Electricity stealing detection method and device based on neural network model

Abstract

The invention provides an electricity stealing detection method and device based on a neural network model, and is suitable for the technical field of power systems, and the method comprises the steps: dividing a to-be-detected region into a plurality of detection regions; acquiring historical power consumption data of each detection area, and training the neural network model to obtain a power consumption prediction model; for each detection area, obtaining the type of each electricity user and the proportion of different types of electricity users, determining the maximum allowable electricity utilization fluctuation value of the detection area based on the proportion, performing electricity utilization prediction on the detection area, and comparing predicted electricity utilization data with actual electricity utilization data to obtain an actual fluctuation value; and determining abnormal regions and a similar region of each abnormal region, screening the abnormal regions based on the similar regions, and determining an electricity stealing region. According to the invention, power consumption prediction is carried out through the power consumption prediction model, the abnormal region is determined, and the electricity stealing region is determined through the similar region, so that the accuracy of the electricity stealing region is ensured, and the investigation range of electricity stealing is effectively reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of power systems, and in particular relates to a method and device for detecting electricity theft based on a neural network model. Background Art

[0002] The act of stealing electricity is extremely harmful to society and will affect the stable supply of the entire society. More serious thefts may also damage power equipment in the process. The damage to power equipment will affect the society's power safety and people's lives.

[0003] Most existing methods for detecting electricity theft analyze the operating status of the power grid. When electricity theft is detected, manual on-site verification is performed to confirm the user who has committed the theft. In practice, the inventors found that this method cannot accurately take into account the influence of multiple factors. Most of them rely on manual experience to determine whether electricity theft has occurred, which has low accuracy. In addition, the detection range of existing electricity theft detection methods is large. When electricity theft occurs, it is not easy to locate the specific location of the theft. Summary of the invention

[0004] The embodiments of the present invention provide a method and device for detecting electricity theft based on a neural network model, so as to solve the problem that the scope of electricity theft detection in the existing electricity theft detection method is large and it is not easy to locate the specific electricity theft location.

[0005] The present invention is achieved through the following technical solutions:

[0006] In a first aspect, an embodiment of the present invention provides a method for detecting electricity theft based on a neural network model, comprising:

[0007] Divide the area to be tested into multiple testing areas;

[0008] Obtaining historical power consumption data of each detection area, and training a neural network model based on the historical power consumption data to obtain a power consumption prediction model for each detection area;

[0009] For each detection area, obtaining the type of each electricity user in the detection area and the ratio of different types of electricity users in the detection area, and determining the maximum allowable electricity consumption fluctuation value of the detection area based on the ratio;

[0010] For each detection area, the power consumption of the detection area is predicted based on the power consumption prediction model, and the predicted power consumption data of the detection area is compared with the actual power consumption data to obtain the actual fluctuation value;

[0011] The abnormal area is determined based on all actual fluctuation values, the similar area of ​​each abnormal area is determined, and the corresponding abnormal area is screened based on the actual power consumption data of the similar area to determine the power theft area.

[0012] In a possible implementation, the area to be detected is divided into multiple detection areas, including:

[0013] Acquire the number of stations and coverage of the stations in the area to be detected, and divide the area to be detected based on the number of stations and coverage of the stations to obtain multiple detection areas;

[0014] Each detection area includes one or more stations.

[0015] In a possible implementation, the acquiring of historical power consumption data of each detection area and training of a neural network model based on the historical power consumption data to obtain a power consumption prediction model for each detection area includes:

[0016] Constructing a neural network model, obtaining historical electricity consumption data of each detection area, obtaining a training set and a validation set based on the historical electricity consumption data, and training the neural network model based on the training set to obtain an initial electricity consumption prediction model;

[0017] The initial power consumption prediction model is verified based on the verification set using a time series cross-validation method, and the power consumption prediction model of the corresponding detection area is determined based on the verification result.

[0018] In a possible implementation, for each detection area, obtaining the type of each electricity user in the detection area and the ratio of different types of electricity users in the detection area, and determining the maximum allowable electricity consumption fluctuation value of the detection area based on the ratio includes:

[0019] For each detection area, perform the following steps:

[0020] Obtain the type of each electricity user in the detection area;

[0021] Obtaining the proportion of each type of electricity users and the historical electricity consumption data of each type of electricity users in a preset historical period;

[0022] Based on the historical electricity consumption data, determining the electricity consumption fluctuation value and the total electricity consumption of each type of electricity user;

[0023] Substituting the power consumption fluctuation value and the total power consumption of each type of power user in the detection area into the power consumption fluctuation prediction formula of the detection area, and calculating the maximum allowable power consumption fluctuation value of the detection area;

[0024] The power consumption fluctuation prediction formula is:

[0025]

[0026] Among them, ki represents the maximum allowable power fluctuation value of the i-th detection area, α w represents the temperature weight of the i-th detection area, α m represents the economic weight of the i-th detection area, ΔW i represents the temperature fluctuation coefficient of the i-th detection area, Δm i represents the economic fluctuation coefficient of the i-th detection area, E i represents the total power consumption in the historical power consumption data of the i-th detection area, p j represents the proportion of electricity users of the jth type in the i-th detection area, β j represents the influence factor of the first influencing factor of the jth type in the i-th detection area, γ j represents the impact factor of the second impact factor of the jth type in the i-th detection area, represents the first impact value of the jth type in the i-th detection area, ΔE(2) j Represents the second impact value of the j-th type in the i-th detection area.

[0027] In a possible implementation, for each detection area, predicting the power consumption of the detection area based on the power consumption prediction model, and comparing the predicted power consumption data of the detection area with the actual power consumption data to obtain the actual fluctuation value include:

[0028] For each detection area, perform the following steps:

[0029] Based on the historical electricity consumption data of the detection area and the electricity consumption prediction model, the electricity consumption of the detection area is predicted to obtain the predicted electricity consumption data within a preset time period;

[0030] Acquire actual power consumption data within a preset period, compare the predicted power consumption data with the actual power consumption data, and obtain an actual fluctuation value based on the comparison result;

[0031] The historical power consumption data and the predicted power consumption data both include the power consumption of the detection area in different time sub-segments and the line loss data of the detection area.

[0032] In a possible implementation, comparing the preset power usage data with the actual power usage data and obtaining the actual fluctuation value based on the comparison result includes:

[0033] For each detection area, the preset power consumption data and the power consumption in different time sub-segments in the actual power consumption data and the line loss data of the detection area are input into the actual fluctuation value calculation formula to obtain the actual fluctuation value;

[0034] The actual fluctuation value calculation formula is:

[0035]

[0036] Among them, k i ′ is the actual fluctuation value of the i-th detection area, t represents the t-th time segment, s(1) it represents the actual power consumption of the i-th detection area in the t-th time period, s(2) it represents the predicted power consumption of the i-th detection area in the t-th time period, δ1 represents the line loss correction coefficient of the i-th detection area in the t-th time period, and x(1) it represents the actual line loss of the i-th detection area in the t-th time period, x(2) it It represents the preset line loss of the i-th detection area in the t-th time period.

[0037] In a possible implementation, determining the abnormal area based on all actual fluctuation values ​​includes:

[0038] Compare the actual fluctuation value of each detection area with the corresponding maximum allowable power fluctuation value, and if the actual fluctuation value exceeds the maximum allowable power fluctuation value, determine the detection area as an abnormal area;

[0039] Otherwise, the detected area is determined as a normal area.

[0040] In a possible implementation, obtaining similar areas of each abnormal area, and screening corresponding abnormal areas based on actual power consumption data of the similar areas to determine the power theft area includes:

[0041] Acquire feature data of each detection area, calculate similarity between each abnormal area and other detection areas based on the feature data, and determine similar areas of each abnormal area based on the similarity;

[0042] For each anomaly area, perform the following steps:

[0043] Acquire actual power consumption data of a similar area to the abnormal area, and compare the actual power consumption data of the abnormal area with the actual power consumption data of the corresponding similar area;

[0044] The difference between the actual power consumption data of the abnormal area and the actual power consumption data of the corresponding similar area is calculated, and each difference is compared with a corresponding threshold value, and it is determined whether the abnormal area is a power theft area based on the comparison result.

[0045] In a possible implementation, the method further includes:

[0046] Based on the electricity theft detection model, each electricity user in each electricity theft area is monitored to identify suspicious electricity users who may be involved in electricity theft;

[0047] The user ID of each suspicious electricity user is sent to the electricity theft alarm system, and staff are notified to conduct on-site verification.

[0048] In a second aspect, an embodiment of the present invention provides an electricity theft detection device based on a neural network model, comprising:

[0049] A division module is used to divide the area to be detected into multiple detection areas;

[0050] A training module is used to obtain historical power consumption data of each detection area, and train a neural network model based on the historical power consumption data to obtain a power consumption prediction model for each detection area;

[0051] A determination module, configured to obtain, for each detection area, the type of each electricity user in the detection area and the ratio of different types of electricity users in the detection area, and determine the maximum allowable electricity consumption fluctuation value of the detection area based on the ratio;

[0052] A comparison module is used to predict the power consumption of each detection area based on the power consumption prediction model, and compare the predicted power consumption data of the detection area with the actual power consumption data to obtain an actual fluctuation value;

[0053] The screening module is used to determine the abnormal area based on all actual fluctuation values, determine the similar area of ​​each abnormal area, and screen the corresponding abnormal area based on the actual power consumption data of the similar area to determine the power theft area.

[0054] The embodiment of the present invention provides a method and device for detecting electricity theft based on a neural network model, which divides the area to be detected into multiple detection areas, uses the historical electricity consumption data of each detection area to train the neural network model to obtain an electricity consumption prediction model, performs electricity consumption prediction to obtain predicted electricity consumption data, compares the predicted electricity consumption data with the actual electricity consumption data to obtain a fluctuation value, determines the proportion of different types of electricity users in each detection area, and determines the maximum allowable electricity consumption fluctuation value, compares with the actual fluctuation value to determine the abnormal area, and verifies and determines the electricity theft area based on similar areas. The maximum allowable electricity consumption fluctuation value is determined by proportion, and the accuracy of the electricity theft area is guaranteed by verification through similar areas. By determining the electricity theft area, the scope of electricity theft investigation is effectively narrowed. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or related technical descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0056] Figure 1 It is a flowchart of a method for detecting electricity theft based on a neural network model provided by one embodiment of the present invention;

[0057] Figure 2 It is an initial area division diagram of a method for detecting electricity theft based on a neural network model provided by an embodiment of the present invention;

[0058] Figure 3 This is a regional division diagram of a method for detecting electricity theft based on a neural network model provided by an embodiment of the present invention;

[0059] Figure 4 It is a structural schematic diagram of a power theft detection device based on a neural network model provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0060] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present invention. However, it should be clear to those skilled in the art that the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present invention.

[0061] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0062] It should also be understood that the term "and / or" used in the present description and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0063] As used in the present specification and the appended claims, the term "if" may be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" may be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]", depending on the context.

[0064] In addition, in the description of the present specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0065] References to "one embodiment" or "some embodiments" etc. described in the present specification mean that one or more embodiments of the present invention include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0066] Figure 1 is a flowchart of a method for detecting electricity theft based on a neural network model provided by an embodiment of the present invention, referring to Figure 1 , the detailed description of the electricity theft detection method based on the neural network model is as follows:

[0067] S110, dividing the area to be detected into multiple detection areas.

[0068] In an optional embodiment, the area to be detected is divided into multiple detection areas, including:

[0069] Obtain the number of stations and the coverage of the stations in the area to be detected, and divide the area to be detected based on the number of stations and the coverage of the stations to obtain multiple detection areas;

[0070] Each detection area includes one or more stations.

[0071] The specific division method is pre-set, such as Figure 2 As shown, if there are 20 substations in the area A to be detected, and each substation covers 50-80 electricity users, then in order to facilitate monitoring, the area A to be detected can be divided into 5 initial areas according to the area, and then the 5 initial areas are adjusted according to the coverage of the substations in the 5 initial areas. For example, the initial area A1 contains 3 complete substations (substations 11, 12, and 13) and some users in 4 substations (substations 14, 15, 16, and 17). When dividing, it can be obtained by how many households in the 4 incomplete substations are within the scope of the initial area 1. When the number of electricity users in a certain incomplete substation within the scope of the initial area 1 exceeds 60% of the total number of electricity users covered by this substation, this substation is divided into the initial area 1, otherwise, this substation is divided into other detection areas.

[0072] For example, one of the four incomplete substations (15) covers a total of 80 electricity users, 67 of which are within the scope of initial area 1, and only 13 are not within the scope of initial area 1. However, the number of electricity users in the initial area 1 of the other substations (substations 14, 16, and 17) does not exceed 60% of the total number of electricity users covered by the corresponding substations. Figure 3 As shown, the initial area 1 is adjusted so that the initial area 1 does not include the other three incomplete areas, and the remaining 13 electricity users of this area (15) are divided into the range of the initial area 1 to obtain the detection area 1. The detection area can also be divided in other ways, and the specific division method is not limited here, but for the convenience of monitoring, each detection area after the division contains a complete area.

[0073] The area to be inspected is divided by the number of substations and the coverage of the substations, and multiple inspection areas are obtained through division, which effectively reduces the possible scope of electricity theft.

[0074] S120, obtaining historical electricity consumption data of each detection area, and training a neural network model based on the historical electricity consumption data to obtain an electricity consumption prediction model for each detection area.

[0075] Optionally, the historical electricity consumption data includes the total electricity consumption of each electricity user within the detection range, electricity consumption in different time periods, harmonic distortion rate, electricity consumption records of solar power generation, abnormal electricity consumption status, and current, voltage and power consumption data at each historical moment.

[0076] When training the neural network model, in addition to the historical electricity consumption data, it is also necessary to detect the weather data, economic data and activity data corresponding to each historical electricity consumption data in the area.

[0077] In this embodiment, the power consumption prediction model is used to predict future power consumption data of the corresponding detection area.

[0078] In an optional embodiment, historical power consumption data of each detection area is obtained, and a neural network model is trained based on the historical power consumption data to obtain a power consumption prediction model for each detection area, including:

[0079] Construct a neural network model, obtain the historical electricity consumption data of each detection area, obtain a training set and a validation set based on the historical electricity consumption data, and train the neural network model based on the training set to obtain an initial electricity consumption prediction model;

[0080] The initial electricity consumption prediction model is verified using the time series cross-validation method based on the validation set, and the electricity consumption prediction model for the corresponding detection area is determined based on the validation results.

[0081] Among them, the training set and the validation set are divided in chronological order. For example, the historical electricity consumption data includes electricity consumption data from January 1, 2022 to December 31, 2022. The electricity consumption data from January 1, 2022 to November 1, 2022 can be used as the training set, and the remaining electricity consumption data can be used as the validation set.

[0082] The time series cross-validation method is used to divide the historical electricity consumption data into a training set and a validation set. After the neural network model is trained in the training set to obtain the initial electricity consumption prediction model, the initial electricity consumption prediction model is verified by the validation set, which can ensure the accuracy of the electricity consumption prediction model.

[0083] S130: For each detection area, obtain the type of each electricity user in the detection area and the ratio of different types of electricity users in the detection area, and determine the maximum allowable electricity consumption fluctuation value of the detection area based on the ratio.

[0084] In this embodiment, the types of electricity users can be urban residential electricity use, industrial electricity use, agricultural electricity use, commercial electricity use, rural residential electricity use, school electricity use, etc., and can also be divided more finely. The finer the division, the more accurate the maximum allowable electricity fluctuation value obtained.

[0085] For example, there are 360 ​​electricity users in detection area 1, of which 200 are urban residents, 30 are schools, 50 are commercial, and 80 are industrial. The proportion of urban residents' electricity consumption is 55.6%, the proportion of schools' electricity consumption is 8.3%, the proportion of commercial electricity consumption is 13.9%, and the proportion of industrial electricity consumption is 22.2%.

[0086] In an optional embodiment, for each detection area, obtaining the type of each electricity user in the detection area and the ratio of different types of electricity users in the detection area, and determining the maximum allowable electricity consumption fluctuation value of the detection area based on the ratio, includes:

[0087] For each detection area, perform the following steps:

[0088] Obtain the type of each electricity user in the detection area;

[0089] Obtain the proportion of each type of electricity users and the historical electricity consumption data of each type of electricity users in a preset historical period;

[0090] Based on historical electricity consumption data, determine the total electricity consumption of each type of electricity user;

[0091] Substituting the total power consumption of each type of power user in the detection area into the power consumption fluctuation prediction formula of the detection area, and calculating the maximum allowable power consumption fluctuation value of the detection area;

[0092] The power consumption fluctuation prediction formula is:

[0093]

[0094] Among them, k i represents the maximum allowable power fluctuation value of the i-th detection area, α w represents the temperature weight of the i-th detection area, α m represents the economic weight of the i-th detection area, △W i represents the temperature fluctuation coefficient of the i-th detection area, Δm i represents the economic fluctuation coefficient of the i-th detection area, E i represents the total power consumption in the historical power consumption data of the i-th detection area, p j represents the proportion of the ,th type of electricity users in the ith detection area, β j represents the influence factor of the first influencing factor of the jth type in the i-th detection area, γ j represents the impact factor of the second influencing factor of the ith type in the ith detection area, represents the first influencing value of the jth type in the ith detection area, ΔE(2) j Represents the second impact value of the j-th type in the i-th detection area.

[0095] In this embodiment, the preset historical time period refers to the time period during which electricity theft detection is required. For example, if the electricity consumption of detection area 1 from August 1, 2023 to August 7, 2023 is to be detected, the preset historical time period may be August 1, 2022 to August 7, 2022 and August 1, 2021 to August 7, 2021. It is necessary to obtain the historical electricity consumption data of each type of electricity user from August 1, 2022 to August 7, 2022 and from August 1, 2021 to August 7, 2021.

[0096] Among them, the first influencing factor and the second influencing factor represent events that will affect the electricity consumption data, which can be natural disasters, holidays, large-scale events and other influencing factors. Each type of electricity user can have 0-2 influencing factors. For example, the influencing factors of schools in detection area 1 can be holidays and large-scale events. For another example, urban residents in detection area 1 have not experienced anything that will affect electricity consumption in a certain period of time, so there is no influencing factor.

[0097] The first impact value and the second impact value represent the values ​​that the influencing factors may have on electricity consumption. The possible impact value is obtained by predicting the impact on electricity consumption when such impact occurred in history. For example, there are three schools in detection area 1, among which school A is going to hold activity D, school B and school C do not hold activities, and school A has held activity D many times before. Then, the impact value of detection area 1 can be determined based on the historical data of school A holding activity D.

[0098] The impact factor refers to the proportion of the average electricity consumption of the part affected by the factors to the total electricity consumption of this type of electricity users. For example, if the electricity consumption of school A is X% on a normal day, then the impact factor of school A is X%.

[0099] By classifying each electricity user into different types and determining the proportion of different types of electricity users in the detection area, the maximum allowable electricity fluctuation value is further determined by the proportion. By taking into account the characteristics of different electricity users, the accuracy of the maximum allowable electricity fluctuation value is effectively improved.

[0100] S140: For each detection area, a power consumption prediction is performed on the detection area based on the power consumption prediction model, and the predicted power consumption data of the detection area is compared with the actual power consumption data to obtain an actual fluctuation value.

[0101] Among them, for each detection area, the data of the detection area is input into the power consumption prediction model, so that the power consumption prediction of the detection area can be realized and the predicted power consumption data of the detection area can be obtained.

[0102] The actual power consumption data refers to the power consumption data recorded in the detection area within a certain period of time.

[0103] For example, to obtain the actual fluctuation value from Monday to Sunday in region A last week, we can first obtain the prediction result of the power consumption prediction model for region A as the predicted power consumption data of region A, and then collect the actual power consumption data of region A in this time period. The actual fluctuation value can be obtained by comparing the predicted power consumption data with the actual power consumption data.

[0104] In an optional embodiment, for each detection area, power consumption prediction is performed for the detection area based on the power consumption prediction model, and the predicted power consumption data of the detection area is compared with the actual power consumption data to obtain an actual fluctuation value, including:

[0105] For each detection area, perform the following steps:

[0106] Based on the historical electricity consumption data of the detection area and the electricity consumption prediction model, the electricity consumption of the detection area is predicted to obtain the predicted electricity consumption data within a preset time period;

[0107] Acquire actual power consumption data within a preset period, compare the predicted power consumption data with the actual power consumption data, and obtain an actual fluctuation value based on the comparison result;

[0108] The historical power consumption data and the predicted power consumption data both include the power consumption of the detection area in different time sub-segments and the line loss data of the detection area.

[0109] In an optional embodiment, the preset power consumption data and the actual power consumption data are compared, and the actual fluctuation value is obtained based on the comparison result, including:

[0110] For each detection area, the preset power consumption data and the power consumption in different time sub-segments in the actual power consumption data and the line loss data of the detection area are input into the actual fluctuation value calculation formula to obtain the actual fluctuation value;

[0111] The actual fluctuation value calculation formula is:

[0112]

[0113] Among them, k i ′ is the actual fluctuation value of the i-th detection area, t represents the t-th time segment, s(1) it represents the actual power consumption of the i-th detection area in the t-th time period, s(2) it represents the predicted power consumption of the i-th detection area in the t-th time period, δ1 represents the line loss correction coefficient of the i-th detection area in the t-th time period, and x(1) it represents the actual line loss of the i-th detection area in the t-th time period, x(2) it It represents the preset line loss of the i-th detection area in the t-th time period.

[0114] In this embodiment, the preset time period is a past time period. For example, if it is August 8, 2023 now, the preset time period can only be the time period before August 8, 2023, and the time period does not include August 1, 2023, and can be August 1, 2023-August 7, 2023.

[0115] Optionally, the actual fluctuation value is a fluctuation value of power consumption, which is obtained through the fluctuation value of power consumption and line loss data.

[0116] The division of time sub-segments is explained by an example. If the preset time period is from August 1, 2023 to August 7, 2023, the first time sub-segment is from 0:00 to 1:00 in the morning from August 1, 2023 to August 7, 2023, the second time sub-segment is from 1:00 to 2:00 in the morning from August 1, 2023 to August 7, 2023, and the third time sub-segment is from 2:00 to 3:00 in the morning from August 1, 2023 to August 7, 2023, and so on, 24 time sub-segments can be obtained. And the right endpoint of each time period is included in this time period.

[0117] By dividing the time segment into time sub-segments, the power consumption and line loss data of each time sub-segment are obtained, and the fluctuation value of each time field is calculated. The fluctuation values ​​of each time sub-segment are superimposed to obtain the actual fluctuation value. The different power consumption patterns in each time period are fully considered, and the fluctuation value of power consumption is corrected by the difference in line loss, ensuring the accuracy of the actual fluctuation value.

[0118] S150, determining an abnormal area based on all actual fluctuation values, determining a similar area of ​​each abnormal area, and screening the corresponding abnormal area based on actual power consumption data of the similar area to determine a power theft area.

[0119] The abnormal area refers to the detection area where the actual fluctuation value is greater than the maximum allowable power fluctuation value.

[0120] In this embodiment, the similarity of the abnormal area is calculated by calculating the similarity between each area and the abnormal area through the number of substations, area, proportion of different types of electricity users, geographical location relationship with the abnormal area, average daily electricity consumption in historical electricity consumption data in a certain period of time in the past, and other data, and arranging the similarities from high to low, and selecting the one with the highest similarity according to the arrangement result. For example, the abnormal area is detection area 1, and the remaining areas are detection areas 2 and 3. The similarity between detection area 1 and detection area 2 and the similarity between detection area 1 and detection area 3 are calculated. If the similarity between detection area 1 and detection area 2 is 78%, and the similarity between detection area 1 and detection area 3 is 80%, then detection area 3 is selected as the similarity of detection area 1. When selecting similar areas, a threshold can also be set, and only areas exceeding the threshold can be similar areas of the abnormal area. Similar areas can also be selected in other ways, which are not limited in this scheme.

[0121] In an optional embodiment, determining the abnormal area based on all actual fluctuation values ​​includes:

[0122] Compare the actual fluctuation value of each detection area with the corresponding maximum allowable power fluctuation value, and if the actual fluctuation value exceeds the maximum allowable power fluctuation value, determine the detection area as an abnormal area;

[0123] Otherwise, the detected area is determined as a normal area.

[0124] In an optional embodiment, obtaining similar areas of each abnormal area, and screening corresponding abnormal areas based on actual power consumption data of similar areas to determine the power theft area includes:

[0125] Acquire feature data of each detection area, calculate similarity between each abnormal area and other detection areas based on the feature data, and determine similar areas of each abnormal area based on the similarity;

[0126] For each anomaly area, perform the following steps:

[0127] Acquire actual power consumption data of a similar area to the abnormal area, and compare the actual power consumption data of the abnormal area with the actual power consumption data of the corresponding similar area;

[0128] The difference between the actual power consumption data of the abnormal area and the actual power consumption data of the corresponding similar area is calculated, and each difference is compared with a corresponding threshold value, and it is determined whether the abnormal area is a power theft area based on the comparison result.

[0129] Among them, the characteristic data of the detection area include the number of substations in the area, the area, the proportion of different types of electricity users, the geographical location relationship between the abnormal area, the average daily electricity consumption in the historical electricity consumption data in a certain period of time in the past, and other data, which are not limited here.

[0130] In this embodiment, the actual electricity consumption data of the abnormal area and the actual electricity consumption data of the similar area can include: total electricity consumption, electricity consumption in different time periods, harmonic distortion rate, electricity records of solar power generation, abnormal electricity consumption status, and current, voltage and power consumption at each historical moment, and may also include other data, which is not limited in this scheme.

[0131] Optionally, an example is used to illustrate the specific scheme of this embodiment: for abnormal area N, if the abnormal area is determined using data such as power consumption, peak-valley electricity prices, power efficiency, and voltage stability, and when two or three differences exceed the threshold, the abnormal area is used as a power theft area. It is necessary to calculate the difference h1 between the power consumption in the actual power consumption data of the abnormal area N and the power consumption in the actual power consumption data of the corresponding similar area, calculate the difference h2 between the peak-valley electricity price in the actual power consumption data of the abnormal area N and the peak-valley electricity price in the actual power consumption data of the corresponding similar area, and calculate the difference h3 between the actual power consumption data of the abnormal area N and the actual power consumption data of the corresponding similar area, and obtain three differences h1, h2, and h3. If h1 and h2 exceed the threshold, the abnormal area N is used as a power theft area. If only h1 exceeds the threshold, the abnormal area N is corrected to a normal area, and the specific correction method is not limited here.

[0132] By comparing the actual fluctuation value of the detection area with the corresponding maximum allowable power consumption fluctuation value, the abnormal area is determined, and the similar area of ​​each abnormal area is calculated through the characteristic data of each detection area, and the actual power consumption data of the abnormal area is compared with the actual power consumption data of the corresponding similar area, and the power theft area is determined according to the comparison result. In actual situations, there are power thieves in both the abnormal area and its similar area, and the situation where the power theft amount of the power thieves is very small is extremely rare and can be almost ignored. Therefore, by correcting the abnormal area through the similar area, the accuracy of the detection results of this scheme can be guaranteed when some emergencies are not taken into account in the calculation of power consumption fluctuations.

[0133] In an optional embodiment, the method further comprises:

[0134] Based on the electricity theft detection model, each electricity user in each electricity theft area is monitored to identify suspicious electricity users who may be involved in electricity theft;

[0135] The user ID of each suspicious electricity user is sent to the electricity theft alarm system, and staff are notified to conduct on-site verification.

[0136] By dividing the area to be detected into multiple detection areas, and obtaining the historical electricity consumption data of each detection area, the neural network model is trained according to these historical electricity consumption data to obtain the electricity consumption prediction model of each detection area, and then the proportion of different types of electricity users in each detection area is obtained, and the maximum allowable electricity consumption fluctuation value is determined according to the proportion, and then the predicted electricity consumption data is compared with the actual electricity consumption data to obtain the actual fluctuation value, and finally, according to the actual electricity consumption data of similar areas of each abnormal area, the abnormal area is excluded to determine the electricity theft area, and each electricity user in the electricity theft area is tested for electricity theft, and the electricity users who can use electricity are determined and the staff are notified to conduct on-site verification. This scheme can effectively determine the area where electricity theft may exist, without the need to detect the entire area, and in the judgment process, the fluctuation effect brought by different types of electricity users is taken into account, ensuring the accuracy of abnormal area judgment, and the abnormal area is excluded according to the similar area to obtain the electricity theft area, eliminating the influence of other factors that have not been taken into account.

[0137] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0138] Corresponding to the above embodiment, a method for detecting electricity theft based on a neural network model is described. Figure 4 A schematic structural diagram of a power theft detection device based on a neural network model provided by an embodiment of the present invention is shown. For ease of description, only the parts related to the embodiment of the present invention are shown.

[0139] See also Figure 4 In an embodiment of the present invention, a power theft detection device 4 based on a neural network model may include:

[0140] A division module 41 is used to divide the area to be detected into multiple detection areas;

[0141] A training module 42 is used to obtain historical power consumption data of each detection area, and train a neural network model based on the historical power consumption data to obtain a power consumption prediction model for each detection area;

[0142] A determination module 43 is used to obtain, for each detection area, the type of each electricity user in the detection area and the ratio of different types of electricity users in the detection area, and determine the maximum allowable electricity consumption fluctuation value of the detection area based on the ratio;

[0143] A comparison module 44 is used to predict the power consumption of each detection area based on the power consumption prediction model, and compare the predicted power consumption data of the detection area with the actual power consumption data to obtain an actual fluctuation value;

[0144] The screening module 45 is used to determine the abnormal area based on all the actual fluctuation values, determine the similar area of ​​each abnormal area, and screen the corresponding abnormal area based on the actual power consumption data of the similar area to determine the power theft area.

[0145] In some embodiments, the partitioning module 41 is specifically configured to:

[0146] Obtain the number of stations and the coverage of the stations in the area to be detected, and divide the area to be detected based on the number of stations and the coverage of the stations to obtain multiple detection areas;

[0147] Each detection area includes one or more stations.

[0148] In some embodiments, the training module 42 is specifically configured to:

[0149] Construct a neural network model, obtain the historical electricity consumption data of each detection area, obtain a training set and a validation set based on the historical electricity consumption data, and train the neural network model based on the training set to obtain an initial electricity consumption prediction model;

[0150] The initial electricity consumption prediction model is verified using the time series cross-validation method based on the validation set, and the electricity consumption prediction model for the corresponding detection area is determined based on the validation results.

[0151] In some embodiments, the determination module 43 is specifically configured to:

[0152] For each detection area, perform the following steps:

[0153] Obtain the type of each electricity user in the detection area;

[0154] Obtain the proportion of each type of electricity users and the historical electricity consumption data of each type of electricity users in a preset historical period;

[0155] Based on historical electricity consumption data, determine the total electricity consumption of each type of electricity user;

[0156] Substituting the total power consumption of each type of power user in the detection area into the power consumption fluctuation prediction formula of the detection area, and calculating the maximum allowable power consumption fluctuation value of the detection area;

[0157] The power consumption fluctuation prediction formula is:

[0158]

[0159] Among them, k i represents the maximum allowable power fluctuation value of the i-th detection area, α w represents the temperature weight of the i-th detection area, α m represents the economic weight of the i-th detection area, ΔWi represents the temperature fluctuation coefficient of the i-th detection area, Δm i represents the economic fluctuation coefficient of the i-th detection area, E i represents the total power consumption in the historical power consumption data of the i-th detection area, p j represents the proportion of electricity users of the jth type in the i-th detection area, β j represents the influence factor of the first influencing factor of the jth type in the i-th detection area, γ j represents the impact factor of the second impact factor of the jth type in the i-th detection area, represents the first impact value of the jth type in the i-th detection area, ΔE(2) j Represents the second impact value of the j-th type in the i-th detection area.

[0160] In some embodiments, the comparison module 44 is specifically configured to:

[0161] For each detection area, perform the following steps:

[0162] Based on the historical electricity consumption data of the detection area and the electricity consumption prediction model, the electricity consumption of the detection area is predicted to obtain the predicted electricity consumption data within a preset time period;

[0163] Acquire actual power consumption data within a preset period, compare the predicted power consumption data with the actual power consumption data, and obtain an actual fluctuation value based on the comparison result;

[0164] The historical power consumption data and the predicted power consumption data both include the power consumption of the detection area in different time sub-segments and the line loss data of the detection area.

[0165] In some embodiments, the comparison module 44 is further configured to:

[0166] For each detection area, the preset power consumption data and the power consumption in different time sub-segments in the actual power consumption data and the line loss data of the detection area are input into the actual fluctuation value calculation formula to obtain the actual fluctuation value;

[0167] The actual fluctuation value calculation formula is:

[0168]

[0169] Among them, k i ′ is the actual fluctuation value of the i-th detection area, t represents the t-th time segment, s(1) it represents the actual power consumption of the i-th detection area in the t-th time period, s(2) it represents the predicted power consumption of the i-th detection area in the t-th time period, δ1 represents the line loss correction coefficient of the i-th detection area in the t-th time period, and x(1) itrepresents the actual line loss of the i-th detection area in the t-th time period, x(2) it It represents the preset line loss of the i-th detection area in the t-th time period.

[0170] In some embodiments, the screening module 45 is specifically configured to:

[0171] Compare the actual fluctuation value of each detection area with the corresponding maximum allowable power fluctuation value, and if the actual fluctuation value exceeds the maximum allowable power fluctuation value, determine the detection area as an abnormal area;

[0172] Otherwise, the detected area is determined as a normal area.

[0173] In some embodiments, the screening module 45 is further configured to:

[0174] Acquire feature data of each detection area, calculate similarity between each abnormal area and other detection areas based on the feature data, and determine similar areas of each abnormal area based on the similarity;

[0175] For each anomaly area, perform the following steps:

[0176] Acquire actual power consumption data of a similar area to the abnormal area, and compare the actual power consumption data of the abnormal area with the actual power consumption data of the corresponding similar area;

[0177] The difference between the actual power consumption data of the abnormal area and the actual power consumption data of the corresponding similar area is calculated, and each difference is compared with a corresponding threshold value, and it is determined whether the abnormal area is a power theft area based on the comparison result.

[0178] In some embodiments, a power theft detection device 4 based on a neural network model can also be used for:

[0179] Based on the electricity theft detection model, each electricity user in each electricity theft area is monitored to identify suspicious electricity users who may be involved in electricity theft;

[0180] The user ID of each suspicious electricity user is sent to the electricity theft alarm system, and staff are notified to conduct on-site verification.

[0181] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0182] Those of ordinary skill in the art will appreciate that the templates, units, and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0183] If the module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, it can implement the steps of each of the above-mentioned embodiments of the method for detecting electricity theft based on the neural network model. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory, random access memory, electrical carrier signal, telecommunication signal and software distribution medium.

[0184] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for detecting electricity theft based on a neural network model, characterized in that: include: Divide the area to be tested into multiple testing areas; Obtaining historical electricity consumption data for each detection area, and training a neural network model based on the historical electricity consumption data to obtain an electricity consumption prediction model for each detection area; For each detection area, obtaining the type of each electricity user in the detection area and the ratio of different types of electricity users in the detection area, and determining the maximum allowable electricity consumption fluctuation value of the detection area based on the ratio; For each detection area, the power consumption of the detection area is predicted based on the power consumption prediction model, and the predicted power consumption data of the detection area is compared with the actual power consumption data to obtain the actual fluctuation value; The abnormal area is determined based on all actual fluctuation values, the similar area of ​​each abnormal area is determined, and the corresponding abnormal area is screened based on the actual power consumption data of the similar area to determine the power theft area.

2. The method for detecting electricity theft based on a neural network model as claimed in claim 1, characterized in that: The area to be tested is divided into multiple testing areas, including: Acquire the number of stations and coverage of the stations in the area to be detected, and divide the area to be detected based on the number of stations and coverage of the stations to obtain multiple detection areas; Each detection area includes one or more stations.

3. The method for detecting electricity theft based on a neural network model as claimed in claim 1, characterized in that: The acquiring of historical power consumption data of each detection area and training of a neural network model based on the historical power consumption data to obtain a power consumption prediction model for each detection area includes: Constructing a neural network model, obtaining historical electricity consumption data of each detection area, obtaining a training set and a validation set based on the historical electricity consumption data, and training the neural network model based on the training set to obtain an initial electricity consumption prediction model; The initial power consumption prediction model is verified based on the verification set using a time series cross-validation method, and the power consumption prediction model of the corresponding detection area is determined based on the verification result.

4. The method for detecting electricity theft based on a neural network model as claimed in claim 1, characterized in that: The step of obtaining, for each detection area, the type of each electricity user in the detection area and the ratio of different types of electricity users in the detection area, and determining the maximum allowable electricity consumption fluctuation value of the detection area based on the ratio includes: For each detection area, perform the following steps: Obtain the type of each electricity user in the detection area; Obtaining the proportion of each type of electricity users and the historical electricity consumption data of each type of electricity users in a preset historical period; Based on the historical electricity consumption data, determining the total electricity consumption of each type of electricity user; Substituting the total power consumption of each type of power user in the detection area into the power consumption fluctuation prediction formula of the detection area, and calculating the maximum allowable power consumption fluctuation value of the detection area; The power consumption fluctuation prediction formula is: Among them, k i represents the maximum allowable power fluctuation value of the i-th detection area, α w represents the temperature weight of the i-th detection area, α m represents the economic weight of the i-th detection area, ΔW i represents the temperature fluctuation coefficient of the i-th detection area, Δm i represents the economic fluctuation coefficient of the i-th detection area, E i represents the total power consumption in the historical power consumption data of the i-th detection area, p j represents the proportion of electricity users of the jth type in the i-th detection area, β j represents the influence factor of the first influencing factor of the jth type in the i-th detection area, γ j represents the influence factor of the second influencing factor of the jth type in the i-th detection area, ΔE(1) j represents the first impact value of the jth type in the i-th detection area, ΔE(2) j Represents the second impact value of the j-th type in the i-th detection area.

5. The method for detecting electricity theft based on a neural network model as claimed in claim 1, characterized in that: For each detection area, the power consumption of the detection area is predicted based on the power consumption prediction model, and the predicted power consumption data of the detection area is compared with the actual power consumption data to obtain the actual fluctuation value, including: For each detection area, perform the following steps: Based on the historical electricity consumption data of the detection area and the electricity consumption prediction model, the electricity consumption of the detection area is predicted to obtain the predicted electricity consumption data within a preset time period; Acquire actual power consumption data within a preset period, compare the predicted power consumption data with the actual power consumption data, and obtain an actual fluctuation value based on the comparison result; The historical power consumption data and the predicted power consumption data both include the power consumption of the detection area in different time sub-segments and the line loss data of the detection area.

6. The method for detecting electricity theft based on a neural network model as claimed in claim 5, characterized in that: The comparing the preset power consumption data with the actual power consumption data and obtaining the actual fluctuation value based on the comparison result includes: For each detection area, the preset power consumption data and the power consumption in different time sub-segments in the actual power consumption data and the line loss data of the detection area are input into the actual fluctuation value calculation formula to obtain the actual fluctuation value; The actual fluctuation value calculation formula is: Among them, k i ′ is the actual fluctuation value of the i-th detection area, t represents the t-th time segment, s(1) it represents the actual power consumption of the i-th detection area in the t-th time period, s(2) it represents the predicted power consumption of the i-th detection area in the t-th time period, δ1 represents the line loss correction coefficient of the i-th detection area in the t-th time period, and x(1) it represents the actual line loss of the i-th detection area in the t-th time period, x(2) it It represents the preset line loss of the i-th detection area in the t-th time period.

7. The method for detecting electricity theft based on a neural network model as claimed in claim 1, characterized in that: The determining of the abnormal area based on all actual fluctuation values ​​comprises: Compare the actual fluctuation value of each detection area with the corresponding maximum allowable power fluctuation value, and if the actual fluctuation value exceeds the maximum allowable power fluctuation value, determine the detection area as an abnormal area; Otherwise, the detected area is determined as a normal area.

8. The method for detecting electricity theft based on a neural network model as claimed in claim 1, characterized in that: The obtaining of similar areas of each abnormal area, and screening the corresponding abnormal area based on actual power consumption data of the similar areas to determine the power theft area includes: Acquire feature data of each detection area, calculate similarity between each abnormal area and other detection areas based on the feature data, and determine similar areas of each abnormal area based on the similarity; For each anomaly area, perform the following steps: Acquire actual power consumption data of a similar area to the abnormal area, and compare the actual power consumption data of the abnormal area with the actual power consumption data of the corresponding similar area; The difference between the actual power consumption data of the abnormal area and the actual power consumption data of the corresponding similar area is calculated, and each difference is compared with a corresponding threshold value, and it is determined whether the abnormal area is a power theft area based on the comparison result.

9. The method for detecting electricity theft based on a neural network model as claimed in claim 1, characterized in that: The method further comprises: Based on the electricity theft detection model, each electricity user in each electricity theft area is monitored to identify suspicious electricity users who may be involved in electricity theft; The user ID of each suspicious electricity user is sent to the electricity theft alarm system, and staff are notified to conduct on-site verification.

10. An electricity theft detection device based on a neural network model, characterized in that: include: A division module is used to divide the area to be detected into multiple detection areas; A training module is used to obtain historical power consumption data of each detection area, and train a neural network model based on the historical power consumption data to obtain a power consumption prediction model for each detection area; A determination module, configured to obtain, for each detection area, the type of each electricity user in the detection area and the ratio of different types of electricity users in the detection area, and determine the maximum allowable electricity consumption fluctuation value of the detection area based on the ratio; A comparison module is used to predict the power consumption of each detection area based on the power consumption prediction model, and compare the predicted power consumption data of the detection area with the actual power consumption data to obtain an actual fluctuation value; The screening module is used to determine the abnormal area based on all actual fluctuation values, determine the similar area of ​​each abnormal area, and screen the corresponding abnormal area based on the actual power consumption data of the similar area to determine the power theft area.

Images