CN115147135B - DRSN-based method, system and device for identifying electricity stealing users in platform area - Google Patents
DRSN-based method, system and device for identifying electricity stealing users in platform area Download PDFInfo
- Publication number
- CN115147135B CN115147135B CN202210923332.1A CN202210923332A CN115147135B CN 115147135 B CN115147135 B CN 115147135B CN 202210923332 A CN202210923332 A CN 202210923332A CN 115147135 B CN115147135 B CN 115147135B
- Authority
- CN
- China
- Prior art keywords
- electricity
- data
- network
- learning
- periodic
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 230000005611 electricity Effects 0.000 title claims abstract description 90
- 238000000034 method Methods 0.000 title claims abstract description 58
- 230000000737 periodic effect Effects 0.000 claims abstract description 40
- 230000006870 function Effects 0.000 claims abstract description 26
- 238000004364 calculation method Methods 0.000 claims abstract description 16
- 238000004458 analytical method Methods 0.000 claims abstract description 10
- 238000007781 pre-processing Methods 0.000 claims abstract description 6
- 230000008569 process Effects 0.000 claims description 16
- 230000004913 activation Effects 0.000 claims description 13
- 239000011159 matrix material Substances 0.000 claims description 10
- 238000013528 artificial neural network Methods 0.000 abstract description 5
- 230000015556 catabolic process Effects 0.000 abstract description 4
- 238000006731 degradation reaction Methods 0.000 abstract description 4
- 239000006185 dispersion Substances 0.000 abstract description 4
- 230000008859 change Effects 0.000 abstract description 3
- 238000010586 diagram Methods 0.000 description 11
- 238000004422 calculation algorithm Methods 0.000 description 9
- 238000003860 storage Methods 0.000 description 8
- 230000000875 corresponding effect Effects 0.000 description 4
- 238000012423 maintenance Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000004590 computer program Methods 0.000 description 3
- 238000010276 construction Methods 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 230000006399 behavior Effects 0.000 description 2
- 230000008602 contraction Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 239000007787 solid Substances 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000009412 basement excavation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000001747 exhibiting effect Effects 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 230000003442 weekly effect Effects 0.000 description 1
- 238000002759 z-score normalization Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/018—Certifying business or products
- G06Q30/0185—Product, service or business identity fraud
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Business, Economics & Management (AREA)
- Economics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Primary Health Care (AREA)
- Tourism & Hospitality (AREA)
- Public Health (AREA)
- Human Resources & Organizations (AREA)
- General Health & Medical Sciences (AREA)
- Water Supply & Treatment (AREA)
- Entrepreneurship & Innovation (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
A method, a system and a device for identifying a station electricity larceny user based on DRSN are characterized in that the method comprises the following steps: step 1, periodic electricity utilization data of a power grid area are collected, and a depth residual error shrinkage network based on the periodic electricity utilization data is constructed; the learning parent class threshold branch in the depth residual error shrinkage network of the periodic electricity data further comprises a learning child class threshold branch; and 2, preprocessing the electricity consumption data and inputting the electricity consumption data into the depth residual error shrinkage network to acquire analysis data, and constructing a Softmax function to realize electricity theft classification of the analysis data. The method of the application greatly reduces the number of layers of the ResNet network, reduces the problems of gradient dispersion and network degradation in the deep neural network, provides secondary characteristic parameters by a double-layer soft thresholding calculation mode, solves the problem of periodic change of data samples, and has full effectiveness and scientificity in experimental results.
Description
Technical Field
The application relates to the field of smart grids, in particular to a DRSN-based method, a DRSN-based system and a DRSN-based device for identifying a station electricity larceny user.
Background
With the rapid development of the economic society in China, the power supply intensity of a platform area is continuously increased due to the increase of the power demand, and the workload of marketing basic operation and maintenance personnel of a power grid company is overlarge, so that the marketing basic operation and maintenance personnel are often caused to be out of the way. Along with the continuous promotion of the construction of a double-high power system and an energy internet, the lean management requirement of the power demand side is continuously improved, and the contradiction between human resources and operation and maintenance amounts is increasingly prominent. In recent years, with the application and popularization of technology progress represented by artificial intelligence and big data in a power grid, the operation and maintenance automation efficiency of a power grid backbone frame is continuously improved, but because of the complexity and the specificity of a user side, the control force of the power grid on the management of a platform area is not high, the accurate grasp of terminal energy consumption is difficult to realize, higher-level value-added service is difficult to provide for the user, and the lean management of the platform area is always a blind area and a weak point of a power grid company.
With the continuous upgrading and reconstruction of an electric energy metering device in the last twenty years, a mechanical electric meter is replaced by an intelligent electric meter, a marketing system business center database is established, a power grid company marketing department area network acquires massive electric energy data in real time, and the effective utilization and excavation of the data are the problems to be solved currently. The electricity stealing behavior of the transformer area is always the focus of striking processing of the power grid company and is also the focus of special action attention of quality improvement and efficiency enhancement. Various algorithms exist in the prior art for identifying electricity consumption data of electricity stealing users in a platform area so as to attempt to acquire accurate electricity stealing type and electricity stealing information. However, due to the huge amount of electricity consumption data information of the power consumer, the general algorithm has difficulty in accurately exploring the useful electricity stealing information.
On the one hand, the electricity consumption data of the power consumer have high periodicity, and weak or strong periodic fluctuation correlation exists among the annual data, the monthly data and the weekly data, so that the periodic fluctuation is not beneficial to the discovery of electricity stealing behavior. In addition, weather information, life, production modes and the like can also have a great influence on the electricity consumption of the transformer area.
On the other hand, the amount of power consumption data information is large and redundant data content is large, which causes that a general deep neural network is difficult to adapt to the degree of data aggregation. In addition, the deep neural network algorithm in the prior art also has the problems of gradient dispersion and network degradation.
In order to solve the problems, the application provides a novel DRSN-based station electricity stealing user identification method.
Disclosure of Invention
In order to solve the defects existing in the prior art, the application aims to provide a DRSN-based method, a DRSN-based system and a DRSN-based device for identifying a station electricity stealing user, wherein the method improves a soft threshold generation process in a DRSN algorithm, and simultaneously establishes a periodic depth residual error shrinkage network according to the periodic characteristics of electricity consumption data.
The application adopts the following technical scheme.
The application relates to a DRSN-based station electricity stealing user identification method, which comprises the following steps: step 1, periodic electricity utilization data of a power grid area are collected, and a depth residual error shrinkage network based on the periodic electricity utilization data is constructed; the learning parent class threshold branch in the depth residual error shrinkage network of the periodic electricity data also comprises a learning child class threshold branch; and 2, preprocessing the electricity consumption data and inputting the electricity consumption data into a depth residual error shrinkage network to acquire analysis data, and constructing a Softmax function to realize electricity theft classification of the analysis data.
Preferably, the periodic electricity consumption data is electricity consumption data of all users in the platform region acquired in a preset time period.
Preferably, the pretreatment in the step 2 is as follows: step 2.1, normalizing the electricity consumption data; step 2.2, converting the normalized electricity consumption data into a gray matrix sample; the gray matrix sample is a gray map of 28×28.
Preferably, the depth residual error shrinkage network of the periodic electricity data is realized by adopting a 12-layer ResNet network; the first layer receives 28 x 28 gray scale image samples of 7 channels, each two layers of the second to eleventh layers realize short link once, each singular layer realizes soft thresholding once, the second to fifth layers receive 28 x 7 gray scale image samples of 7 channels, and the sixth to eleventh layers receive 7*7 gray scale image samples of 7 channels; the twelfth layer is a full connection layer.
Preferably, a learning parent class threshold branch in the depth residual error shrinkage network is used for realizing soft thresholding; and a second soft threshold calculation process also exists between the full connection layer of the learning parent class threshold branch and the Sigmoid activation function.
Preferably, in the second soft threshold calculation process, the soft threshold calculation is performed based on the learning sub-class threshold branch.
Preferably, the learning sub-class threshold branch sequentially comprises a first full-connection layer, a batch standardization, a rectification linear unit, a second full-connection layer, a Sigmoid activation function and a convolution layer; the first full-connection layer sequentially enters a batch standardization and rectification linear unit, a second full-connection layer and a Sigmoid activation function after being output; the output of the Sigmoid activation function and the output of the first full-connection layer pass through the convolution layer to realize the output of the second soft threshold.
Preferably, the sample width in the threshold branch of the study father class is 1/n of the sample width in the ResNet network; the sample width in the learning sub-class threshold branch is 1/m of the sample width in the learning parent class threshold branch; wherein n is the month number of the periodic electricity data in a preset time period, and m=4.
The second aspect of the application relates to a DRSN-based station electricity stealing user identification system, wherein the system is realized by adopting the DRSN-based station electricity stealing user identification method in the first aspect of the application.
The third aspect of the application relates to a DRSN-based station electricity stealing user identification device, wherein the device comprises a processor; the processor is configured to implement a method for identifying a power-stealing user in a station area based on a DRSN according to the first aspect of the application.
Compared with the prior art, the DRSN-based method, system and device for identifying the station electricity stealing users improve the soft threshold generation process in the DRSN algorithm, and establish a periodic depth residual error shrinkage network according to the periodic characteristics of electricity consumption data. The method of the application greatly reduces the number of layers of the ResNet network, reduces the problems of gradient dispersion and network degradation in the deep neural network, provides secondary characteristic parameters by a double-layer soft thresholding calculation mode, solves the problem of periodic change of data samples, and has full effectiveness and scientificity in experimental results.
Drawings
FIG. 1 is a schematic flow chart of steps of a method for identifying a power stealing user in a station area based on DRSN in the application;
FIG. 2 is a schematic diagram of a gray matrix sample in a DRSN-based method for identifying a power-stealing user in a station area;
FIG. 3 is a schematic diagram of a 12-layer ResNet network in a DRSN-based method for identifying a power stealing subscriber of a station area in the application;
FIG. 4 is a schematic diagram of an improved DRSN network in a DRSN-based method for identifying a power-stealing subscriber in a region of the present application;
FIG. 5 is a schematic diagram showing the change of classification accuracy along with the iteration number in a DRSN-based power-stealing user identification method in a platform;
fig. 6 is a schematic diagram of a topology structure of a station according to an embodiment of the method for identifying a station electricity stealing user based on DRSN in the present application.
Detailed Description
The application is further described below with reference to the accompanying drawings. The following examples are only for more clearly illustrating the technical aspects of the present application, and are not intended to limit the scope of the present application.
Fig. 1 is a schematic flow chart of steps of a method for identifying a power stealing user in a station area based on a DRSN in the application. As shown in fig. 1, the first aspect of the present application relates to a method for identifying a station area electricity stealing user based on a DRSN, which comprises a step 1 and a step 2.
Step 1, periodic electricity utilization data of a power grid area are collected, and a depth residual error shrinkage network based on the periodic electricity utilization data is constructed; and the learning parent class threshold branch in the depth residual error shrinkage network of the periodic electricity data also comprises a learning child class threshold branch.
First, it is understood that the electricity data collected according to various modes in the grid section may be data collected through various modes. For example, the acquisition of the sample data may be achieved by acquiring the power consumption of the load and the power consumption of the gateway table. Generally, when the line loss index of a certain line in the transformer area is found to be unqualified, the electricity consumption data of a period of time before and after the line loss is unqualified can be collected. And the acquired data are used for being input into a depth residual error shrinkage network, so that the study of electricity stealing classification is realized.
Preferably, the periodic electricity consumption data is electricity consumption data of all users in the platform region collected in a preset time period.
In the application, the power utilization data information of all power utilization users in the current station area can be acquired by taking the day as a unit. Such collection may be based on the data content returned by the power meter. In addition, the acquisition of the actual real-time electricity utilization data on a certain line or area can be realized through the gateway table. Therefore, the acquired data can be electricity consumption data with periodic characteristics.
Because the data content has periodicity, the construction of the depth residual error contraction network is realized based on the characteristics of the data.
Preferably, the depth residual error shrinkage network of the periodic electricity data is realized by adopting a 12-layer ResNet network; the first layer receives 28 x 28 gray scale image samples of 7 channels, each two layers of the second to eleventh layers realize short link once, each singular layer realizes soft thresholding once, the second to fifth layers receive 28 x 7 gray scale image samples of 7 channels, and the sixth to eleventh layers receive 7*7 gray scale image samples of 7 channels; the twelfth layer is a full connection layer.
Fig. 3 is a schematic diagram of a 12-layer res net network in a method for identifying a power-stealing user in a station based on DRSN in the present application. As shown in fig. 3, it can be appreciated that in the prior art, a plurality of different layers of res net (residual) networks are used to implement sample computation of the deep neural network. In the application, the number of layers is reduced to 12 layers in order to reduce the gradient dispersion and degradation problems of the network. Specifically, in layers 2 to 11, short linking is realized once every two layers, while soft thresholding can be realized once every two lower layers.
In addition, in the above-described multi-layer, the size of the gradation pattern sample also changes a plurality of times. In the first layer, samples were 28 x 28 in size, and 28 x 7 in layers 2-5, and decreased to 7*7 after 6 layers. In the present application, the sample size is also set according to the periodic characteristics of the data itself. The depth residual shrink network of the present application is thus able to adapt to the periodic characteristics of the input data, planarize such periodic characteristics to adjacent pixels in the sample image that are correlated and eliminate the effects caused by such periodic characteristics.
Preferably, a learning parent class threshold branch in the depth residual error shrink network is used for realizing soft thresholding; and a second soft threshold calculation process also exists between the full connection layer of the learning parent class threshold branch and the Sigmoid activation function.
It can be appreciated that in the depth residual contraction network, a learning threshold branch is generally adopted to realize the output of a soft threshold, and the removal of non-critical characteristic information is realized through the soft threshold. In the application, the learning threshold branch is correspondingly improved to include a learning parent class threshold branch and a learning child class threshold branch.
Fig. 4 is a schematic diagram of an improved DRSN network in a method for identifying a power-stealing subscriber in a cell based on a DRSN in the present application. As shown in fig. 4, a learning sub-class threshold branch is further added to the learning parent class threshold branch, and the learning sub-class threshold branch can output and perform Sigmoid activation after performing a second soft thresholding operation on the data output by the second full connection layer on the parent class threshold branch, so as to generate a soft threshold of the whole res net network.
Preferably, in the second soft threshold calculation process, the soft threshold calculation is performed based on the learning sub-class threshold branch.
Specifically, the second soft threshold calculation can enable the training times required by the operation process to be less, errors to be reduced, accuracy to be higher, and the algorithm to be more stable. In addition, the second soft threshold calculation can adapt to the characteristics of the periodic electricity data, so that the calculation result beats out the periodic influence.
Preferably, the learning sub-class threshold branch sequentially comprises a first full-connection layer, a batch standardization, a rectification linear unit, a second full-connection layer, a Sigmoid activation function and a convolution layer; the first full-connection layer sequentially enters a batch standardization and rectification linear unit, a second full-connection layer and a Sigmoid activation function after being output; the output of the Sigmoid activation function and the output of the first full-connection layer pass through the convolution layer to realize the output of the second soft threshold.
It can be understood that the learning sub-class threshold branch in the present application is substantially the same as the original learning threshold branch in the prior art, but the nested structure adopted in the present application makes the soft thresholding process more accurate.
Preferably, the sample width in the threshold branch of the learning parent class is 1/n of the sample width in the ResNet network; the sample width in the learning sub-class threshold branch is 1/m of the sample width in the learning parent class threshold branch; wherein n is the month number of the periodic electricity data in a preset time period, and m=4.
In the application, n represents the month number contained in the sample data, and m represents the cycle number in each month, so that the nested or independent periodic characteristic of the sample data is relieved through the generation process of the soft threshold value twice. For example, for some large plant users, the electricity consumption is relatively fixed in terms of output per month in addition to exhibiting periodicity on weekdays and non-weekdays of the week, so that the electricity consumption data at the beginning of the month, in the month, and at the end of the month also exhibit fixed fluctuation characteristics. Therefore, the application adopts learning threshold branches with different scales, so that the periodicity is fully eliminated.
And 2, preprocessing electricity consumption data and inputting the electricity consumption data into the depth residual error shrinkage network to acquire analysis data, and constructing a Softmax function to realize electricity theft classification of the analysis data.
After the data acquisition and the model construction are realized, the method can realize the data preprocessing process according to the characteristics of the model.
Preferably, the pretreatment in the step 2 is as follows: step 2.1, normalizing the electricity consumption data; step 2.2, converting the normalized electricity consumption data into a gray matrix sample; the gray matrix sample is a gray map of 28×28.
Fig. 2 is a schematic diagram of a gray matrix sample in a method for identifying a power-stealing user in a station based on DRSN in the present application. It will be appreciated that fig. 2 is a matrix gray scale map implemented using the preprocessing method described above. The generation of the grey-scale map itself may be based on periodic data, including 4 grey-scale matrices 7*7 in the map, so that the periodicity of the data can be interpreted in the form of a grey-scale map and calculated by subsequent processes. It should be noted that 7*7 herein may be obtained by reasonably arranging and combining corresponding data of seven days of a week in the power data. Therefore, the periodic data can be reasonably organized by the arrangement mode of the matrix gray level diagram, and the periodicity rule in the periodic data is considered in the actual calculation process.
The application can also classify the electricity stealing modes according to different electricity stealing means, realize the characteristic analysis of electricity stealing data according to different classifications, establish a Softmax function according to the analysis, and input the data output by the DRSN algorithm of the application into the Softmax function so as to realize the classification of specific electricity stealing types.
Fig. 5 is a schematic diagram of a classification accuracy rate changing with iteration times in a method for identifying a power-stealing user in a station based on a DRSN in the present application. Fig. 6 is a schematic diagram of a topology structure of a station according to an embodiment of the method for identifying a station electricity stealing user based on DRSN in the present application. As shown in fig. 5 and fig. 6, the method of the present application is verified by using line loss failure data in two years of a city power supply company as an experimental sample. The company area comprises an S11 type transformer with 200kVA, the low-voltage lines are composed of LJ-50, LJ-35 and LJ-25, the maximum power supply radius is Wie535m, and the total of 18 low-voltage load metering points is except a gateway table. In addition, the method collects gateway table data and user power consumption data of 15 days before and after the line loss failure corresponding to the transformer area.
The normalization mode adopted in the application is a Z-score normalization method, and the compression process of the sample size adopts pooling calculation. By utilizing the method, the method of the application can check and treat the electricity larceny event 921 in two years, and the method relates to the 1209.17 ten thousand yuan, and compared with the original electricity larceny inquiry modes in 2017 and 2018, the judging effect of the electricity larceny event is respectively improved by 336% and 289%. Table 1 table for checking fraudulent use of electricity
In addition, in fig. 5, the double-layer depth residual shrinkage network, that is, the algorithm in the application, has relatively faster starting chain speed and higher accuracy.
The second aspect of the application relates to a DRSN-based station electricity stealing user identification system, wherein the system is realized by adopting the DRSN-based station electricity stealing user identification method in the first aspect of the application.
The third aspect of the present application relates to a device for identifying a power-stealing user in a station based on a DRSN, which is used for implementing a method for identifying a power-stealing user in a station based on a DRSN in the first aspect of the present application.
It can be understood that, in order to implement the functions in the method provided by the embodiment of the present application, the device for identifying a power-stealing user in a platform area includes a hardware structure and/or a software module for executing the corresponding functions. Those of skill in the art will readily appreciate that the various illustrative algorithm steps described in connection with the embodiments disclosed herein may be implemented as hardware or combinations of hardware and computer software. Whether a function is implemented as hardware or computer software driven hardware depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
The embodiment of the application can divide the functional modules of the power-stealing user identifying device according to the method example, for example, each functional module can be divided corresponding to each function, and two or more functions can be integrated in one processing module. The integrated modules may be implemented in hardware or in software functional modules. It should be noted that, in the embodiment of the present application, the division of the modules is schematic, which is merely a logic function division, and other division manners may be implemented in actual implementation.
The apparatus includes at least one processor, a bus system, and at least one communication interface.
The processor may be a central processing unit (Central Processing Unit, CPU), or replaced by other hardware such as a field programmable gate array (Field Programmable Gate Array, FPGA) or Application-specific integrated circuit (ASIC), or may be an FPGA or other hardware and CPU together.
The memory may be a read-only memory (ROM) or other type of static storage device that may store static information and instructions, a random access memory (random access memory, RAM) or other type of dynamic storage device that may store information and instructions. An electrically erasable programmable read-only memory (EEPROM), compact disk read-only memory (CD-ROM) or other optical disk storage, optical disk storage (including compact disk, laser disk, optical disk, digital versatile disk, blu-ray disk, etc.), magnetic disk storage media or other magnetic storage devices, any other medium capable of carrying or storing desired program code in the form of instructions or data structures and of being accessed by a computer, may also be implemented as the memory in the present application, but is not limited to such. The memory may be stand alone and coupled to the processor via a bus. In the alternative, the memory may be integral to the processor.
The hard disk may be a mechanical disk or a solid state disk (Solid State Drive, SSD), etc. The embodiment of the present application is not limited thereto. The interface card in the hard disk module is communicated with the hard disk. The storage node communicates with an interface card of the hard disk module to access the hard disk in the hard disk module.
When the above-described functions in the present application are implemented using a software program, they may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in accordance with embodiments of the present application are produced in whole or in part. The computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable apparatus. The computer instructions may be stored in or transmitted from one computer-readable storage medium to another, for example, by wired (e.g., coaxial cable, fiber optic, digital subscriber line (digital subscriber line, simply DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means from one website, computer, server, or data center.
In addition, the terms "first," "second," are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include one or more such feature. In the description of the embodiments of the present application, unless otherwise indicated, the meaning of "a plurality" is two or more.
Although the application is described herein in connection with various embodiments, other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed application, from a study of the drawings, the disclosure, and the appended claims.
While the applicant has described and illustrated the embodiments of the present application in detail with reference to the drawings, it should be understood by those skilled in the art that the above embodiments are only preferred embodiments of the present application, and the detailed description is only for the purpose of helping the reader to better understand the spirit of the present application, and not to limit the scope of the present application, but any improvements or modifications based on the spirit of the present application should fall within the scope of the present application.
Claims (3)
1. The method for identifying the electricity stealing users of the platform area based on the DRSN is characterized by comprising the following steps of: step 1, periodic electricity utilization data of a power grid area are collected, and a depth residual error shrinkage network based on the periodic electricity utilization data is constructed; the learning parent class threshold branch in the depth residual error shrinkage network of the periodic electricity data further comprises a learning child class threshold branch;
step 2, preprocessing the electricity consumption data and inputting the electricity consumption data into the depth residual error shrinkage network to obtain analysis data, and constructing a Softmax function to realize electricity theft classification of the analysis data;
the periodic electricity utilization data are collected in a preset time period and are the electricity utilization data of all users in the platform area;
the pretreatment process in the step 2 is as follows:
step 2.1, normalizing the electricity consumption data;
step 2.2, converting the normalized electricity consumption data into a gray matrix sample;
the gray matrix sample is a gray graph of 28 x 28;
the depth residual error shrinkage network of the periodic electricity data is realized by adopting a 12-layer ResNet network; the first layer receives 28 x 28 gray scale image samples of 7 channels, each two layers of the second to eleventh layers realize short link once, each singular layer realizes soft thresholding once, the second to fifth layers receive 28 x 7 gray scale image samples of 7 channels, and the sixth to eleventh layers receive 7*7 gray scale image samples of 7 channels;
the twelfth layer is a full-connection layer;
the learning parent class threshold branch in the depth residual error shrink network is used for realizing the soft thresholding;
a second soft threshold calculation process is also arranged between the full-connection layer of the learning parent class threshold branch and the Sigmoid activation function;
in the second soft threshold calculation process, soft threshold calculation is realized based on a learning sub-class threshold branch; the learning sub-class threshold branch comprises a first full-connection layer, a batch standardization, rectification linear unit, a second full-connection layer, a Sigmoid activation function and a convolution layer in sequence;
the first full-connection layer sequentially enters a batch standardization and rectification linear unit, a second full-connection layer and a Sigmoid activation function after being output;
the output of the Sigmoid activation function and the output of the first full-connection layer pass through a convolution layer to realize the output of a second soft threshold;
the sample width in the threshold branch of the learning parent class is 1/n of the sample width in the ResNet network;
the sample width in the learning sub-class threshold branch is 1/m of the sample width in the learning parent class threshold branch;
and n is the month number of the periodic electricity consumption data in a preset time period, and m=4.
2. A DRSN-based station electricity stealing user identification system is characterized in that: the system is realized by adopting the DRSN-based platform electricity stealing user identification method in claim 1.
3. A platform district electricity stealing user identification device based on DRSN, characterized in that: the apparatus includes a processor;
the processor is configured to implement the DRSN-based method for identifying a power-stealing user in a platform.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210923332.1A CN115147135B (en) | 2022-08-02 | 2022-08-02 | DRSN-based method, system and device for identifying electricity stealing users in platform area |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210923332.1A CN115147135B (en) | 2022-08-02 | 2022-08-02 | DRSN-based method, system and device for identifying electricity stealing users in platform area |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115147135A CN115147135A (en) | 2022-10-04 |
| CN115147135B true CN115147135B (en) | 2023-10-24 |
Family
ID=83413707
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210923332.1A Active CN115147135B (en) | 2022-08-02 | 2022-08-02 | DRSN-based method, system and device for identifying electricity stealing users in platform area |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115147135B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112098714A (en) * | 2020-08-12 | 2020-12-18 | 国网江苏省电力有限公司南京供电分公司 | ResNet-LSTM-based electricity stealing detection method and system |
| CN113393103A (en) * | 2021-06-03 | 2021-09-14 | 西南科技大学 | Anti-electricity-stealing system based on genetic algorithm optimization BP neural network |
| CN113901977A (en) * | 2020-06-22 | 2022-01-07 | 中国电力科学研究院有限公司 | Deep learning-based power consumer electricity stealing identification method and system |
-
2022
- 2022-08-02 CN CN202210923332.1A patent/CN115147135B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113901977A (en) * | 2020-06-22 | 2022-01-07 | 中国电力科学研究院有限公司 | Deep learning-based power consumer electricity stealing identification method and system |
| CN112098714A (en) * | 2020-08-12 | 2020-12-18 | 国网江苏省电力有限公司南京供电分公司 | ResNet-LSTM-based electricity stealing detection method and system |
| CN113393103A (en) * | 2021-06-03 | 2021-09-14 | 西南科技大学 | Anti-electricity-stealing system based on genetic algorithm optimization BP neural network |
Non-Patent Citations (3)
| Title |
|---|
| GMA-DRSNs: A novel fault diagnosis method with global multi-attention deep residual shrinkage networks;Zhijin Zhang;《Measurement》;第1-14页 * |
| Robust Deep Residual Shrinkage Networks for Online Fault Classifications;Alireza Salimy;《2021 29th European Signal Processing Conference》;第1-5页 * |
| Signal Model-Based Fault Detection and Diagnosis for Induction Motors Using Features of Vibration Signal in Two-Dimension Domain;Van Tuan Do;《Journal of Mechanical Engineering》;第655-666页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115147135A (en) | 2022-10-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108985330B (en) | Self-coding network and training method thereof, and abnormal power utilization detection method and system | |
| CN111738462B (en) | Fault first-aid repair active service early warning method for electric power metering device | |
| CN113283175A (en) | Photovoltaic power station joint fault diagnosis method based on asynchronous decentralized federal learning | |
| CN113361559B (en) | Multi-mode data knowledge information extraction method based on deep-width combined neural network | |
| CN115730749B (en) | Power dispatching risk early warning method and device based on fusion power data | |
| CN113469266A (en) | Electricity stealing behavior detection method based on improved deep convolutional neural network | |
| CN113516501A (en) | User communication behavior prediction method and device based on graph neural network | |
| CN115758151A (en) | Combined diagnosis model establishing method and photovoltaic module fault diagnosis method | |
| CN115456093A (en) | High-performance graph clustering method based on attention-graph neural network | |
| CN116826710A (en) | Peak clipping strategy recommendation method and device based on load prediction and storage medium | |
| CN114880538A (en) | Attribute graph community detection method based on self-supervision | |
| CN114679372A (en) | Node similarity-based attention network link prediction method | |
| CN114266676A (en) | Abnormal electricity utilization detection method of genetic optimization Bagging heterogeneous integration model | |
| CN115147135B (en) | DRSN-based method, system and device for identifying electricity stealing users in platform area | |
| CN114021425A (en) | Power system operation data modeling and feature selection method and device, electronic equipment and storage medium | |
| CN113891342B (en) | Base station inspection method and device, electronic equipment and storage medium | |
| CN117495422A (en) | Cost management system and method based on power communication network construction | |
| CN112508734A (en) | Method and device for predicting power generation capacity of power enterprise based on convolutional neural network | |
| CN116774089A (en) | Convolutional neural network battery state of health estimation method and system based on feature fusion | |
| CN115687788A (en) | Intelligent business opportunity recommendation method and system | |
| CN113837486B (en) | RNN-RBM-based distribution network feeder long-term load prediction method | |
| Prashanthi et al. | A comparative study of the performance of machine learning based load forecasting methods | |
| CN114915563A (en) | Network flow prediction method and system | |
| CN113744081B (en) | Analysis method for electricity stealing behavior | |
| CN116340765B (en) | Electricity larceny user prediction method and device, storage medium and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |