Electricity stealing behavior identification method, device, equipment, medium and program product

Abstract

The invention discloses an electricity stealing behavior identification method and device, equipment, a medium and a program product, and the method comprises the steps: in an electricity stealing behavior identification process, carrying out the analysis through considering the electricity utilization characteristics, station information and historical electricity utilization data of a to-be-detected communication station and combining an electricity utilization characteristic cluster; since the electricity utilization characteristic cluster is formed by clustering the electricity utilization characteristics of the plurality of communication stations, the electricity utilization characteristic difference of different communication stations can be fully considered, and the electricity stealing behavior can be effectively and accurately identified according to the electricity utilization characteristics of the communication station to be detected.

Description

Technical Field

[0001] This invention relates to the field of power detection technology, and in particular to a method, device, electronic equipment, computer-readable storage medium, and computer program product for identifying electricity theft. Background Technology

[0002] With the development of communication networks and the continuous expansion of communication site construction, the power consumption of communication sites has increased significantly. Communication management units have an increasingly urgent need for energy conservation and cost reduction at communication sites. However, in daily life, there are instances of electricity theft at base stations and other communication sites for the purpose of illegally occupying electricity and saving on electricity bills, resulting in economic losses and safety hazards. Existing methods for identifying electricity theft typically do not consider the differences in power consumption characteristics among different communication sites, making it difficult for current technologies to effectively and accurately identify electricity theft at communication sites. Summary of the Invention

[0003] This invention provides a method, apparatus, device, medium, and program product for identifying electricity theft, in order to solve the technical problem that existing technologies are unable to effectively and accurately identify electricity theft at communication sites.

[0004] To address the aforementioned technical problems, a first aspect of this invention provides a method for identifying electricity theft, comprising: Based on the historical power consumption data of the communication site to be tested, the power consumption characteristics of the communication site to be tested are obtained; Based on the site information of the communication site to be detected, the electricity consumption characteristics, the historical electricity consumption data, and several preset electricity consumption characteristic clusters, the electricity theft behavior identification model is used to identify the electricity theft behavior of the communication site to be detected, and the electricity theft behavior identification result of the communication site to be detected is obtained. The electricity consumption feature cluster is formed by clustering the electricity consumption features of several communication sites; the electricity theft behavior identification model is composed of a gray prediction model and a random forest model.

[0005] As a preferred embodiment, the method specifically obtains the electricity consumption characteristics through the following steps: Based on the historical electricity consumption data of any communication station, the electricity consumption change characteristics for each preset time period are obtained; wherein, the electricity consumption change characteristics include the first-order differential characteristics of electricity consumption, the second-order differential characteristics of electricity consumption, and the electricity consumption fluctuation characteristics. Based on the historical electricity consumption data, obtain the daily electricity consumption of the computer room and the average daily electricity consumption of the computer room for any one of the communication sites within a preset statistical period; Based on the preset statistical period, the daily power consumption of the computer room, and the average daily power consumption of the computer room, the daily power consumption fluctuation characteristics of any one communication site are determined. Based on the electricity consumption change characteristics, the daily electricity consumption fluctuation characteristics, and the rated power of the air conditioner in the computer room of any one communication site, the electricity consumption characteristics of any one communication site are determined.

[0006] As a preferred embodiment, the step of obtaining the electricity consumption change characteristics for each preset time period based on the historical electricity consumption data of any communication station specifically includes: Obtain the electricity consumption sequence for each preset time period from the historical electricity consumption data; Based on the electricity consumption sequence of two adjacent preset time periods, determine the first-order differential feature of electricity consumption for each preset time period; Based on the first-order differential characteristics of electricity consumption in two adjacent preset time periods, determine the second-order differential characteristics of electricity consumption in each preset time period. Based on a preset first time interval and a preset second time interval, calculate the first tilt angle corresponding to the line connecting the electricity consumption sequence of each preset time period and the electricity consumption sequence that follows it and is separated by the first time interval, and the second tilt angle corresponding to the line connecting the electricity consumption sequence of each preset time period and the electricity consumption sequence that follows it and is separated by the second time interval; wherein, the first time interval is less than the second time interval. The power consumption fluctuation characteristics for each preset time period are determined based on the difference between the first tilt angle and the second tilt angle.

[0007] As a preferred embodiment, determining the daily power consumption fluctuation characteristics of any communication site based on the preset statistical period, the daily power consumption of the data center, and the average daily power consumption of the data center specifically includes: Calculate the standard deviation of the daily power consumption of the data center based on the preset statistical period, the daily power consumption of the data center, and the average daily power consumption of the data center; The coefficient of variation of daily power consumption is obtained by comparing the standard deviation with the average daily power consumption of the computer room. Based on the preset statistical period, the daily power consumption of the computer room, the average daily power consumption of the computer room, and the standard deviation, the daily power consumption kurtosis and daily power consumption skewness are calculated respectively. The daily electricity consumption fluctuation characteristics are determined based on the daily electricity consumption variation coefficient, the daily electricity consumption kurtosis, and the daily electricity consumption skewness.

[0008] As a preferred embodiment, the method specifically obtains the electricity consumption characteristic clusters through the following steps: The similarity of the electricity consumption change characteristics of each of the communication stations in each of the preset time periods is calculated to determine the feature similarity between the communication stations. Based on the feature similarity, the electricity consumption change features of each of the communication stations in each of the preset time periods are clustered to obtain several initial electricity consumption feature clusters. The mean value of several electricity consumption change features in each preset time period in the initial electricity consumption feature cluster is calculated to determine the station electricity consumption features in each preset time period in the initial electricity consumption feature cluster. Based on the electricity consumption characteristics of several stations in each of the initial electricity consumption characteristic clusters, several electricity consumption characteristic clusters are generated.

[0009] As a preferred embodiment, the step of identifying electricity theft behavior of the communication site under test based on the site information, electricity consumption characteristics, historical electricity consumption data, and several preset electricity consumption characteristic clusters, using an electricity theft behavior identification model, and obtaining the electricity theft behavior identification result of the communication site under test, specifically includes: Based on the gray prediction model, the similarity calculation is performed between the electricity consumption change characteristics of the communication site to be detected and the electricity consumption characteristics of each site in several electricity consumption feature clusters, so as to obtain the target electricity consumption feature cluster that matches the electricity consumption change characteristics from several electricity consumption feature clusters; Based on the site information, the electricity consumption change characteristics, and the target electricity consumption feature cluster, the gray prediction model is used to calculate the real-time electricity consumption jitter prediction value and the initial electricity theft behavior identification result of the communication site to be detected. When the initial electricity theft behavior identification result does not indicate the existence of electricity theft behavior, the historical electricity consumption data corresponding to the communication site to be detected, the target electricity consumption feature cluster, the real-time electricity consumption fluctuation prediction value, the daily electricity consumption fluctuation feature, and the rated power of the computer room air conditioner are input into the random forest model to identify electricity theft behavior and obtain the electricity theft behavior identification result of the communication site to be detected.

[0010] As a preferred embodiment, the method specifically obtains the real-time power consumption fluctuation prediction value through the following steps: Based on the site information, obtain the number of service terminals added to the communication site to be detected, the power consumption calculation data caused by computing load, and the site air conditioning temperature; Obtain the station's electricity consumption fluctuation characteristics for the time period corresponding to the current prediction time from the target electricity consumption characteristic cluster; Based on the gray prediction model, and according to the number of new service terminals, the electricity consumption calculation data, the site air conditioning temperature, and the site electricity consumption fluctuation characteristics, the real-time electricity consumption fluctuation prediction value is calculated using the following expression: ; in, a , b , c and d These are the preset first fitting coefficient, second fitting coefficient, third fitting coefficient, and fourth fitting coefficient, respectively; This indicates the increase in the number of service terminals; This represents the electricity consumption calculation data; This indicates the air conditioning temperature at the site; This indicates the fluctuation characteristics of the power consumption at the site; This represents the preset prediction residual; This represents the predicted real-time power consumption fluctuation value.

[0011] As a preferred embodiment, the method specifically obtains the initial electricity theft behavior identification result through the following steps: Based on the gray prediction model, the difference identifier of the first-order differential feature of the electricity consumption of the communication station to be detected in each preset time period is determined by multiplying the first-order differential feature of the electricity consumption of the communication station to be detected with the first-order differential feature of the electricity consumption of the station in the target electricity consumption feature cluster. The difference identifier of the second-order differential feature of the electricity consumption of the communication station to be detected in each preset time period is determined by multiplying the second-order differential feature of the electricity consumption of the station to be detected with the second-order differential feature of the electricity consumption of the station in the target electricity consumption feature cluster. Based on the first-order differential characteristic difference identifier of electricity consumption and the second-order differential characteristic difference identifier of electricity consumption, the total differential characteristic difference value of electricity consumption of the communication station to be detected is determined. The total difference value of the power consumption fluctuation feature of the communication station under test is determined based on the difference between the preset power consumption fluctuation feature threshold and the power consumption fluctuation feature of the communication station under test in each preset time period. When the total difference value of the differential characteristics of electricity consumption is less than the preset first differential characteristic difference threshold of electricity consumption, the total difference value of the fluctuation characteristics of electricity consumption is less than the preset fluctuation characteristic difference threshold of electricity consumption, and the predicted value of the real-time fluctuation of electricity consumption is less than the preset fluctuation threshold of electricity consumption, the initial electricity theft behavior identification result is determined to be that there is no electricity theft behavior. When the total difference value of the differential characteristics of electricity consumption is greater than the first differential characteristic difference threshold of electricity consumption and less than the preset second differential characteristic difference threshold of electricity consumption, or the total difference value of the differential characteristics of electricity consumption is less than the second differential characteristic difference threshold of electricity consumption and the total difference value of the fluctuation characteristics of electricity consumption is greater than the fluctuation characteristic difference threshold of electricity consumption, or the total difference value of the differential characteristics of electricity consumption is greater than the first differential characteristic difference threshold of electricity consumption and the predicted value of real-time fluctuation of electricity consumption is less than the fluctuation threshold of electricity consumption, or the total difference value of the fluctuation characteristics of electricity consumption is greater than the fluctuation characteristic difference threshold of electricity consumption and the predicted value of real-time fluctuation of electricity consumption is less than the fluctuation threshold of electricity consumption, the initial electricity theft behavior identification result is determined to be suspected electricity theft behavior; When the total difference value of the differential feature of electricity consumption is greater than the first differential feature difference threshold of electricity consumption, the total difference value of the fluctuation feature of electricity consumption is greater than the threshold of the fluctuation feature of electricity consumption, and the predicted value of the fluctuation of real-time electricity consumption is greater than the threshold of the fluctuation of electricity consumption, the initial electricity theft behavior identification result is determined to be that there is electricity theft behavior. Wherein, the second electricity consumption differential characteristic difference threshold is greater than the first electricity consumption differential characteristic difference threshold.

[0012] As a preferred embodiment, the step of inputting the historical electricity consumption data corresponding to the communication site to be detected, the target electricity consumption feature cluster, the real-time electricity consumption fluctuation prediction value, the daily electricity consumption fluctuation characteristics, and the rated power of the computer room air conditioner into the random forest model to identify electricity theft behavior and obtain the electricity theft behavior identification result of the communication site to be detected specifically includes: The historical electricity consumption data corresponding to the communication site to be detected, the target electricity consumption feature cluster, the real-time electricity consumption fluctuation prediction value, the daily electricity consumption fluctuation feature, and the rated power of the computer room air conditioner are input into several decision trees in the random forest model to identify electricity theft behavior and obtain the electricity theft behavior prediction label output by each decision tree. Each of the predicted electricity theft behavior labels is voted on, and the electricity theft behavior identification result of the communication site to be detected is determined based on the predicted electricity theft behavior label with the most votes.

[0013] As a preferred embodiment, the method specifically generates the random forest model through the following steps: Based on the site information corresponding to several communication sites, the power consumption characteristics and the power consumption characteristic clusters, the gray prediction model is used to calculate the predicted power consumption fluctuation value of each communication site. Based on the historical electricity consumption data corresponding to several communication sites, the electricity theft behavior tags corresponding to the historical electricity consumption data, the daily electricity consumption fluctuation characteristics, the rated power of the computer room air conditioner, the electricity consumption feature clusters, and the predicted electricity consumption fluctuation values, a decision tree training data is constructed. Based on the decision tree training data, a number of decision trees are generated using the Bagging algorithm; The random forest model is constructed based on several decision trees.

[0014] As a preferred embodiment, the method further includes: When the initial electricity theft behavior identification result is that electricity theft behavior exists, the initial electricity theft behavior identification result is used as the electricity theft behavior identification result of the communication station to be detected.

[0015] A second aspect of the present invention provides a device for identifying electricity theft, comprising: The electricity consumption characteristic acquisition module is used to acquire the electricity consumption characteristics of the communication site to be detected based on the historical electricity consumption data of the communication site to be detected. The electricity theft behavior identification module is used to identify the electricity theft behavior of the communication site under test based on the site information, the electricity consumption characteristics, the historical electricity consumption data and a number of preset electricity consumption characteristic clusters, and to obtain the electricity theft behavior identification result of the communication site under test. The electricity consumption feature cluster is formed by clustering the electricity consumption features of several communication sites; the electricity theft behavior identification model is composed of a gray prediction model and a random forest model.

[0016] A third aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the electricity theft detection method described in any of the first aspects.

[0017] A fourth aspect of the present invention provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the electricity theft identification method according to any one of the first aspects.

[0018] A fifth aspect of the present invention provides a computer program product, including a computer program / instructions, wherein when the computer program / instructions are executed by a processor, they implement the steps of the electricity theft behavior identification method described in any of the first aspects.

[0019] Compared with the prior art, the beneficial effect of the embodiments of the present invention is that, in the process of identifying electricity theft, by considering the electricity consumption characteristics, site information and historical electricity consumption data of the communication site to be detected, and combining the analysis with electricity consumption feature clusters, since the electricity consumption feature clusters are formed by clustering the electricity consumption characteristics of several communication sites, the differences in electricity consumption characteristics of different communication sites can be fully considered, thereby effectively and accurately identifying electricity theft based on the electricity consumption characteristics of the communication site to be detected. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating the electricity theft identification method in an embodiment of the present invention; Figure 2 This is a flowchart of the electricity theft identification process in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of the electricity theft detection device in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of the electronic device in an embodiment of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Please see Figure 1 The first aspect of this invention provides a method for identifying electricity theft, comprising the following steps S1 and S2: Step S1: Obtain the power consumption characteristics of the communication site to be tested based on its historical power consumption data; Step S2: Based on the site information of the communication site to be detected, the electricity consumption characteristics, the historical electricity consumption data, and several preset electricity consumption characteristic clusters, the electricity theft behavior identification model is used to identify the electricity theft behavior of the communication site to be detected, and the electricity theft behavior identification result of the communication site to be detected is obtained. The electricity consumption feature cluster is formed by clustering the electricity consumption features of several communication sites; the electricity theft behavior identification model is composed of a gray prediction model and a random forest model.

[0023] Specifically, due to differences in hardware and services carried by different communication sites, their electricity consumption varies across different time periods. To improve the accuracy of electricity theft detection, it is first necessary to obtain the electricity consumption characteristics of the monitored communication sites based on their historical electricity consumption data. Understandably, the daily electricity consumption data of communication sites mainly comes from the power supply bureau's data collection system or the operator's self-built energy consumption data collection platform. By obtaining the meter data of each communication site within a period (e.g., one month), the data obtained consists of the meter number and the daily electricity consumption. This requires internal data matching between the communication site and the meter number.

[0024] Furthermore, this embodiment combines the grey prediction algorithm and the random forest algorithm to construct an electricity theft behavior identification model. By considering the site information, electricity consumption characteristics, historical electricity consumption data, and several preset electricity consumption characteristic clusters of the communication site to be detected, the model integrates multiple data to identify electricity theft behavior. On the one hand, it can solve the problem of the grey prediction algorithm relying too much on historical data. By combining the random forest algorithm, it can reduce the situation of missed identification and false identification. On the other hand, it can fully consider the differences in electricity consumption characteristics of different communication sites, thereby improving the accuracy of electricity theft behavior identification.

[0025] The electricity theft identification method provided in this invention considers the electricity consumption characteristics, site information, and historical electricity consumption data of the communication site to be detected during the electricity theft identification process, and combines the analysis with electricity consumption feature clusters. Since the electricity consumption feature clusters are formed by clustering the electricity consumption characteristics of several communication sites, they can fully consider the differences in electricity consumption characteristics of different communication sites, thereby effectively and accurately identifying electricity theft behavior based on the electricity consumption characteristics of the communication site to be detected.

[0026] As a preferred embodiment, the method specifically obtains the electricity consumption characteristics through the following steps: Based on the historical electricity consumption data of any communication station, the electricity consumption change characteristics for each preset time period are obtained; wherein, the electricity consumption change characteristics include the first-order differential characteristics of electricity consumption, the second-order differential characteristics of electricity consumption, and the electricity consumption fluctuation characteristics. Based on the historical electricity consumption data, obtain the daily electricity consumption of the computer room and the average daily electricity consumption of the computer room for any one of the communication sites within a preset statistical period; Based on the preset statistical period, the daily power consumption of the computer room, and the average daily power consumption of the computer room, the daily power consumption fluctuation characteristics of any one communication site are determined. Based on the electricity consumption change characteristics, the daily electricity consumption fluctuation characteristics, and the rated power of the air conditioner in the computer room of any one communication site, the electricity consumption characteristics of any one communication site are determined.

[0027] Specifically, whether for the electricity consumption characteristics of the communication site to be detected, or for the electricity consumption characteristics of different communication sites required for training the electricity theft identification model, this embodiment first needs to obtain the electricity consumption change characteristics for each preset time period based on the historical electricity consumption data of the communication site. In this embodiment, the electricity consumption change characteristics include first-order differential characteristics, second-order differential characteristics, and electricity fluctuation characteristics. It can be understood that the first-order differential characteristics characterize the rate of change of electricity consumption within different time periods, the second-order differential characteristics characterize the acceleration of change of electricity consumption within different time periods, and the electricity fluctuation characteristics characterize the severity of electricity fluctuations in future time periods. These electricity consumption change characteristics can reflect the common characteristics of various types of communication sites, thereby being used to determine electricity theft behavior.

[0028] Furthermore, considering that the power consumption of a communication equipment room should fluctuate steadily within a certain range under normal circumstances during a certain period, this embodiment determines the daily power consumption fluctuation characteristics of the communication site based on a preset statistical period and the daily power consumption and average daily power consumption of the equipment room within that preset statistical period. Combined with the rated power of the equipment room air conditioner, this further improves the accuracy of identifying abnormal power consumption at the communication site.

[0029] As a preferred embodiment, the step of obtaining the electricity consumption change characteristics for each preset time period based on the historical electricity consumption data of any communication station specifically includes: Obtain the electricity consumption sequence for each preset time period from the historical electricity consumption data; Based on the electricity consumption sequence of two adjacent preset time periods, determine the first-order differential feature of electricity consumption for each preset time period; Based on the first-order differential characteristics of electricity consumption in two adjacent preset time periods, determine the second-order differential characteristics of electricity consumption in each preset time period. Based on a preset first time interval and a preset second time interval, calculate the first tilt angle corresponding to the line connecting the electricity consumption sequence of each preset time period and the electricity consumption sequence that follows it and is separated by the first time interval, and the second tilt angle corresponding to the line connecting the electricity consumption sequence of each preset time period and the electricity consumption sequence that follows it and is separated by the second time interval; wherein, the first time interval is less than the second time interval. The power consumption fluctuation characteristics for each preset time period are determined based on the difference between the first tilt angle and the second tilt angle.

[0030] Specifically, in this embodiment, the first-order differential characteristic of electricity consumption for each preset time period is calculated using the following expression: ; in, ; The time interval is a preset value, such as half an hour or one hour. This embodiment does not specify a particular time interval. For Time series of time intervals i The corresponding time, For the first i The time after The electricity consumption data sequence before the time interval, i.e., the time series. i Time-series electricity consumption data, Then it is the first i -1 hour onwards A sequence of electricity consumption data prior to the time interval; Indicates Time series of time intervals i The first-order differential characteristic of electricity consumption.

[0031] Furthermore, in this embodiment, the second-order differential characteristic of electricity consumption for each preset time period is calculated using the following expression: ; in, Indicates Time series of time intervals i The second-order differential characteristics of electricity consumption.

[0032] Furthermore, in this embodiment, the first tilt angle and the second tilt angle are calculated using the following expressions: ; ; in, No. i+s The time after A sequence of electricity consumption data prior to the time interval; S The first time interval mentioned above, i.e. t i+s - t i ; The first tilt angle is used to characterize the degree of fluctuation in electricity consumption in the near future, for example, the first time interval. S It can contain two of the above time intervals. 3 time intervals Etc., this embodiment is not specifically limited here; No. i+l The time after A sequence of electricity consumption data prior to the time interval; L The second time interval mentioned above, i.e. ti+l - t i ; The second tilt angle is used to characterize the degree of fluctuation in electricity consumption over a long future period, for example, the second time interval. L It can contain four of the above time intervals. 5 time intervals Etc., this embodiment is not specifically limited here.

[0033] Then, through the expression: This allows us to determine the fluctuation characteristics of electricity consumption in each preset time period. .

[0034] As a preferred embodiment, determining the daily power consumption fluctuation characteristics of any communication site based on the preset statistical period, the daily power consumption of the data center, and the average daily power consumption of the data center specifically includes: Calculate the standard deviation of the daily power consumption of the data center based on the preset statistical period, the daily power consumption of the data center, and the average daily power consumption of the data center; The coefficient of variation of daily power consumption is obtained by comparing the standard deviation with the average daily power consumption of the computer room. Based on the preset statistical period, the daily power consumption of the computer room, the average daily power consumption of the computer room, and the standard deviation, the daily power consumption kurtosis and daily power consumption skewness are calculated respectively. The daily electricity consumption fluctuation characteristics are determined based on the daily electricity consumption variation coefficient, the daily electricity consumption kurtosis, and the daily electricity consumption skewness.

[0035] Specifically, this embodiment introduces the coefficient of variation to observe the fluctuation of daily power consumption in the computer room, and introduces the kurtosis and skewness of daily power consumption to eliminate noise data such as power outages and service interruptions, so as to improve the accuracy of identifying abnormal power consumption at communication sites.

[0036] Understandably, since the power consumption of each communication site varies, the power consumption fluctuation characteristics of low-power-consumption sites differ significantly from those of high-power-consumption sites, requiring normalization. Therefore, this embodiment introduces the coefficient of variation through the following expression. CV (Coefficient of Variation) When electricity consumption fluctuates steadily, the coefficient of variation should be relatively concentrated and tend to be 0.

[0037] ; ; ; in, nIndicates the preset statistical period; x i Indicates the first i Daily electricity consumption of the computer room; This indicates the average daily power consumption of the computer room; This represents the average daily electricity consumption of the computer room within a preset statistical period.

[0038] Furthermore, kurtosis is used to describe the peak value of a set of data, reflecting the relative sharpness or flatness of a distribution compared to a normal distribution. Positive peak values ​​indicate a relatively sharp distribution, while negative peak values ​​indicate a relatively flat distribution. In this embodiment, the kurtosis of daily electricity consumption is calculated using the following expression. KURT : ; Furthermore, skewness is used to describe the degree of skewness of the distribution, indicating the degree of asymmetry of the distribution relative to the mean. Positive skewness indicates that the asymmetric tails of the distribution tend to have more positive values, while negative skewness indicates that the asymmetric tails of the distribution tend to have more negative values. In this embodiment, the daily electricity consumption skewness is calculated using the following expression. SKEW : ; Furthermore, in this embodiment, the rated power of the computer room air conditioner is the rated power of all air conditioners in the communication computer room.

[0039] As a preferred embodiment, the method specifically obtains the electricity consumption characteristic clusters through the following steps: The similarity of the electricity consumption change characteristics of each of the communication stations in each of the preset time periods is calculated to determine the feature similarity between the communication stations. Based on the feature similarity, the electricity consumption change features of each of the communication stations in each of the preset time periods are clustered to obtain several initial electricity consumption feature clusters. The mean value of several electricity consumption change features in each preset time period in the initial electricity consumption feature cluster is calculated to determine the station electricity consumption features in each preset time period in the initial electricity consumption feature cluster. Based on the electricity consumption characteristics of several stations in each of the initial electricity consumption characteristic clusters, several electricity consumption characteristic clusters are generated.

[0040] Specifically, because the number of terminals and the services carried by each communication site are different, the power consumption characteristics of the corresponding communication equipment are different, which will cause problems in power consumption characteristic identification. Therefore, this embodiment adopts a multi-data clustering method to reduce the feature description error caused by different service volumes, different hardware power consumption, and short-term jitter. That is, the first-order differential features of power consumption, the second-order differential features of power consumption, and the jitter features of power consumption are clustered separately, and finally multiple clusters of first-order differential features of power consumption, multiple clusters of second-order differential features of power consumption, and multiple clusters of jitter features of power consumption are formed.

[0041] First, this embodiment calculates the similarity of the electricity consumption change characteristics of each communication site in each preset time period using the following expression: ; in, loss For communication sites s 1. Communication site s 2 at time a At that time b Similarity of electricity consumption change characteristics; For communication sites s 1 in the i The characteristics of electricity consumption changes over a preset time period (e.g., first-order differential characteristics of electricity consumption, second-order differential characteristics of electricity consumption, or fluctuating characteristics of electricity consumption). For communication sites s 2 in the i The characteristics of electricity consumption changes over a preset time period (e.g., first-order differential characteristics of electricity consumption, second-order differential characteristics of electricity consumption, or fluctuating characteristics of electricity consumption).

[0042] By analyzing the similarity between the first-order differential characteristics, second-order differential characteristics, and fluctuation characteristics of power consumption at different communication sites... loss Clustering can yield several initial electricity consumption characteristic clusters, including multiple initial electricity consumption first-order differential characteristic clusters, multiple initial electricity consumption second-order differential characteristic clusters, and multiple initial electricity consumption fluctuation characteristic clusters.

[0043] Furthermore, the time-series data of electricity consumption variation characteristics within the same cluster are averaged to obtain multiple clusters of first-order differential electricity consumption characteristics, multiple clusters of second-order differential electricity consumption characteristics, and multiple clusters of electricity consumption fluctuation characteristics. A single cluster of first-order differential electricity consumption characteristics can be represented as: A single second-order differential feature cluster of electricity consumption can be represented as: A cluster of individual electricity consumption fluctuation features can be represented as: ,in, This represents the average value of the first-order differential characteristics of electricity consumption in the first preset time period within the first-order differential characteristic cluster of electricity consumption. This represents the average value of the second-order differential characteristics of electricity consumption in the first preset time period within the second-order differential characteristic cluster of electricity consumption. The average value of the electricity consumption fluctuation features in the first preset time period in the electricity consumption fluctuation feature cluster is used. Other parameters are similar, and will not be described in detail here.

[0044] Ultimately, these can be further combined to form cluster feature values: ,in, Clusters of the same type of communication sites a Clustering feature values, For clusters a The electricity consumption first-order differential feature clusters in the data, For clusters a The second-order differential characteristic cluster of electricity consumption in the data, For clusters a Clustering of electricity consumption fluctuation characteristics.

[0045] As a preferred embodiment, the step of identifying electricity theft behavior of the communication site under test based on the site information, electricity consumption characteristics, historical electricity consumption data, and several preset electricity consumption characteristic clusters, using an electricity theft behavior identification model, and obtaining the electricity theft behavior identification result of the communication site under test, specifically includes: Based on the gray prediction model, the similarity calculation is performed between the electricity consumption change characteristics of the communication site to be detected and the electricity consumption characteristics of each site in several electricity consumption feature clusters, so as to obtain the target electricity consumption feature cluster that matches the electricity consumption change characteristics from several electricity consumption feature clusters; Based on the site information, the electricity consumption change characteristics, and the target electricity consumption feature cluster, the gray prediction model is used to calculate the real-time electricity consumption jitter prediction value and the initial electricity theft behavior identification result of the communication site to be detected. When the initial electricity theft behavior identification result does not indicate the existence of electricity theft behavior, the historical electricity consumption data corresponding to the communication site to be detected, the target electricity consumption feature cluster, the real-time electricity consumption fluctuation prediction value, the daily electricity consumption fluctuation feature, and the rated power of the computer room air conditioner are input into the random forest model to identify electricity theft behavior and obtain the electricity theft behavior identification result of the communication site to be detected.

[0046] Specifically, in order to accurately identify whether a communication site under test is stealing electricity, this embodiment first uses a grey prediction model to calculate the similarity between the electricity consumption change characteristics of the communication site under test and the electricity consumption characteristics of each site in several electricity consumption feature clusters, as shown in the following expression: ; in, loss u Communication site to be tested u Clustering with electricity consumption characteristics s At any moment a At that time b Similarity of electricity consumption change characteristics; Communication site to be tested u In the i The characteristics of electricity consumption changes over a preset time period (e.g., first-order differential characteristics of electricity consumption, second-order differential characteristics of electricity consumption, or fluctuating characteristics of electricity consumption). Clustering based on electricity consumption characteristics s In the i The characteristics of electricity consumption changes over a preset time period (e.g., first-order differential characteristics, second-order differential characteristics, or fluctuation characteristics) are then analyzed. loss u Minimal electricity consumption characteristic cluster s Clusters based on target electricity consumption characteristics.

[0047] Furthermore, because the hardware, services, and temperatures of communication sites, such as equipment rooms, vary and change in real time, algorithms trained on large datasets are insufficient to represent the unique characteristics of each site. Therefore, a model for identifying electricity theft based on the specific characteristics of each communication site is necessary. In this embodiment, the grey prediction model calculates the predicted real-time electricity consumption fluctuations of the communication site under test based on its site information and target electricity consumption feature clusters. Since the target electricity consumption feature clusters represent common electricity consumption characteristics of communication sites of the same type as the site under test, the model outputs an initial electricity theft identification result based on the electricity consumption change characteristics of the communication site under test.

[0048] Furthermore, since the grey prediction model mainly analyzes data trends, and the data is in hourly units, it is more sensitive to data and can easily obtain the identification results of electricity theft behavior in a specific time period. However, its output results rely excessively on historical data. Therefore, in this embodiment, when the initial electricity theft behavior identification result does not indicate the existence of electricity theft behavior, the random forest model is further used to perform robust analysis on real-time data. By performing robust data analysis on a daily basis, the identification results of electricity theft behavior over a long period can be obtained.

[0049] As a preferred embodiment, the method specifically obtains the real-time power consumption fluctuation prediction value through the following steps: Based on the site information, obtain the number of service terminals added to the communication site to be detected, the power consumption calculation data caused by computing load, and the site air conditioning temperature; Obtain the station's electricity consumption fluctuation characteristics for the time period corresponding to the current prediction time from the target electricity consumption characteristic cluster; Based on the gray prediction model, and according to the number of new service terminals, the electricity consumption calculation data, the site air conditioning temperature, and the site electricity consumption fluctuation characteristics, the real-time electricity consumption fluctuation prediction value is calculated using the following expression: ; in, a , b , c and d These are the preset first fitting coefficient, second fitting coefficient, third fitting coefficient, and fourth fitting coefficient, respectively; This indicates the increase in the number of service terminals; This represents the electricity consumption calculation data; This indicates the air conditioning temperature at the site; This indicates the fluctuation characteristics of the power consumption at the site; This represents the preset prediction residual; This represents the predicted real-time power consumption fluctuation value.

[0050] Specifically, the above-mentioned electricity consumption fluctuation prediction expression can be obtained by fitting historical electricity consumption fluctuation data of different communication sites and corresponding site information (the number of service terminals added in different time periods, electricity consumption calculation data caused by computing load and site air conditioning temperature) and the electricity consumption fluctuation characteristics of sites in different time periods in the electricity consumption feature cluster.

[0051] As a preferred embodiment, the method specifically obtains the initial electricity theft behavior identification result through the following steps: Based on the gray prediction model, the difference identifier of the first-order differential feature of the electricity consumption of the communication station to be detected in each preset time period is determined by multiplying the first-order differential feature of the electricity consumption of the communication station to be detected with the first-order differential feature of the electricity consumption of the station in the target electricity consumption feature cluster. The difference identifier of the second-order differential feature of the electricity consumption of the communication station to be detected in each preset time period is determined by multiplying the second-order differential feature of the electricity consumption of the station to be detected with the second-order differential feature of the electricity consumption of the station in the target electricity consumption feature cluster. Based on the first-order differential characteristic difference identifier of electricity consumption and the second-order differential characteristic difference identifier of electricity consumption, the total differential characteristic difference value of electricity consumption of the communication station to be detected is determined. The total difference value of the power consumption fluctuation feature of the communication station under test is determined based on the difference between the preset power consumption fluctuation feature threshold and the power consumption fluctuation feature of the communication station under test in each preset time period. When the total difference value of the differential characteristics of electricity consumption is less than the preset first differential characteristic difference threshold of electricity consumption, the total difference value of the fluctuation characteristics of electricity consumption is less than the preset fluctuation characteristic difference threshold of electricity consumption, and the predicted value of the real-time fluctuation of electricity consumption is less than the preset fluctuation threshold of electricity consumption, the initial electricity theft behavior identification result is determined to be that there is no electricity theft behavior. When the total difference value of the differential characteristics of electricity consumption is greater than the first differential characteristic difference threshold of electricity consumption and less than the preset second differential characteristic difference threshold of electricity consumption, or the total difference value of the differential characteristics of electricity consumption is less than the second differential characteristic difference threshold of electricity consumption and the total difference value of the fluctuation characteristics of electricity consumption is greater than the fluctuation characteristic difference threshold of electricity consumption, or the total difference value of the differential characteristics of electricity consumption is greater than the first differential characteristic difference threshold of electricity consumption and the predicted value of real-time fluctuation of electricity consumption is less than the fluctuation threshold of electricity consumption, or the total difference value of the fluctuation characteristics of electricity consumption is greater than the fluctuation characteristic difference threshold of electricity consumption and the predicted value of real-time fluctuation of electricity consumption is less than the fluctuation threshold of electricity consumption, the initial electricity theft behavior identification result is determined to be suspected electricity theft behavior; When the total difference value of the differential feature of electricity consumption is greater than the first differential feature difference threshold of electricity consumption, the total difference value of the fluctuation feature of electricity consumption is greater than the threshold of the fluctuation feature of electricity consumption, and the predicted value of the fluctuation of real-time electricity consumption is greater than the threshold of the fluctuation of electricity consumption, the initial electricity theft behavior identification result is determined to be that there is electricity theft behavior. Wherein, the second electricity consumption differential characteristic difference threshold is greater than the first electricity consumption differential characteristic difference threshold.

[0052] Specifically, in this embodiment, the first-order differential characteristic difference identifier and the second-order differential characteristic difference identifier of electricity consumption are determined by the following expressions: ; ; in, and These represent the differences in the first-order differential characteristics of electricity consumption and the differences in the second-order differential characteristics of electricity consumption, respectively. Indicates the communication site to be tested u In the i The first-order differential characteristic of electricity consumption over a preset time period; Indicates the communication site to be tested u In the i The second-order differential characteristic of electricity consumption over a preset time period; Clusters representing target electricity consumption characteristics s In the i First-order differential characteristics of station electricity consumption over a preset time period; Clusters representing target electricity consumption characteristics s In thei The second-order differential characteristics of electricity consumption at stations within a preset time period.

[0053] Furthermore, in this embodiment, the total difference value of the differential characteristic of the power consumption of the communication site to be detected is determined by the following expression: ; in, th 1 represents the total difference in the differential characteristics of electricity consumption; T This represents the total number of preset time periods, which is the cycle for identifying whether electricity theft is occurring. Preferably, T =48, meaning the time-series electricity consumption data is divided into half-hour intervals, with a total of 48 sets of time-series electricity consumption data per day, and one day as the minimum period for detecting electricity theft. It can be understood that the larger the total difference value of this electricity consumption differential characteristic, the greater the deviation between the electricity consumption differential characteristics of the detected communication site and the electricity consumption differential characteristics of similar communication sites under normal conditions, thus increasing the likelihood of electricity theft.

[0054] Furthermore, in this embodiment, the total difference value of electricity consumption fluctuation characteristics is determined by the following expression: ; in, th 2 represents the total difference in electricity consumption fluctuation characteristics; The threshold for electricity consumption fluctuation features is preferably set as follows: based on the historical electricity consumption data and corresponding electricity consumption fluctuation features of each communication station in the target electricity consumption feature cluster, the electricity consumption fluctuation features of each preset time period are obtained in the historical electricity consumption data (in days) in which there is no electricity theft, and the largest daily electricity consumption fluctuation feature is used as the electricity consumption fluctuation feature threshold. Indicates the communication station to be detected at the th i The power consumption fluctuation characteristics are measured over a preset time period. It is understandable that the larger the total difference value of these power consumption fluctuation characteristics, the greater the deviation between the power consumption fluctuation characteristics of the communication site under test and those of similar communication sites under normal conditions, thus increasing the likelihood of electricity theft.

[0055] Furthermore, in this embodiment, a first power consumption differential feature difference threshold and a second power consumption differential feature difference threshold are preset. Preferably, the first power consumption differential feature difference threshold and the second power consumption differential feature difference threshold are set as follows: based on the historical power consumption data of each communication station in the target power consumption feature cluster and the corresponding first-order differential feature and second-order differential feature of power consumption, the total difference value of the differential feature of power consumption of each communication station in the historical power consumption data (in days) in which there is no power theft is calculated, and 90% of the maximum value of the total difference ...

[0056] Furthermore, this embodiment also pre-sets a threshold for the difference in electricity consumption fluctuation characteristics. Preferably, the threshold for the difference in electricity consumption fluctuation characteristics is set as follows: based on the historical electricity consumption data and corresponding electricity consumption fluctuation characteristics of each communication station in the target electricity consumption feature cluster, the total difference value of the electricity consumption fluctuation characteristics of each communication station in the historical electricity consumption data (in days) in which there is no electricity theft is calculated, and 90% of the average value of the total difference value of each electricity consumption fluctuation characteristic is taken as the threshold for the difference in electricity consumption fluctuation characteristics.

[0057] Furthermore, this embodiment also pre-sets a power consumption fluctuation threshold for the real-time power consumption fluctuation prediction value. Preferably, the power consumption fluctuation threshold is set by: calculating the absolute difference between the real-time power consumption fluctuation prediction value and the power consumption fluctuation characteristics of the communication station to be detected in each preset time period, and selecting the maximum value of the absolute difference as the power consumption fluctuation threshold.

[0058] It is worth noting that the setting methods for the first power consumption differential characteristic difference threshold, the second power consumption differential characteristic difference threshold, the power consumption fluctuation characteristic difference threshold, and the power consumption fluctuation threshold in this embodiment are not limited to the setting methods mentioned above. Other methods can also be used, such as setting based on human experience. This embodiment will not elaborate further on these methods.

[0059] As a preferred embodiment, the step of inputting the historical electricity consumption data corresponding to the communication site to be detected, the target electricity consumption feature cluster, the real-time electricity consumption fluctuation prediction value, the daily electricity consumption fluctuation characteristics, and the rated power of the computer room air conditioner into the random forest model to identify electricity theft behavior and obtain the electricity theft behavior identification result of the communication site to be detected specifically includes: The historical electricity consumption data corresponding to the communication site to be detected, the target electricity consumption feature cluster, the real-time electricity consumption fluctuation prediction value, the daily electricity consumption fluctuation feature, and the rated power of the computer room air conditioner are input into several decision trees in the random forest model to identify electricity theft behavior and obtain the electricity theft behavior prediction label output by each decision tree. Each of the predicted electricity theft behavior labels is voted on, and the electricity theft behavior identification result of the communication site to be detected is determined based on the predicted electricity theft behavior label with the most votes.

[0060] Specifically, in this embodiment, the historical electricity consumption data, target electricity consumption feature clusters, real-time electricity consumption fluctuation predictions, daily electricity consumption fluctuation characteristics, and rated power of the computer room air conditioner corresponding to the communication site to be detected are used as input feature data for each decision tree in the random forest model. Each decision tree is used to identify abnormal sites based on the input feature data, thereby obtaining the electricity theft behavior prediction label output by each decision tree. It can be understood that the electricity theft behavior prediction label is used to characterize whether the current communication site to be detected has electricity theft behavior. For example, the value of the electricity theft behavior prediction label is 0 or 1. When the value of the electricity theft behavior prediction label is 0, it means that the current communication site to be detected does not have electricity theft behavior; when the value of the electricity theft behavior prediction label is 1, it means that the current communication site to be detected has electricity theft behavior.

[0061] Furthermore, the votes for each electricity theft prediction tag are counted, including the votes for tags indicating the presence of electricity theft and the votes for tags indicating the absence of electricity theft. If the tag with the most votes indicates the presence of electricity theft, then the electricity theft identification result for the communication site under test is that electricity theft exists; if the tag with the most votes indicates the absence of electricity theft, then the electricity theft identification result for the communication site under test is that electricity theft does not exist. Preferably, this embodiment uses a weighted voting method to vote on each electricity theft prediction tag. Assuming that the number of votes for the electricity theft prediction tag indicating the presence of electricity theft is n, the number of votes for the electricity theft prediction tag indicating the absence of electricity theft is m, and the weighting coefficient is α, where α is greater than 1, then if α × n > m, then the electricity theft identification result for the communication site under test is determined to be that electricity theft exists; otherwise, the electricity theft identification result for the communication site under test is determined to be that electricity theft does not exist.

[0062] As a preferred embodiment, the method specifically generates the random forest model through the following steps: Based on the site information corresponding to several communication sites, the power consumption characteristics and the power consumption characteristic clusters, the gray prediction model is used to calculate the predicted power consumption fluctuation value of each communication site. Based on the historical electricity consumption data corresponding to several communication sites, the electricity theft behavior tags corresponding to the historical electricity consumption data, the daily electricity consumption fluctuation characteristics, the rated power of the computer room air conditioner, the electricity consumption feature clusters, and the predicted electricity consumption fluctuation values, a decision tree training data is constructed. Based on the decision tree training data, a number of decision trees are generated using the Bagging algorithm; The random forest model is constructed based on several decision trees.

[0063] Specifically, this embodiment uses historical electricity consumption data corresponding to several communication sites, electricity theft behavior labels corresponding to the historical electricity consumption data (used to indicate whether historical electricity consumption data in different time periods contains electricity theft behavior), daily electricity consumption fluctuation characteristics, rated power of computer room air conditioners, electricity consumption feature clusters, and predicted electricity consumption fluctuation values ​​to construct decision tree training data. Then, the Bagging algorithm (Bootstrap aggregating) is used to randomly select data from the decision tree training data with replacement, while randomly selecting some feature data as input to achieve the training and generation of the decision tree. During the training and generation process of the decision tree, the splitting threshold for each feature data is determined by information gain, information gain ratio, Gini coefficient, etc. Then, node splitting is performed, that is, after selecting the root node, features are selected to form internal nodes, and the dataset is divided into different subsets by the internal node features and thresholds. Then, the feature selection and node splitting process is recursively applied to each subset until the stopping condition is met. Finally, overfitting is checked, and redundant branch nodes are pruned.

[0064] Furthermore, during the training of the random forest model, this embodiment trained the model using four factors: the daily electricity consumption variation coefficient as a single factor, three factors in the daily electricity consumption fluctuation characteristics, and the daily electricity consumption fluctuation characteristics and the rated power of the data center air conditioner. The model was then validated using a test set. The results showed that using all four factors—the daily electricity consumption fluctuation characteristics and the rated power of the data center air conditioner—resulted in the best electricity theft detection performance, as shown in Table 1 below. Table 1 Comparison of Model Training Results

[0065] Among them, the accuracy rate of electricity theft is the ratio of the number of accurately identified sites to the number of sites identified by the model, and the detection rate is the ratio of the number of accurately identified sites to the number of sites with electricity theft in the test set. As shown in Table 1 above, when using four factors including daily electricity consumption fluctuation characteristics and the rated power of the computer room air conditioner, the detection rate of electricity theft reaches 71.43% and the accuracy rate reaches 80.65%. Finally, by combining the grey prediction model and the random forest model generated using the four factors including daily electricity consumption fluctuation characteristics and the rated power of the computer room air conditioner to identify electricity theft behavior, the detection rate of electricity theft reaches 82.85% and the accuracy rate reaches 93.55%, indicating that the identification effect of electricity theft behavior is good.

[0066] As a preferred embodiment, the method further includes: When the initial electricity theft behavior identification result is that electricity theft behavior exists, the initial electricity theft behavior identification result is used as the electricity theft behavior identification result of the communication station to be detected.

[0067] Specifically, such as Figure 2 As shown, in this embodiment, if the initial power theft behavior identification result is that power theft behavior exists, the communication site to be detected is directly determined to have power theft behavior. If the initial power theft behavior identification result is that power theft behavior is suspected or that power theft behavior does not exist, the decision tree in the random forest model is further used to calculate and output the final power theft behavior identification result.

[0068] Please see Figure 3 A second aspect of the present invention provides a device 100 for identifying electricity theft, comprising: The electricity consumption feature acquisition module 11 is used to acquire the electricity consumption features of the communication site to be detected based on the historical electricity consumption data of the communication site to be detected. The electricity theft behavior identification module 12 is used to identify the electricity theft behavior of the communication site to be detected based on the site information of the communication site to be detected, the electricity consumption characteristics, the historical electricity consumption data and a number of preset electricity consumption characteristic clusters, and to obtain the electricity theft behavior identification result of the communication site to be detected. The electricity consumption feature cluster is formed by clustering the electricity consumption features of several communication sites; the electricity theft behavior identification model is composed of a gray prediction model and a random forest model.

[0069] As a preferred embodiment, the electricity consumption characteristic acquisition module 11 acquires the electricity consumption characteristic through the following steps: Based on the historical electricity consumption data of any communication station, the electricity consumption change characteristics for each preset time period are obtained; wherein, the electricity consumption change characteristics include the first-order differential characteristics of electricity consumption, the second-order differential characteristics of electricity consumption, and the electricity consumption fluctuation characteristics. Based on the historical electricity consumption data, obtain the daily electricity consumption of the computer room and the average daily electricity consumption of the computer room for any one of the communication sites within a preset statistical period; Based on the preset statistical period, the daily power consumption of the computer room, and the average daily power consumption of the computer room, the daily power consumption fluctuation characteristics of any one communication site are determined. Based on the electricity consumption change characteristics, the daily electricity consumption fluctuation characteristics, and the rated power of the air conditioner in the computer room of any one communication site, the electricity consumption characteristics of any one communication site are determined.

[0070] As a preferred embodiment, the electricity consumption characteristic acquisition module 11 is used to acquire the electricity consumption change characteristics for each preset time period based on the historical electricity consumption data of any communication station, specifically including: Obtain the electricity consumption sequence for each preset time period from the historical electricity consumption data; Based on the electricity consumption sequence of two adjacent preset time periods, determine the first-order differential feature of electricity consumption for each preset time period; Based on the first-order differential characteristics of electricity consumption in two adjacent preset time periods, determine the second-order differential characteristics of electricity consumption in each preset time period. Based on a preset first time interval and a preset second time interval, calculate the first tilt angle corresponding to the line connecting the electricity consumption sequence of each preset time period and the electricity consumption sequence that follows it and is separated by the first time interval, and the second tilt angle corresponding to the line connecting the electricity consumption sequence of each preset time period and the electricity consumption sequence that follows it and is separated by the second time interval; wherein, the first time interval is less than the second time interval. The power consumption fluctuation characteristics for each preset time period are determined based on the difference between the first tilt angle and the second tilt angle.

[0071] As a preferred embodiment, the electricity consumption characteristic acquisition module 11 is used to determine the daily electricity consumption fluctuation characteristics of any one communication site based on the preset statistical period, the daily electricity consumption of the computer room, and the average daily electricity consumption of the computer room, specifically including: Calculate the standard deviation of the daily power consumption of the data center based on the preset statistical period, the daily power consumption of the data center, and the average daily power consumption of the data center; The coefficient of variation of daily power consumption is obtained by comparing the standard deviation with the average daily power consumption of the computer room. Based on the preset statistical period, the daily power consumption of the computer room, the average daily power consumption of the computer room, and the standard deviation, the daily power consumption kurtosis and daily power consumption skewness are calculated respectively. The daily electricity consumption fluctuation characteristics are determined based on the daily electricity consumption variation coefficient, the daily electricity consumption kurtosis, and the daily electricity consumption skewness.

[0072] As a preferred embodiment, the device further includes an electricity consumption characteristic clustering module, used for: The similarity of the electricity consumption change characteristics of each of the communication stations in each of the preset time periods is calculated to determine the feature similarity between the communication stations. Based on the feature similarity, the electricity consumption change features of each of the communication stations in each of the preset time periods are clustered to obtain several initial electricity consumption feature clusters. The mean value of several electricity consumption change features in each preset time period in the initial electricity consumption feature cluster is calculated to determine the station electricity consumption features in each preset time period in the initial electricity consumption feature cluster. Based on the electricity consumption characteristics of several stations in each of the initial electricity consumption characteristic clusters, several electricity consumption characteristic clusters are generated.

[0073] As a preferred embodiment, the electricity theft behavior identification module 12 is used to identify electricity theft behavior of the communication site under test based on the site information of the communication site to be detected, the electricity consumption characteristics, the historical electricity consumption data, and several preset electricity consumption characteristic clusters, using an electricity theft behavior identification model to obtain the electricity theft behavior identification result of the communication site to be detected, specifically including: Based on the gray prediction model, the similarity calculation is performed between the electricity consumption change characteristics of the communication site to be detected and the electricity consumption characteristics of each site in several electricity consumption feature clusters, so as to obtain the target electricity consumption feature cluster that matches the electricity consumption change characteristics from several electricity consumption feature clusters; Based on the site information, the electricity consumption change characteristics, and the target electricity consumption feature cluster, the gray prediction model is used to calculate the real-time electricity consumption jitter prediction value and the initial electricity theft behavior identification result of the communication site to be detected. When the initial electricity theft behavior identification result does not indicate the existence of electricity theft behavior, the historical electricity consumption data corresponding to the communication site to be detected, the target electricity consumption feature cluster, the real-time electricity consumption fluctuation prediction value, the daily electricity consumption fluctuation feature, and the rated power of the computer room air conditioner are input into the random forest model to identify electricity theft behavior and obtain the electricity theft behavior identification result of the communication site to be detected.

[0074] As a preferred embodiment, the electricity theft detection module 12 obtains the real-time electricity consumption fluctuation prediction value through the following steps: Based on the site information, obtain the number of service terminals added to the communication site to be detected, the power consumption calculation data caused by computing load, and the site air conditioning temperature; Obtain the station's electricity consumption fluctuation characteristics for the time period corresponding to the current prediction time from the target electricity consumption characteristic cluster; Based on the gray prediction model, and according to the number of new service terminals, the electricity consumption calculation data, the site air conditioning temperature, and the site electricity consumption fluctuation characteristics, the real-time electricity consumption fluctuation prediction value is calculated using the following expression: ; in, a , b , c and d These are the preset first fitting coefficient, second fitting coefficient, third fitting coefficient, and fourth fitting coefficient, respectively; This indicates the increase in the number of service terminals; This represents the electricity consumption calculation data; This indicates the air conditioning temperature at the site; This indicates the fluctuation characteristics of the power consumption at the site; This represents the preset prediction residual; This represents the predicted real-time power consumption fluctuation value.

[0075] As a preferred embodiment, the electricity theft behavior identification module 12 obtains the initial electricity theft behavior identification result through the following steps: Based on the gray prediction model, the difference identifier of the first-order differential feature of the electricity consumption of the communication station to be detected in each preset time period is determined by multiplying the first-order differential feature of the electricity consumption of the communication station to be detected with the first-order differential feature of the electricity consumption of the station in the target electricity consumption feature cluster. The difference identifier of the second-order differential feature of the electricity consumption of the communication station to be detected in each preset time period is determined by multiplying the second-order differential feature of the electricity consumption of the station to be detected with the second-order differential feature of the electricity consumption of the station in the target electricity consumption feature cluster. Based on the first-order differential characteristic difference identifier of electricity consumption and the second-order differential characteristic difference identifier of electricity consumption, the total differential characteristic difference value of electricity consumption of the communication station to be detected is determined. The total difference value of the power consumption fluctuation feature of the communication station under test is determined based on the difference between the preset power consumption fluctuation feature threshold and the power consumption fluctuation feature of the communication station under test in each preset time period. When the total difference value of the differential characteristics of electricity consumption is less than the preset first differential characteristic difference threshold of electricity consumption, the total difference value of the fluctuation characteristics of electricity consumption is less than the preset fluctuation characteristic difference threshold of electricity consumption, and the predicted value of the real-time fluctuation of electricity consumption is less than the preset fluctuation threshold of electricity consumption, the initial electricity theft behavior identification result is determined to be that there is no electricity theft behavior. When the total difference value of the differential characteristics of electricity consumption is greater than the first differential characteristic difference threshold of electricity consumption and less than the preset second differential characteristic difference threshold of electricity consumption, or the total difference value of the differential characteristics of electricity consumption is less than the second differential characteristic difference threshold of electricity consumption and the total difference value of the fluctuation characteristics of electricity consumption is greater than the fluctuation characteristic difference threshold of electricity consumption, or the total difference value of the differential characteristics of electricity consumption is greater than the first differential characteristic difference threshold of electricity consumption and the predicted value of real-time fluctuation of electricity consumption is less than the fluctuation threshold of electricity consumption, or the total difference value of the fluctuation characteristics of electricity consumption is greater than the fluctuation characteristic difference threshold of electricity consumption and the predicted value of real-time fluctuation of electricity consumption is less than the fluctuation threshold of electricity consumption, the initial electricity theft behavior identification result is determined to be suspected electricity theft behavior; When the total difference value of the differential feature of electricity consumption is greater than the first differential feature difference threshold of electricity consumption, the total difference value of the fluctuation feature of electricity consumption is greater than the threshold of the fluctuation feature of electricity consumption, and the predicted value of the fluctuation of real-time electricity consumption is greater than the threshold of the fluctuation of electricity consumption, the initial electricity theft behavior identification result is determined to be that there is electricity theft behavior. Wherein, the second electricity consumption differential characteristic difference threshold is greater than the first electricity consumption differential characteristic difference threshold.

[0076] As a preferred embodiment, the electricity theft behavior identification module 12 is used to input the historical electricity consumption data corresponding to the communication site to be detected, the target electricity consumption feature cluster, the real-time electricity consumption fluctuation prediction value, the daily electricity consumption fluctuation characteristics, and the rated power of the computer room air conditioner into the random forest model to identify electricity theft behavior, and obtain the electricity theft behavior identification result of the communication site to be detected, specifically including: The historical electricity consumption data corresponding to the communication site to be detected, the target electricity consumption feature cluster, the real-time electricity consumption fluctuation prediction value, the daily electricity consumption fluctuation feature, and the rated power of the computer room air conditioner are input into several decision trees in the random forest model to identify electricity theft behavior and obtain the electricity theft behavior prediction label output by each decision tree. Each of the predicted electricity theft behavior labels is voted on, and the electricity theft behavior identification result of the communication site to be detected is determined based on the predicted electricity theft behavior label with the most votes.

[0077] As a preferred embodiment, the device further includes a random forest model construction module, used for: Based on the site information corresponding to several communication sites, the power consumption characteristics and the power consumption characteristic clusters, the gray prediction model is used to calculate the predicted power consumption fluctuation value of each communication site. Based on the historical electricity consumption data corresponding to several communication sites, the electricity theft behavior tags corresponding to the historical electricity consumption data, the daily electricity consumption fluctuation characteristics, the rated power of the computer room air conditioner, the electricity consumption feature clusters, and the predicted electricity consumption fluctuation values, a decision tree training data is constructed. Based on the decision tree training data, a number of decision trees are generated using the Bagging algorithm; The random forest model is constructed based on several decision trees.

[0078] As a preferred embodiment, the electricity theft detection module 12 is further used for: When the initial electricity theft behavior identification result is that electricity theft behavior exists, the initial electricity theft behavior identification result is used as the electricity theft behavior identification result of the communication station to be detected.

[0079] The electricity theft identification device 100 provided in this embodiment of the invention, in the process of electricity theft identification, considers the electricity consumption characteristics, site information and historical electricity consumption data of the communication site to be detected, and analyzes them in combination with electricity consumption feature clusters. Since the electricity consumption feature clusters are formed by clustering the electricity consumption characteristics of several communication sites, they can fully consider the differences in electricity consumption characteristics of different communication sites, thereby effectively and accurately identifying electricity theft behavior based on the electricity consumption characteristics of the communication site to be detected.

[0080] Please see Figure 4 The third aspect of the present invention provides an electronic device 200, including a memory 22, a processor 21, and a computer program stored in the memory 22 and executable on the processor 21. When the processor 21 executes the computer program, it implements the electricity theft behavior identification method described in any embodiment of the first aspect.

[0081] For example, the computer program may be divided into one or more modules / units, which are stored in the memory 22 and executed by the processor 21 to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device 200.

[0082] The electronic device 200 may include, but is not limited to, a processor 21 and a memory 22. Those skilled in the art will understand that the schematic diagram is merely an example of the electronic device 200 and does not constitute a limitation on the electronic device 200. It may include more or fewer components than illustrated, or combine certain components, or different components. For example, the electronic device 200 may also include input / output devices, network access devices, buses, etc.

[0083] The processor 21 can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor, or the processor 21 can be any conventional processor 21. The processor 21 is the control center of the electronic device 200, connecting various parts of the electronic device 200 via various interfaces and lines.

[0084] The memory 22 can be used to store the computer programs and / or modules. The processor 21 implements various functions of the electronic device 200 by running or executing the computer programs and / or modules stored in the memory 22 and calling the data stored in the memory 22. The memory 22 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory 22 may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0085] A fourth aspect of the present invention provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the electricity theft identification method described in any embodiment of the first aspect.

[0086] A fifth aspect of the present invention provides a computer program product, including a computer program / instructions, wherein when the computer program / instructions are executed by a processor, the steps of the electricity theft behavior identification method described in any embodiment of the first aspect are implemented.

[0087] Wherein, if the modules / units integrated in the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0088] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for identifying electricity theft, characterized in that, include: Based on the historical power consumption data of the communication site to be tested, the power consumption characteristics of the communication site to be tested are obtained; Based on the site information of the communication site to be detected, the electricity consumption characteristics, the historical electricity consumption data, and several preset electricity consumption characteristic clusters, the electricity theft behavior identification model is used to identify the electricity theft behavior of the communication site to be detected, and the electricity theft behavior identification result of the communication site to be detected is obtained. The electricity consumption feature cluster is formed by clustering the electricity consumption features of several communication sites; the electricity theft behavior identification model is composed of a gray prediction model and a random forest model.

2. The method for identifying electricity theft as described in claim 1, characterized in that, The method specifically obtains the electricity consumption characteristics through the following steps: Based on the historical electricity consumption data of any communication station, obtain the electricity consumption change characteristics for each preset time period; wherein, the electricity consumption change characteristics include the first-order differential characteristics of electricity consumption, the second-order differential characteristics of electricity consumption, and the electricity consumption fluctuation characteristics. Based on the historical electricity consumption data, obtain the daily electricity consumption of the computer room and the average daily electricity consumption of the computer room for any one of the communication sites within a preset statistical period; Based on the preset statistical period, the daily power consumption of the computer room, and the average daily power consumption of the computer room, the daily power consumption fluctuation characteristics of any one communication site are determined. Based on the electricity consumption change characteristics, the daily electricity consumption fluctuation characteristics, and the rated power of the air conditioner in the computer room of any one communication site, the electricity consumption characteristics of any one communication site are determined.

3. The method for identifying electricity theft as described in claim 2, characterized in that, The step of obtaining the electricity consumption change characteristics for each preset time period based on the historical electricity consumption data of any communication station specifically includes: Obtain the electricity consumption sequence for each preset time period from the historical electricity consumption data; Based on the electricity consumption sequence of two adjacent preset time periods, determine the first-order differential feature of electricity consumption for each preset time period; Based on the first-order differential characteristics of electricity consumption in two adjacent preset time periods, determine the second-order differential characteristics of electricity consumption in each preset time period. Based on a preset first time interval and a preset second time interval, calculate the first tilt angle corresponding to the line connecting the electricity consumption sequence of each preset time period and the electricity consumption sequence that follows it and is separated by the first time interval, and the second tilt angle corresponding to the line connecting the electricity consumption sequence of each preset time period and the electricity consumption sequence that follows it and is separated by the second time interval; wherein, the first time interval is less than the second time interval. The power consumption fluctuation characteristics for each preset time period are determined based on the difference between the first tilt angle and the second tilt angle.

4. The method for identifying electricity theft as described in claim 2, characterized in that, The step of determining the daily power consumption fluctuation characteristics of any communication site based on the preset statistical period, the daily power consumption of the data center, and the average daily power consumption of the data center specifically includes: Calculate the standard deviation of the daily power consumption of the data center based on the preset statistical period, the daily power consumption of the data center, and the average daily power consumption of the data center; The coefficient of variation of daily power consumption is obtained by comparing the standard deviation with the average daily power consumption of the computer room. Based on the preset statistical period, the daily power consumption of the computer room, the average daily power consumption of the computer room, and the standard deviation, the daily power consumption kurtosis and daily power consumption skewness are calculated respectively. The daily electricity consumption fluctuation characteristics are determined based on the daily electricity consumption variation coefficient, the daily electricity consumption kurtosis, and the daily electricity consumption skewness.

5. The method for identifying electricity theft as described in claim 2, characterized in that, The method specifically obtains the electricity consumption characteristic clusters through the following steps: The similarity of the electricity consumption change characteristics of each of the communication stations in each of the preset time periods is calculated to determine the feature similarity between the communication stations. Based on the feature similarity, the electricity consumption change features of each of the communication stations in each of the preset time periods are clustered to obtain several initial electricity consumption feature clusters. The average value of several electricity consumption change features in each preset time period in the initial electricity consumption feature cluster is calculated to determine the station electricity consumption features in each preset time period in the initial electricity consumption feature cluster. Based on the electricity consumption characteristics of several stations in each of the initial electricity consumption characteristic clusters, several electricity consumption characteristic clusters are generated.

6. The method for identifying electricity theft as described in claim 5, characterized in that, The step of identifying electricity theft behavior of the communication site under test based on the site information, electricity consumption characteristics, historical electricity consumption data, and several preset electricity consumption characteristic clusters, using an electricity theft behavior identification model, and obtaining the electricity theft behavior identification result of the communication site under test, specifically includes: Based on the gray prediction model, the similarity calculation is performed between the electricity consumption change characteristics of the communication site to be detected and the electricity consumption characteristics of each site in several electricity consumption feature clusters, so as to obtain the target electricity consumption feature cluster that matches the electricity consumption change characteristics from several electricity consumption feature clusters; Based on the site information, the electricity consumption change characteristics, and the target electricity consumption feature cluster, the gray prediction model is used to calculate the real-time electricity consumption jitter prediction value and the initial electricity theft behavior identification result of the communication site to be detected. When the initial electricity theft behavior identification result does not indicate the existence of electricity theft behavior, the historical electricity consumption data corresponding to the communication site to be detected, the target electricity consumption feature cluster, the real-time electricity consumption fluctuation prediction value, the daily electricity consumption fluctuation feature, and the rated power of the computer room air conditioner are input into the random forest model to identify electricity theft behavior and obtain the electricity theft behavior identification result of the communication site to be detected.

7. The method for identifying electricity theft as described in claim 6, characterized in that, The method specifically obtains the real-time electricity consumption fluctuation prediction value through the following steps: Based on the site information, obtain the number of service terminals added to the communication site to be detected, the power consumption calculation data caused by computing load, and the site air conditioning temperature; Obtain the station's electricity consumption fluctuation characteristics for the time period corresponding to the current prediction time from the target electricity consumption characteristic cluster; Based on the gray prediction model, and according to the number of new service terminals, the electricity consumption calculation data, the site air conditioning temperature, and the site electricity consumption fluctuation characteristics, the real-time electricity consumption fluctuation prediction value is calculated using the following expression: ； in, a , b , c and d These are the preset first fitting coefficient, second fitting coefficient, third fitting coefficient, and fourth fitting coefficient, respectively; This indicates the increase in the number of service terminals; This represents the electricity consumption calculation data; This indicates the air conditioning temperature at the site; This indicates the fluctuation characteristics of the power consumption at the site; This represents the preset prediction residual; This represents the predicted real-time power consumption fluctuation value.

8. The method for identifying electricity theft as described in claim 6, characterized in that, The method specifically obtains the initial electricity theft behavior identification result through the following steps: Based on the gray prediction model, the difference identifier of the first-order differential feature of the electricity consumption of the communication station to be detected in each preset time period is determined by multiplying the first-order differential feature of the electricity consumption of the communication station to be detected with the first-order differential feature of the electricity consumption of the station in the target electricity consumption feature cluster. The difference identifier of the second-order differential feature of the electricity consumption of the communication station to be detected in each preset time period is determined by multiplying the second-order differential feature of the electricity consumption of the station to be detected with the second-order differential feature of the electricity consumption of the station in the target electricity consumption feature cluster. Based on the first-order differential characteristic difference identifier of electricity consumption and the second-order differential characteristic difference identifier of electricity consumption, the total differential characteristic difference value of electricity consumption of the communication station to be detected is determined. The total difference value of the power consumption fluctuation feature of the communication station under test is determined based on the difference between the preset power consumption fluctuation feature threshold and the power consumption fluctuation feature of the communication station under test in each preset time period. When the total difference value of the differential feature of electricity consumption is less than the preset first differential feature difference threshold of electricity consumption, the total difference value of the fluctuation feature of electricity consumption is less than the preset fluctuation feature difference threshold of electricity consumption, and the predicted value of the real-time fluctuation of electricity consumption is less than the preset fluctuation threshold of electricity consumption, the initial electricity theft behavior identification result is determined to be that there is no electricity theft behavior. When the total difference value of the differential characteristics of electricity consumption is greater than the first differential characteristic difference threshold of electricity consumption and less than the preset second differential characteristic difference threshold of electricity consumption, or the total difference value of the differential characteristics of electricity consumption is less than the second differential characteristic difference threshold of electricity consumption and the total difference value of the fluctuation characteristics of electricity consumption is greater than the fluctuation characteristic difference threshold of electricity consumption, or the total difference value of the differential characteristics of electricity consumption is greater than the first differential characteristic difference threshold of electricity consumption and the predicted value of real-time fluctuation of electricity consumption is less than the fluctuation threshold of electricity consumption, or the total difference value of the fluctuation characteristics of electricity consumption is greater than the fluctuation characteristic difference threshold of electricity consumption and the predicted value of real-time fluctuation of electricity consumption is less than the fluctuation threshold of electricity consumption, the initial electricity theft behavior identification result is determined to be suspected electricity theft behavior; When the total difference value of the differential feature of electricity consumption is greater than the first differential feature difference threshold of electricity consumption, the total difference value of the fluctuation feature of electricity consumption is greater than the threshold of the fluctuation feature of electricity consumption, and the predicted value of the fluctuation of real-time electricity consumption is greater than the threshold of the fluctuation of electricity consumption, the initial electricity theft behavior identification result is determined to be that there is electricity theft behavior. Wherein, the second electricity consumption differential characteristic difference threshold is greater than the first electricity consumption differential characteristic difference threshold.

9. The method for identifying electricity theft as described in claim 6, characterized in that, The process of inputting the historical electricity consumption data corresponding to the communication site to be detected, the target electricity consumption feature cluster, the real-time electricity consumption fluctuation prediction value, the daily electricity consumption fluctuation characteristics, and the rated power of the computer room air conditioner into the random forest model to identify electricity theft behavior and obtain the electricity theft behavior identification result of the communication site to be detected, specifically includes: The historical electricity consumption data corresponding to the communication site to be detected, the target electricity consumption feature cluster, the real-time electricity consumption fluctuation prediction value, the daily electricity consumption fluctuation feature, and the rated power of the computer room air conditioner are input into several decision trees in the random forest model to identify electricity theft behavior and obtain the electricity theft behavior prediction label output by each decision tree. Each of the predicted electricity theft behavior labels is voted on, and the electricity theft behavior identification result of the communication site to be detected is determined based on the predicted electricity theft behavior label with the most votes.

10. The method for identifying electricity theft as described in claim 6, characterized in that, The method specifically generates the random forest model through the following steps: Based on the site information corresponding to several communication sites, the power consumption characteristics and the power consumption characteristic clusters, the gray prediction model is used to calculate the predicted power consumption fluctuation value of each communication site. Based on the historical electricity consumption data corresponding to several communication sites, the electricity theft behavior tags corresponding to the historical electricity consumption data, the daily electricity consumption fluctuation characteristics, the rated power of the computer room air conditioner, the electricity consumption feature clusters, and the predicted electricity consumption fluctuation values, a decision tree training data is constructed. Based on the decision tree training data, several decision trees are generated using the Bagging algorithm; The random forest model is constructed based on several decision trees.

11. The method for identifying electricity theft as described in claim 6, characterized in that, The method further includes: When the initial electricity theft behavior identification result is that electricity theft behavior exists, the initial electricity theft behavior identification result is used as the electricity theft behavior identification result of the communication station to be detected.

12. A device for identifying electricity theft, characterized in that, include: The electricity consumption feature acquisition module is used to acquire the electricity consumption features of the communication site to be detected based on the historical electricity consumption data of the communication site to be detected. The electricity theft behavior identification module is used to identify the electricity theft behavior of the communication site under test based on the site information, the electricity consumption characteristics, the historical electricity consumption data and several preset electricity consumption characteristic clusters, and to obtain the electricity theft behavior identification result of the communication site under test. The electricity consumption feature cluster is formed by clustering the electricity consumption features of several communication sites; the electricity theft behavior identification model is composed of a gray prediction model and a random forest model.

13. An electronic device, characterized in that, The device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the electricity theft identification method according to any one of claims 1 to 11.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the electricity theft identification method according to any one of claims 1 to 11.

15. A computer program product, characterized in that, It includes a computer program / instruction that, when executed by a processor, implements the steps of the electricity theft identification method according to any one of claims 1 to 11.

Images