain the work and rest component data corresponding to each group of working hour sets. All seasonal data and work-rest component data are regarded as periodic component data.

[0009] Furthermore, the method for extracting random component data includes: Based on the electricity data of each user, a corresponding real-time electricity consumption model is constructed, and the real-time electricity consumption model corresponds to the user one-to-one; the method for constructing the real-time electricity consumption model is consistent with the method for constructing the electricity consumption benchmark model; each electricity consumption in each real-time electricity consumption model is subtracted from the corresponding trend component data, and then from the corresponding periodic component data to obtain a random quantity; all random quantities corresponding to each real-time electricity consumption model are used as the random component data corresponding to each user.

[0010] Furthermore, the method of identifying the reverse fluctuation pattern includes: Define a sliding window with a sliding window size of Periods, is an integer greater than 0; the sliding window starts from the first data in each group of random component data and slides backward step by step, sliding one period each time until the sliding window covers the last data in the corresponding random component data, and the sliding ends; wherein, after each sliding, the continuous The random amount of each time period is combined into a set of window segments, where the first data is the earliest random amount in the random component data for the corresponding time period, and the last data is the latest random amount in the random component data for the corresponding time period; each user is combined with each user in the corresponding adjacency relationship set as a user pair, and the window segments corresponding to each user pair are combined into a set of window sets in turn, with a one-to-one correspondence between window sets and user pairs; The average of the starting time point and the corresponding ending time point corresponding to each time period is used as the time period average of each time period; the time period average of the random quantity corresponding to each window segment is averaged to obtain the window average corresponding to each window segment; a time lag threshold is preset, and the two groups of window segments corresponding to the difference between the window means in each group of window sets is less than the time lag threshold, and are respectively used as a group of window time sets; wherein, the two groups of window segments in each group of window time sets correspond to different users; if the random quantities in one group of window segments are all positive numbers, and the random quantities in the other group of window segments are all negative numbers, then it is determined that there is a reverse fluctuation pattern in the corresponding window time set; if the random quantities in one group of window segments are not all positive numbers or not all negative numbers, then it is determined that there is no reverse fluctuation pattern in the corresponding window time set; if there is a reverse fluctuation pattern in the window time set corresponding to the user pair, then it is determined that there is a reverse fluctuation pattern in the corresponding user pair; if there is no reverse fluctuation pattern in the window time set corresponding to the user pair, then it is determined that there is no reverse fluctuation pattern in the corresponding user pair; The method for obtaining the suspected user pair is as follows: if there is a reverse fluctuation pattern in the user pair, the corresponding user pair is marked as a suspected user pair; if there is no reverse fluctuation pattern in the user pair, the corresponding user pair is not marked.

[0011] Furthermore, the method of obtaining the correlation coefficient includes: The window time sets with reverse fluctuation patterns in each group of suspected user pairs are marked as reverse sets, and the reverse sets with the same window means corresponding to two groups of window segments are marked as real-time sets. The Pearson correlation coefficients between the two groups of window segments corresponding to each real-time set are calculated in sequence and marked as segment coefficients. The number of real-time sets corresponding to each group of suspected user pairs is counted and marked as the real-time number. The segment coefficients of each group of suspected user pairs are added in sequence and then divided by the corresponding real-time number to obtain the correlation coefficient of each group of suspected user pairs. Methods for obtaining the time-lagged correlation coefficient include: All reverse sets not marked as real-time sets are marked as time-delay sets. The segment coefficients between the two sets of window segments corresponding to each set of time-delay sets are calculated in sequence and marked as time-delay coefficients. The absolute values ​​of each time-delay coefficient are compared, and the time-delay coefficient with the largest absolute value is taken as the time-delay correlation coefficient. Methods for comprehensively obtaining the anti-correlation of electricity consumption include: Obtain the time lag set corresponding to the time lag correlation coefficient and mark it as the maximum set; mark the window segment with a later window mean in the maximum set as the first segment, and mark the window segment with an earlier window mean as the second segment; subtract the window mean of the second segment from the window mean of the first segment to obtain the time lag value; preset an attenuation parameter, multiply the opposite of the attenuation parameter by the absolute value of the time lag value to obtain a weight factor; use a natural constant as the base and the weight factor as the exponent to perform an exponential operation to obtain a weight coefficient; subtract the weight coefficient from 1 and multiply it by the time lag correlation coefficient to obtain a time lag weight; multiply the weight coefficient by the correlation coefficient to obtain a real-time weight; add the time lag weight to the real-time weight to obtain the anti-correlation degree of electricity consumption.

[0012] Furthermore, the method for extracting the electricity usage feature set includes: The random quantities in each reverse set are marked as abnormal quantities, and the users in each group of suspected user pairs are marked as suspected users; the abnormal quantities corresponding to each suspected user are averaged to obtain the abnormal degree of power fluctuation of each suspected user; the power factor corresponding to the same power data as the abnormal quantity in the power data corresponding to each suspected user is marked as the abnormal factor; the abnormal factors corresponding to each suspected user are averaged to obtain the average power factor of each suspected user; the abnormal degree of power fluctuation and the average power factor of each suspected user are used as the power feature set of each suspected user.

[0013] Furthermore, the method for identifying electricity theft users from suspected user pairs includes: The electricity usage feature set and the electricity usage anti-correlation corresponding to each group of suspected user pairs are used as a set of electricity theft identification data, and the electricity theft identification data is corresponding to the suspected user pairs one by one. Each set of electricity theft identification data is input into the trained electricity theft identification model, and the corresponding identification label is output. The identification label is a digital label corresponding to the identification result. Different identification results have different digital labels. The identification results include the presence of electricity theft and the absence of electricity theft. The corresponding identification results are obtained based on the identification labels, and the suspected user pairs identified as having electricity theft are marked as electricity theft user pairs. In each pair of electricity theft users, the user corresponding to the first segment is marked as the first user, and the user corresponding to the second segment is marked as the second user; different digital labels are set for the first user and the second user, and marked as user labels; the electricity usage feature set and user label corresponding to each first user are used as a set of first data, and the first data correspond to the first user one-to-one; the electricity usage feature set and user label corresponding to each second user are used as a set of second data, and the second data correspond to the second user one-to-one; the first data and second data corresponding to each pair of electricity theft users are used as a set of electricity theft analysis data, and the electricity theft analysis data correspond to the electricity theft user pairs one-to-one; each set of electricity theft analysis data is input into a trained user identification model, the corresponding user label is output, and the user corresponding to the user label is regarded as the electricity theft user; wherein both the electricity theft identification model and the user identification model are deep neural network models.

[0014] A system for processing electric power monitoring data, implementing the method for processing electric power monitoring data, comprises: Data collection module, used to collect power data from different users in residential areas; The spatial association module is used to obtain user location information and power grid topology, build a user spatial association network, and determine the adjacency set of each user; The model building module is used to obtain the historical data of each user and build a corresponding electricity consumption benchmark model, and extract trend component data and periodic component data from each electricity consumption benchmark model; The time decomposition module is used to decompose the time series of each user's power data based on the trend component data and the periodic component data, and extract the random component data; The preliminary identification module is used to apply sliding window analysis based on random component data to identify the reverse fluctuation patterns between different users in the same adjacency set and obtain suspected user pairs; The correlation calculation module is used to perform in-depth correlation analysis on the random component data of each group of suspected user pairs, calculate the correlation coefficient and time-lagged correlation coefficient, and comprehensively calculate the anti-correlation degree of electricity consumption; The electricity theft identification module is used to extract the electricity usage feature set from the power data of each group of suspected user pairs, and identify the electricity theft users from the suspected user pairs based on the electricity usage anti-correlation.

[0015] The technical effects and advantages of the method and system for processing electric energy monitoring data of the present invention are as follows: By constructing a user spatial association network and comprehensively leveraging users' spatial and electrical proximity relationships, the system can more comprehensively characterize the connections between users. By extracting trend and periodic component data, it can accurately identify the inherent trend and periodic characteristics of electricity usage behavior, reflecting changes in users' electricity usage habits and patterns. Using time series decomposition, random component data is extracted from electricity usage data, helping to identify complex patterns inherent in abnormal electricity usage. Based on sliding window analysis and correlation calculation, it can accurately identify inverse fluctuation patterns between different users within the same adjacency set and precisely calculate the anti-correlation of electricity usage, enabling preliminary screening of electricity theft users. Deep learning technology, combined with the extracted electricity usage feature set, can identify electricity theft and identify users, achieving intelligent and precise detection of electricity theft. This system accurately identifies electricity theft in residential areas, significantly improving the accuracy and timeliness of theft detection and enhancing the system's adaptability and generalization capabilities for complex electricity usage scenarios. This not only reduces the operation and maintenance costs of traditional manual identification but also effectively improves the security, stability, and resource utilization efficiency of the distribution network. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a schematic diagram of a system for processing electric energy monitoring data according to embodiment 1 of the present invention; Figure 2 This is a flow chart of a method for processing electric energy monitoring data according to embodiment 2 of the present invention. DETAILED DESCRIPTION

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0018] Example 1 See also Figure 1 As shown, this embodiment provides a system for processing electric energy monitoring data, including a data acquisition module, a spatial correlation module, a model building module, a time decomposition module, a preliminary identification module, a correlation calculation module, and an electricity theft identification module; each module is connected by wired and / or wireless means to realize data transmission between modules.

[0019] The data acquisition module is used to collect power data from different users in residential areas.

[0020] Electricity data refers to electric energy monitoring data. Electricity energy monitoring data refers to various types of data collected within the power system and is used to detect and analyze electricity usage. Electricity data includes electricity consumption data for the day, which includes electricity consumption and power factor. Electricity consumption is the amount of electricity consumed by the user within the preset sampling frequency, and the power factor is the ratio of active power to apparent power, which is used to reflect electricity efficiency. Electricity consumption data is obtained through the smart meter installed at the main incoming line entrance of each user according to the sampling frequency. The sampling frequency is pre-set by technical personnel in this field according to actual conditions, such as 15 minutes, 30 minutes, 1 hour, etc.

[0021] The spatial association module is used to obtain user location information and power grid topology, build a user spatial association network, and determine the adjacency set of each user.

[0022] User location information is the user's geographical location in physical space, that is, the user's specific address, including community name, unit number, floor number, door number, etc.; user location information is obtained through the user information management system within the power company; the power grid topology is the connection relationship between each node (such as transformer node, distribution unit node, user node) in the power system within the residential area; the power grid topology is obtained through the distribution network management system within the power company.

[0023] Methods for building a user space association network include: The electrical connection relationships between all distribution unit nodes and user nodes are obtained from the power grid topology to form a basic network framework to reflect the power supply path. The user location information of each user is sequentially mapped to the corresponding user node in the basic network framework, and spatial attributes are added to each user node to construct a user space association network. The steps for determining the adjacency set for each user include: Step S101: randomly select a user who is not marked as a selected user and mark it as the current user; Step S102: Mark the user node of the current user as the current node, obtain user nodes adjacent to the current node from the user spatial association network, and mark them as spatially adjacent nodes. In this embodiment, spatially adjacent nodes refer to user nodes located on the same floor of the same unit building as the current node (i.e., horizontally adjacent), or user nodes located on adjacent floors (i.e., above or below) of the current node in the same unit building (i.e., vertically adjacent). Step S103: Mark the power distribution unit node connected to the current node in the user space association network as the current unit, and mark all user nodes connected to the current unit as electrical adjacent nodes; Step S104: construct an adjacency relationship set corresponding to the current user based on all spatially adjacent nodes and electrically adjacent nodes, and mark the current user as a selected user; Step S105: looping steps S101 to S104 until all users are marked as selected users, and then the loop ends, and the adjacency relationship set of each user is determined.

[0024] It should be noted that the reason why the adjacency relationship set includes both spatial adjacency and electrical adjacency is that by combining the relationships in the two dimensions of spatial adjacency and electrical adjacency, the adjacency relationship set can more comprehensively characterize the correlation between users. In electricity theft identification, spatially adjacent and electrically adjacent users may have correlations in their electricity usage patterns. This dual correlation helps to more accurately identify electricity theft.

[0025] The model building module is used to obtain the historical data of each user and build a corresponding electricity consumption benchmark model, and extract trend component data and periodic component data from each electricity consumption benchmark model.

[0026] Methods for building a baseline model for electricity consumption include: Historical data refers to electricity consumption data collected at historical moments. All historical data of each user are taken as a set of data sets, and the data sets correspond to users one by one. According to the sampling frequency, the time of the day is divided into time periods to obtain Periods, is an integer greater than 1; for example, if the sampling frequency is 8 hours, the time of the day is divided into 3 periods, period 1 is 00:00-08:00, period 2 is 08:00-16:00, and period 3 is 16:00-24:00; according to the built-in clock of the smart meter, the collection time corresponding to each set of historical data is obtained; according to the collection time, the historical data corresponding to the same date in each set of data is taken as a set of same-day sets, and the same-day sets correspond to the dates one by one; according to the collection time, the period corresponding to each set of historical data is obtained; according to the same-day sets corresponding to each set of data and the period corresponding to each set of historical data, the electricity consumption benchmark model corresponding to each user is constructed; wherein the electricity consumption benchmark model is A matrix, each element represents a set of historical data, is the number of sets on the same day; each row in the matrix represents a date, and the dates corresponding to each row are arranged in ascending time order, that is, from the earliest date to the latest date, to ensure the continuity of the time series, and each column represents a time period.

[0027] For example, the data set includes 2 same-day sets, each of which includes 3 sets of historical data; wherein, same-day set 1 includes 、 and , Same Day Collection 2 includes 、 and , and Corresponding to period 1, and Corresponding to period 2, and For period 3, the same-day collection 1 corresponds to 2025-01-01, and the same-day collection 2 corresponds to 2025-01-02. Therefore, the corresponding electricity consumption benchmark model is: .

[0028] Methods for extracting trend component data from the electricity consumption benchmark model include: The electricity consumption benchmark model is set to have a cycle of one week, and the electricity consumption at the same cycle position in the electricity consumption benchmark model is taken as a group of same-week sets, and the same-week sets correspond to the cycle positions (such as Monday, Tuesday, Wednesday, etc.) one-to-one. For example, the electricity consumption benchmark model includes the electricity consumption of Monday of the first week, Tuesday of the first week, Wednesday of the first week, Monday of the second week, Tuesday of the second week, and Thursday of the second week, respectively. Then the electricity consumption of Monday of the first week and Monday of the second week is taken as a group of same-week sets, the electricity consumption of Tuesday of the first week and Tuesday of the second week is taken as a group of same-week sets, the electricity consumption of Wednesday of the first week is taken as a group of same-week sets, and the electricity consumption of Thursday of the second week is taken as a group of same-week sets. The electricity consumption in the same time period in each group of same-week sets is taken as a group of simultaneous sets, and the simultaneous sets correspond to the time period one-to-one. The amount of electricity consumption in each group of simultaneous sets is counted and marked as a first amount. The electricity consumption in each group of simultaneous sets is sequentially added and then divided by the first amount to obtain the trend component data corresponding to each group of simultaneous sets, that is, the electricity consumption trend of each time period of each day in a week is obtained.

[0029] Methods for extracting cycle component data from the electricity consumption benchmark model include: Those skilled in the art may obtain the seasonal division standard through the official website of the National Meteorological Department, meteorological professional textbooks, meteorological academic journals, technical specifications issued by meteorological service agencies, etc.; according to the seasonal division standard, all months of the year are divided into four seasons, and each season consists of three months with similar climatic characteristics; in this embodiment, spring preferably corresponds to March to May, summer corresponds to June to August, autumn corresponds to September to November, and winter corresponds to December to February; according to the collection time corresponding to each set of historical data, the season corresponding to each electricity consumption in the electricity consumption benchmark model is obtained; the electricity consumption corresponding to the same season in the electricity consumption benchmark model is taken as a group of seasonal sets, and the seasonal sets correspond to the seasons one-to-one; the electricity consumption corresponding to the same time period in each group of seasonal sets is taken as a group of seasonal time sets, and the seasonal time sets correspond to the time periods one-to-one; the amount of electricity consumption in each group of seasonal time sets is counted and marked as a second amount; the electricity consumption in each group of seasonal time sets is sequentially added, and then divided by the second amount to obtain the seasonal component data corresponding to each group of seasonal time sets, that is, the electricity consumption pattern of each time period in a season is obtained; Monday to Friday are defined as working days, and Saturday and Sunday are defined as rest days; according to the collection time corresponding to each set of historical data, each electricity consumption in the electricity consumption benchmark model is divided into work and rest, that is, the division into working days or rest days; according to the division result, all electricity consumption in the electricity consumption benchmark model is divided into two groups of work and rest sets, one of which corresponds to the electricity consumption of working days, and the other corresponds to the electricity consumption of rest days; the electricity consumption with the same corresponding time period in each group of work and rest sets is regarded as a group of working hour sets, and the working hour sets correspond to the time periods one by one; the amount of electricity consumption in each group of working hour sets is counted and marked as the third amount; the electricity consumption in each group of working hour sets is added in sequence, and then divided by the third amount to obtain the work and rest component data corresponding to each group of working hour sets, that is, to obtain the electricity consumption pattern of each time period on working days or rest days; All seasonal data and work-rest component data are regarded as periodic component data.

[0030] The time decomposition module is used to perform time series decomposition on the power data of each user based on trend component data and periodic component data, and extract random component data.

[0031] Methods for extracting random component data include: Based on the electricity data of each user, a corresponding real-time electricity consumption model is constructed, and the real-time electricity consumption model corresponds to the user one-to-one; the method for constructing the real-time electricity consumption model is consistent with the method for constructing the electricity consumption benchmark model; each electricity consumption in each real-time electricity consumption model is subtracted from the corresponding trend component data, and then from the corresponding periodic component data to obtain a random quantity; all random quantities corresponding to each real-time electricity consumption model are used as the random component data corresponding to each user.

[0032] The preliminary identification module is used to apply sliding window analysis based on random component data to identify the reverse fluctuation patterns between different users in the same adjacency set and obtain suspected user pairs.

[0033] Methods for identifying counter-swing patterns include: Define a sliding window with a sliding window size of Periods, is an integer greater than 0; the sliding window starts from the first data in each group of random component data and slides backward step by step, sliding one period each time until the sliding window covers the last data in the corresponding random component data, and the sliding ends; wherein, after each sliding, the continuous The random amount of each time period is combined to form a set of window segments. The first data is the earliest random amount in the random component data for the corresponding time period, and the last data is the latest random amount in the random component data for the corresponding time period. Each user is combined with each user in the corresponding adjacency relationship set as a user pair, and the window segments corresponding to each user pair are combined as a set of window sets. The window sets correspond to the user pairs one by one. It should be noted that a set of user pairs includes two users. The mean of the start time point and the end time point corresponding to each time period is taken as the time period mean of each time period. For example, if the time period is 00:00-08:00, where 00:00 is the start time point and 08:00 is the end time point, the time mean is 04:00. The time period means of the random amount corresponding to each window segment are added in sequence and then divided by , obtain the window mean corresponding to each window fragment; preset a time lag threshold, which is preset by a person skilled in the art according to actual conditions; in each group of window sets, the difference between the window means is less than the time lag threshold, and the corresponding two groups of window fragments are respectively regarded as a group of window time sets; wherein, the two groups of window fragments in each group of window time sets correspond to different users; if the random quantities in one group of window fragments are all positive numbers, and the random quantities in the other group of window fragments are all negative numbers, then it is determined that there is a reverse fluctuation pattern in the corresponding window time set; if the random quantities in one group of window fragments are not all positive numbers or not all negative numbers, then it is determined that there is no reverse fluctuation pattern in the corresponding window time set; if there is a reverse fluctuation pattern in the window time set corresponding to the user pair, then it is determined that there is a reverse fluctuation pattern in the corresponding user pair; if there is no reverse fluctuation pattern in the window time set corresponding to the user pair, then it is determined that there is no reverse fluctuation pattern in the corresponding user pair.

[0034] The method for obtaining the suspected user pair is as follows: if there is a reverse fluctuation pattern in the user pair, the corresponding user pair is marked as a suspected user pair; if there is no reverse fluctuation pattern in the user pair, the corresponding user pair is not marked.

[0035] The correlation calculation module is used to perform in-depth correlation analysis on the random component data of each group of suspected user pairs, calculate the correlation coefficient and time-lag correlation coefficient, and comprehensively calculate the anti-correlation degree of electricity consumption.

[0036] Methods for calculating correlation coefficients include: The window time sets with reverse fluctuation patterns in each group of suspected user pairs are marked as reverse sets, and the reverse sets with the same window means corresponding to the two groups of window segments are marked as real-time sets; the Pearson correlation coefficients between the two groups of window segments corresponding to each group of real-time sets are calculated in turn and marked as segment coefficients; the number of real-time sets corresponding to each group of suspected user pairs is counted and marked as real-time number; the segment coefficients of each group of suspected user pairs are added in turn, and then divided by the corresponding real-time number to obtain the correlation coefficient of each group of suspected user pairs; it should be noted that the calculation method of the Pearson correlation coefficient is an existing technology, and the specific calculation process will not be described in detail here.

[0037] Methods for calculating the time-lagged correlation coefficient include: All reverse sets that are not marked as real-time sets are marked as time-lag sets. The segment coefficients between the two groups of window segments corresponding to each group of time-lag sets are calculated in turn and marked as time-lag coefficients. The absolute values ​​of each time-lag coefficient are compared respectively, and the time-lag coefficient with the largest absolute value is taken as the time-lag correlation coefficient.

[0038] The methods for comprehensively calculating the anti-correlation of electricity consumption include: Obtain a time lag set corresponding to the time lag correlation coefficient and mark it as the maximum set; mark the window segment with a later window mean in the maximum set as the first segment, and mark the window segment with an earlier window mean as the second segment; subtract the window mean of the second segment from the window mean of the first segment to obtain a time lag value; preset an attenuation parameter, and the attenuation parameter is pre-set by technical personnel in this field according to actual conditions; multiply the opposite of the attenuation parameter by the absolute value of the time lag value to obtain a weight factor; use a natural constant as the base and the weight factor as the exponent to perform an exponential operation to obtain a weight coefficient; subtract the weight coefficient from 1 and multiply it by the time lag correlation coefficient to obtain a time lag weight; multiply the weight coefficient by the correlation coefficient to obtain a real-time weight; add the time lag weight to the real-time weight to obtain the anti-correlation degree of electricity consumption.

[0039] The electricity theft identification module is used to extract the electricity usage feature set from the power data of each group of suspected user pairs, and identify the electricity theft users from the suspected user pairs based on the electricity usage anti-correlation.

[0040] Methods for extracting electricity usage feature sets include: The random quantities in each reverse set are marked as abnormal quantities, and the users in each group of suspected user pairs are marked as suspected users; the number of abnormal quantities corresponding to each suspected user is counted and marked as the abnormal number; the abnormal quantities corresponding to each suspected user are added up in sequence, and then divided by the abnormal number to obtain the abnormal degree of power fluctuation of each suspected user; in the power data corresponding to each suspected user, the power factor corresponding to the same power data as the abnormal quantity is marked as the abnormal factor; the abnormal factors corresponding to each suspected user are added up in sequence, and then divided by the abnormal number to obtain the average power factor of each suspected user; the abnormal degree of power fluctuation and the average power factor of each suspected user are used as the power feature set of each suspected user.

[0041] Methods for identifying electricity theft users from suspected user pairs include: The electricity usage feature set and the electricity usage anti-correlation corresponding to each group of suspected user pairs are used as a set of electricity theft identification data, and the electricity theft identification data is corresponding to the suspected user pairs one by one. Each set of electricity theft identification data is input into the trained electricity theft identification model, and the corresponding identification label is output. The identification label is a digital label corresponding to the identification result. Different identification results have different digital labels. The identification results include the presence of electricity theft and the absence of electricity theft. The corresponding identification results are obtained based on the identification labels, and the suspected user pairs identified as having electricity theft are marked as electricity theft user pairs. In each pair of electricity theft users, the user corresponding to the first segment is marked as the first user, and the user corresponding to the second segment is marked as the second user; different digital labels are set for the first user and the second user, and marked as user labels; the electricity usage feature set and user label corresponding to each first user are used as a set of first data, and the first data correspond to the first user one-to-one; the electricity usage feature set and user label corresponding to each second user are used as a set of second data, and the second data correspond to the second user one-to-one; the first data and second data corresponding to each pair of electricity theft users are used as a set of electricity theft analysis data, and the electricity theft analysis data correspond to the electricity theft user pair one-to-one; each set of electricity theft analysis data is input into the trained user identification model, the corresponding user label is output, and the user corresponding to the user label is regarded as the electricity theft user; wherein, the electricity theft identification model and the user identification model are both deep neural network models, and the deep neural network model is a prior art, and the specific training process is not described in detail here.

[0042] This embodiment constructs a user spatial association network and comprehensively utilizes users' spatial and electrical proximity relationships to more comprehensively characterize the associations between users. By extracting trend and periodic component data, it accurately identifies the inherent trend and periodic characteristics of electricity usage behavior, reflecting changes in users' electricity usage habits and patterns. It uses time series decomposition to extract random component data from electricity usage data, helping to identify complex patterns inherent in abnormal electricity usage. Sliding window analysis and correlation calculations accurately identify inverse fluctuation patterns between different users within the same adjacency set and precisely calculate the anti-correlation of electricity usage, enabling preliminary screening of electricity theft users. Deep learning technology, combined with the extracted electricity usage feature set, enables identification of electricity theft and user identification, enabling intelligent and precise detection of electricity theft. This system accurately identifies electricity theft in residential areas, significantly improving the accuracy and timeliness of detection and enhancing the system's adaptability and generalization capabilities for complex electricity usage scenarios. This not only reduces the operational and maintenance costs of traditional manual identification but also effectively improves the security, stability, and resource utilization efficiency of the distribution network.

[0043] Example 2 See also Figure 2 As shown, for the parts not described in detail in this embodiment, please refer to the description of Example 1. A method for processing electric energy monitoring data is provided, and the method includes: Collect electricity data from different users in residential areas; Obtain user location information and grid topology, build a user spatial association network, and determine the adjacency set of each user; Obtain historical data for each user and build a corresponding electricity usage benchmark model. Extract trend component data and periodic component data from each electricity usage benchmark model. Based on the trend component data and periodic component data, the time series of each user's electricity data is decomposed to extract the random component data; Based on random component data, sliding window analysis is applied to identify reverse fluctuation patterns between different users in the same adjacency set and obtain suspected user pairs; Perform deep correlation analysis on the random component data of each group of suspected user pairs, calculate the correlation coefficient and time-lagged correlation coefficient, and comprehensively calculate the anti-correlation of electricity consumption; The electricity usage feature set is extracted from the power data of each group of suspected user pairs, and combined with the electricity usage anti-correlation, the electricity theft users are identified from the suspected user pairs.

[0044] Example 3 The present application also provides an electronic device. The electronic device may include one or more processors and one or more memories. The memories may store computer-readable code that, when executed by the one or more processors, may execute the above-described method for processing electric energy monitoring data.

[0045] The method or system according to the embodiment of the present application can also be implemented with the help of the architecture of the electronic device shown in this application. The electronic device may include a bus, one or more CPUs, ROM, RAM, a communication port connected to a network, input / output, a hard disk, etc. The storage device in the electronic device, such as a ROM or a hard disk, can store a method for processing electric energy monitoring data provided in this application. Furthermore, the electronic device may also include a user interface. Of course, the architecture shown in this application is only exemplary. When implementing different devices, one or more components in the electronic device shown in this application can be omitted according to actual needs.

[0046] Example 4 One embodiment of the present application discloses a computer-readable storage medium. The computer-readable storage medium stores computer-readable instructions. When the computer-readable instructions are executed by a processor, a method for processing electric energy monitoring data according to an embodiment of the present application described with reference to the above figures can be executed. The storage medium includes, but is not limited to, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory. Non-volatile memory may include, for example, read-only memory (ROM), a hard disk, a flash memory, etc.

[0047] In addition, according to embodiments of the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the present application provides a non-transitory machine-readable storage medium storing machine-readable instructions capable of being executed by a processor to perform instructions corresponding to the steps of the method provided herein, such as a method for processing electric power monitoring data. When the computer program is executed by a central processing unit (CPU), the functions defined in the method of the present application are performed.

[0048] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art will be able to modify the technical solutions described in the foregoing embodiments or to substitute equivalents for some of the technical features. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

[0049] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or apparatus comprising the element.

[0050] In the description of the present invention, it should be understood that the terms "first", "second", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0051] In the description of the present invention, unless otherwise specified, "plurality" means two or more.

[0052] In the description of the present invention, “several” means one or more, and “a large number” means two or more.

[0053] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0054] The formulas in this manual are all dimensionless and calculated using numerical values. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters and thresholds in the formulas are set by technicians in this field based on actual conditions.

[0055] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.

Claims

1. A method for processing electric energy monitoring data, characterized in that: include: Collect electricity data from different users in residential areas; Obtain user location information and grid topology, build a user spatial association network, and determine the adjacency set of each user; Obtain historical data for each user and build a corresponding electricity usage benchmark model. Extract trend component data and periodic component data from each electricity usage benchmark model. Based on the trend component data and periodic component data, the time series of each user's electricity data is decomposed to extract the random component data; Based on random component data, sliding window analysis is applied to identify reverse fluctuation patterns between different users in the same adjacency set and obtain suspected user pairs; Perform deep correlation analysis on the random component data of each group of suspected user pairs to obtain the correlation coefficient and time-lagged correlation coefficient, and comprehensively obtain the electricity consumption anti-correlation degree; The electricity usage feature set is extracted from the power data of each group of suspected user pairs, and combined with the electricity usage anti-correlation, the electricity theft users are identified from the suspected user pairs.

2. The method for processing electric power monitoring data according to claim 1, characterized in that: The power data includes power consumption data for the day, including power consumption and power factor, and the power consumption data is acquired according to a preset sampling frequency; The method for constructing a user space association network includes: The electrical connection relationship between all distribution unit nodes and user nodes is obtained from the power grid topology to form a basic network framework; the user location information of each user is sequentially mapped to the corresponding user node in the basic network framework to build a user space association network; The steps for determining the adjacency set for each user include: Step S101: randomly select a user who is not marked as a selected user and mark it as the current user; Step S102: Mark the user node of the current user as the current node, obtain the user nodes adjacent to the current node from the user space association network, and mark them as spatial adjacent nodes; Step S103: Mark the power distribution unit node connected to the current node in the user space association network as the current unit, and mark all user nodes connected to the current unit as electrical adjacent nodes; Step S104: construct an adjacency relationship set corresponding to the current user based on all spatially adjacent nodes and electrically adjacent nodes, and mark the current user as a selected user; Step S105: looping steps S101 to S104 until all users are marked as selected users, and then the loop ends, and the adjacency relationship set of each user is determined.

3. The method for processing electric power monitoring data according to claim 2, characterized in that: Methods for building a baseline model for electricity consumption include: Historical data refers to electricity consumption data collected at historical moments. All historical data of each user are taken as a set of data sets, and the data sets correspond to users one by one. According to the sampling frequency, the time of the day is divided into time periods to obtain Periods, is an integer greater than 1; obtain the collection time corresponding to each set of historical data, and according to the collection time, take the historical data corresponding to the same date in each set of data as a set of same-day sets, and the same-day sets correspond to the dates one by one; according to the collection time, obtain the time period corresponding to each set of historical data; according to the same-day sets corresponding to each set of data and the time period corresponding to each set of historical data, construct the electricity consumption benchmark model corresponding to each user; wherein, the electricity consumption benchmark model is A matrix, each element represents a set of historical data, is the number of sets on the same day; each row in the matrix represents a date, and the dates corresponding to each row are arranged in ascending order of time, and each column represents a time period; Methods for extracting trend component data from the electricity consumption benchmark model include: The cycle of the electricity consumption benchmark model is set to one week, and the electricity consumption at the same cycle position in the electricity consumption benchmark model is taken as a group of same-week sets, and the same-week sets correspond to the cycle positions one-to-one; the electricity consumption in the corresponding time period in each group of same-week sets is taken as a group of simultaneous sets, and the simultaneous sets correspond to the time period one-to-one; the electricity consumption in each group of simultaneous sets is averaged to obtain the trend component data corresponding to each group of simultaneous sets.

4. The method for processing electric power monitoring data according to claim 3, characterized in that: Methods for extracting cycle component data from the electricity consumption benchmark model include: Obtain the seasonal division standard, and divide all months of the year into four seasons based on the seasonal division standard, with each season consisting of three months; obtain the season corresponding to each electricity consumption in the electricity consumption benchmark model based on the collection time corresponding to each set of historical data; treat the electricity consumption corresponding to the same season in the electricity consumption benchmark model as a set of seasonal sets, with the seasonal sets corresponding to the seasons one-to-one; treat the electricity consumption corresponding to the same time period in each set of seasonal sets as a set of seasonal time sets, with the seasonal time sets corresponding to the time periods one-to-one; count the amount of electricity consumption in each set of seasonal time sets and mark it as the second amount; perform mean processing on the electricity consumption in each set of seasonal time sets to obtain the seasonal component data corresponding to each set of seasonal time sets; Monday to Friday are defined as working days, and Saturday and Sunday are defined as rest days. Based on the collection time corresponding to each set of historical data, each electricity consumption in the electricity consumption benchmark model is divided into work and rest days, that is, divided into work days or rest days. Based on the division results, all electricity consumption in the electricity consumption benchmark model is divided into two groups of work and rest sets. The electricity consumption with the same corresponding time period in each group of work and rest sets is regarded as a group of working hour sets, and the working hour sets correspond to the time periods one by one. The electricity consumption in each group of working hour sets is averaged to obtain the work and rest component data corresponding to each group of working hour sets. All seasonal data and work-rest component data are regarded as periodic component data.

5. The method for processing electric power monitoring data according to claim 4, characterized in that: The method for extracting random component data includes: Based on the electricity data of each user, a corresponding real-time electricity consumption model is constructed, and the real-time electricity consumption model corresponds to the user one-to-one; the method for constructing the real-time electricity consumption model is consistent with the method for constructing the electricity consumption benchmark model; each electricity consumption in each real-time electricity consumption model is subtracted from the corresponding trend component data, and then from the corresponding periodic component data to obtain a random quantity; all random quantities corresponding to each real-time electricity consumption model are used as the random component data corresponding to each user.

6. The method for processing electric power monitoring data according to claim 5, characterized in that: Methods for identifying counter-swing patterns include: Define a sliding window with a sliding window size of Periods, is an integer greater than 0; the sliding window starts from the first data in each group of random component data and slides backward step by step, sliding one period each time until the sliding window covers the last data in the corresponding random component data, and the sliding ends; wherein, after each sliding, the continuous The random amount of each time period is combined into a set of window segments, where the first data is the earliest random amount in the random component data for the corresponding time period, and the last data is the latest random amount in the random component data for the corresponding time period; each user is combined with each user in the corresponding adjacency relationship set as a user pair, and the window segments corresponding to each user pair are combined into a set of window sets in turn, with a one-to-one correspondence between window sets and user pairs; The average of the starting time point and the corresponding ending time point corresponding to each time period is used as the time period average of each time period; the time period average of the random quantity corresponding to each window segment is averaged to obtain the window average corresponding to each window segment; a time lag threshold is preset, and the two groups of window segments corresponding to the difference between the window means in each group of window sets is less than the time lag threshold, and are respectively used as a group of window time sets; wherein, the two groups of window segments in each group of window time sets correspond to different users; if the random quantities in one group of window segments are all positive numbers, and the random quantities in the other group of window segments are all negative numbers, then it is determined that there is a reverse fluctuation pattern in the corresponding window time set; if the random quantities in one group of window segments are not all positive numbers or not all negative numbers, then it is determined that there is no reverse fluctuation pattern in the corresponding window time set; if there is a reverse fluctuation pattern in the window time set corresponding to the user pair, then it is determined that there is a reverse fluctuation pattern in the corresponding user pair; if there is no reverse fluctuation pattern in the window time set corresponding to the user pair, then it is determined that there is no reverse fluctuation pattern in the corresponding user pair; The method for obtaining the suspected user pair is as follows: if there is a reverse fluctuation pattern in the user pair, the corresponding user pair is marked as a suspected user pair; if there is no reverse fluctuation pattern in the user pair, the corresponding user pair is not marked.

7. The method for processing electric power monitoring data according to claim 6, characterized in that: Methods for obtaining correlation coefficients include: The window time sets with reverse fluctuation patterns in each group of suspected user pairs are marked as reverse sets, and the reverse sets with the same window means corresponding to two groups of window segments are marked as real-time sets. The Pearson correlation coefficients between the two groups of window segments corresponding to each real-time set are calculated in sequence and marked as segment coefficients. The number of real-time sets corresponding to each group of suspected user pairs is counted and marked as the real-time number. The segment coefficients of each group of suspected user pairs are added in sequence and then divided by the corresponding real-time number to obtain the correlation coefficient of each group of suspected user pairs. Methods for obtaining the time-lagged correlation coefficient include: All reverse sets not marked as real-time sets are marked as time-delay sets. The segment coefficients between the two sets of window segments corresponding to each set of time-delay sets are calculated in sequence and marked as time-delay coefficients. The absolute values ​​of each time-delay coefficient are compared, and the time-delay coefficient with the largest absolute value is taken as the time-delay correlation coefficient. Methods for comprehensively obtaining the anti-correlation of electricity consumption include: Obtain the time lag set corresponding to the time lag correlation coefficient and mark it as the maximum set; mark the window segment with a later window mean in the maximum set as the first segment, and mark the window segment with an earlier window mean as the second segment; subtract the window mean of the second segment from the window mean of the first segment to obtain the time lag value; preset an attenuation parameter, multiply the opposite of the attenuation parameter by the absolute value of the time lag value to obtain a weight factor; use a natural constant as the base and the weight factor as the exponent to perform an exponential operation to obtain a weight coefficient; subtract the weight coefficient from 1 and multiply it by the time lag correlation coefficient to obtain a time lag weight; multiply the weight coefficient by the correlation coefficient to obtain a real-time weight; add the time lag weight to the real-time weight to obtain the anti-correlation degree of electricity consumption.

8. The method for processing electric power monitoring data according to claim 7, characterized in that: Methods for extracting electricity usage feature sets include: The random quantities in each reverse set are marked as abnormal quantities, and the users in each group of suspected user pairs are marked as suspected users; the abnormal quantities corresponding to each suspected user are averaged to obtain the abnormal degree of power fluctuation of each suspected user; the power factor corresponding to the same power data as the abnormal quantity in the power data corresponding to each suspected user is marked as the abnormal factor; the abnormal factors corresponding to each suspected user are averaged to obtain the average power factor of each suspected user; the abnormal degree of power fluctuation and the average power factor of each suspected user are used as the power feature set of each suspected user.

9. The method for processing electric power monitoring data according to claim 8, characterized in that: The method for identifying electricity theft users from suspected user pairs includes: The electricity usage feature set and the electricity usage anti-correlation corresponding to each group of suspected user pairs are used as a set of electricity theft identification data, and the electricity theft identification data is corresponding to the suspected user pairs one by one. Each set of electricity theft identification data is input into the trained electricity theft identification model, and the corresponding identification label is output. The identification label is a digital label corresponding to the identification result. Different identification results have different digital labels. The identification results include the presence of electricity theft and the absence of electricity theft. The corresponding identification results are obtained based on the identification labels, and the suspected user pairs identified as having electricity theft are marked as electricity theft user pairs. In each pair of electricity theft users, the user corresponding to the first segment is marked as the first user, and the user corresponding to the second segment is marked as the second user; different digital labels are set for the first user and the second user, and marked as user labels; the electricity usage feature set and user label corresponding to each first user are used as a set of first data, and the first data correspond to the first user one-to-one; the electricity usage feature set and user label corresponding to each second user are used as a set of second data, and the second data correspond to the second user one-to-one; the first data and second data corresponding to each pair of electricity theft users are used as a set of electricity theft analysis data, and the electricity theft analysis data correspond to the electricity theft user pairs one-to-one; each set of electricity theft analysis data is input into a trained user identification model, the corresponding user label is output, and the user corresponding to the user label is regarded as the electricity theft user; wherein both the electricity theft identification model and the user identification model are deep neural network models.

10. A system for processing electric power monitoring data, implementing the method for processing electric power monitoring data according to any one of claims 1 to 9, characterized in that: include: Data collection module, used to collect power data from different users in residential areas; The spatial association module is used to obtain user location information and power grid topology, build a user spatial association network, and determine the adjacency set of each user; The model building module is used to obtain the historical data of each user and build a corresponding electricity consumption benchmark model, and extract trend component data and periodic component data from each electricity consumption benchmark model; The time decomposition module is used to decompose the time series of each user's power data based on the trend component data and the periodic component data, and extract the random component data; The preliminary identification module is used to apply sliding window analysis based on random component data to identify the reverse fluctuation patterns between different users in the same adjacency set and obtain suspected user pairs; The correlation calculation module is used to perform in-depth correlation analysis on the random component data of each group of suspected user pairs, calculate the correlation coefficient and time-lagged correlation coefficient, and comprehensively calculate the anti-correlation degree of electricity consumption; The electricity theft identification module is used to extract the electricity usage feature set from the power data of each group of suspected user pairs, and identify the electricity theft users from the suspected user pairs based on the electricity usage anti-correlation.

Images