CN117390546A - Multimode database fusion calculation model for instant anti-electricity-theft detection - Google Patents
Multimode database fusion calculation model for instant anti-electricity-theft detection Download PDFInfo
- Publication number
- CN117390546A CN117390546A CN202311362622.4A CN202311362622A CN117390546A CN 117390546 A CN117390546 A CN 117390546A CN 202311362622 A CN202311362622 A CN 202311362622A CN 117390546 A CN117390546 A CN 117390546A
- Authority
- CN
- China
- Prior art keywords
- electricity
- larceny
- stealing
- sample
- data
- 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.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 26
- 238000004364 calculation method Methods 0.000 title claims abstract description 22
- 230000004927 fusion Effects 0.000 title claims abstract description 19
- 230000005611 electricity Effects 0.000 claims abstract description 410
- 238000000034 method Methods 0.000 claims abstract description 45
- 230000005856 abnormality Effects 0.000 claims description 33
- 238000004422 calculation algorithm Methods 0.000 claims description 25
- 230000002159 abnormal effect Effects 0.000 claims description 22
- 230000008859 change Effects 0.000 claims description 16
- 238000012549 training Methods 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 9
- 238000013528 artificial neural network Methods 0.000 claims description 8
- 238000007689 inspection Methods 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 7
- 230000000750 progressive effect Effects 0.000 claims description 7
- 238000012216 screening Methods 0.000 claims description 7
- 238000004088 simulation Methods 0.000 claims description 6
- 238000003066 decision tree Methods 0.000 claims description 4
- 230000007123 defense Effects 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 claims description 4
- 239000003550 marker Substances 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 abstract description 4
- 238000005065 mining Methods 0.000 abstract description 3
- 238000012544 monitoring process Methods 0.000 abstract description 2
- 238000013527 convolutional neural network Methods 0.000 description 10
- 230000006870 function Effects 0.000 description 7
- 238000007405 data analysis Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000001939 inductive effect Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
- G06F18/15—Statistical pre-processing, e.g. techniques for normalisation or restoring missing data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/0895—Weakly supervised learning, e.g. semi-supervised or self-supervised learning
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application provides a multimode database fusion calculation model for instant anti-electricity-stealing detection, and belongs to the technical field of power grid monitoring. The method comprises the steps of based on exploration of historical electricity larceny samples, mining electricity consumption characteristics of special-purpose power-change electricity larceny users, summarizing and summarizing the electricity consumption characteristics, generating electricity consumption figures of the users, merging the historical electricity larceny users, arranging to form a classical electricity larceny case library, generating electricity larceny fingerprints of the users based on conversion rules according to the electricity consumption characteristics of the users, arranging different electricity larceny methods to generate electricity larceny fingerprint libraries, building a special-purpose power-change electricity larceny identification model based on the electricity larceny fingerprint libraries and the electricity consumption signals (electricity consumption data characteristics or electricity consumption phenomenon characteristics) of the users, accurately positioning the electricity larceny users, and giving out suspected electricity larceny methods; therefore, by analyzing big data of the electricity larceny sample, customers for electricity larceny and illegal electricity consumption are accurately positioned, the anti-electricity larceny resource is optimized and integrated, and the smooth and efficient operation of the workflow is ensured.
Description
Technical Field
The application relates to the field of power grid monitoring, in particular to a multimode database fusion calculation model for instant anti-electricity-stealing detection.
Background
The detection of the anti-electricity-theft is a difficult problem in practice, the electricity-theft means are also continuously refined, and the traditional method mainly based on equipment inspection is difficult to obtain evidence and is difficult to deal with, prevent and control. In recent years, as large-scale power companies such as national power grids and southern power grids are connected with an electric energy and electric quantity collection system, based on collected electronic data, the power consumption condition of a customer is analyzed by utilizing a data analysis and mining method, and abnormal customers with large electric quantity fluctuation are detected at present, so that the informatization means based on the large data gradually show obvious advantages.
The current anti-electricity-theft attack defense simulation expands anti-electricity-theft research work based on the data of the electricity consumption information acquisition system, but the data processing method is mostly shallow data analysis based on manual experience, and deep relevance of each data is difficult to find; the existing collected data has massive unlabeled data, the tens of thousands of electric energy meters are manually identified and marked with low efficiency, and subjective misjudgment exists; and the case data of the anti-electricity-theft knowledge database is relatively less, even if the current electricity-theft type is completely identified, with the continuous development of information technology, new electricity-theft types can continuously appear, and the identification of the new electricity-theft types is difficult to realize only according to the case data of the anti-electricity-theft knowledge database.
Disclosure of Invention
In order to make up for the defects, the application provides a multimode database fusion calculation model for instant anti-electricity-theft detection, and aims to solve the problems in the background technology.
The embodiment of the application provides a multimode database fusion calculation model for instant anti-electricity-theft detection, which comprises the following steps:
s1: respectively designing and realizing decision trees in a service database for acquiring electric energy and electric quantity;
s2: preprocessing the electricity consumption sample data of the user, and storing the preprocessed data into a historical sample database;
s3: selecting relevant characteristics of user electricity consumption data of the electric energy and electric quantity collection industry score data vehicle and actual credit rating of the user as a model training data set;
s4: performing marker screening on the database samples;
s5: model training is performed by a combination method, and model training is performed based on a real-time and historical electricity utilization data set;
s6: extracting electricity stealing characteristics from corresponding user electricity data or user phenomena according to the electricity stealing information, determining corresponding electricity stealing methods according to the corresponding electricity stealing characteristics, generating electricity using images of the users according to the electricity using data or the electricity stealing phenomena of the electricity stealing users in combination with the electricity stealing information, merging the electricity stealing users to form an electricity stealing case library, generating electricity stealing fingerprints of the users according to the electricity using data characteristics or the electricity stealing characteristics of the users, arranging different electricity stealing methods according to the electricity stealing fingerprints to generate electricity stealing fingerprint libraries, and establishing a special change user electricity stealing identification prediction model.
In a specific embodiment, S6 comprises the steps of:
s61: the abnormal detection comprises the steps of acquiring power consumption data or power consumption phenomenon of a specially-changed user, detecting power consumption abnormality, capturing abnormal data, extracting characteristic quantity of power consumption data characteristics of the user if the abnormal data is captured, comparing and judging the acquired abnormal data with corresponding set standard data or comparison data, and calculating rule violation probability;
s62: if the electricity larceny probability reaches or exceeds the preset rule violation probability, identifying a suspected electricity larceny user according to a special change user electricity larceny identification prediction model, matching electricity consumption data characteristics of the suspected electricity larceny user with electricity larceny fingerprints in an electricity larceny fingerprint library, primarily judging a suspected electricity larceny mode and suspected electricity larceny time, and generating a suspected electricity larceny user list;
s63: the electricity larceny checking comprises the steps of generating an electricity larceny checking work list according to a suspected electricity larceny user list, issuing a dispatch work list for checking, obtaining checking information, recording the checking information according to a user, extracting electricity consumption data characteristics or electricity consumption phenomenon characteristics of the user according to the checking information of the user and electricity consumption data or electricity consumption phenomenon, and updating an electricity larceny case library or electricity larceny fingerprint library;
s64: and electricity stealing processing, namely obtaining inspection conditions according to the inspection information to perform electricity stealing processing.
In a specific embodiment, S4 comprises the steps of:
s41: screening out an electricity stealing sample, classifying the existing electricity stealing category, and marking the existing electricity stealing sample and a normal sample by category;
s42: removing normal samples to form a suspected electricity larceny sample set, and marking the class;
s43: and constructing a suspected electricity larceny identification model of a shop in the GRU algorithm aiming at the suspected electricity larceny sample set.
In a specific embodiment, S4 further comprises the steps of:
s44: calculating the recognition accuracy rate of the suspected electricity larceny identification model according to the output result of the suspected electricity larceny identification model and the class mark of the suspected electricity larceny collection, and carrying out feedback adjustment on model parameters;
s45: and determining an anti-electricity-stealing defense simulation model for self-adaptive class identification according to the adjusted model parameters.
In a specific embodiment, S5 comprises the steps of:
s51: the result election of each pre-measured electricity user in each period is carried out;
s52: predicting the model by a progressive method, namely predicting the model obtained in the step S51;
s53: combining and de-duplicating the predicted credit level users obtained by the combination method and the progressive method and then picking up the service system to perform early warning detection.
In a specific embodiment, S64 further comprises a fee calculation step of checking whether electricity is stolen according to the checking information, and if so, calculating basic electricity fee and illegal fine according to electricity consumption data of a user or electricity consumption phenomenon and electricity stealing operation and electricity stealing time.
In a specific embodiment, S61 includes one or more of power anomaly, line loss anomaly, phase anomaly, voltage anomaly, current anomaly, power factor anomaly, power anomaly including: the power is super-capacity, the daily active power is changed abnormally, the power is unbalanced abnormally, and the electric quantity abnormality comprises: the daily electricity quantity change abnormality includes: the daily electricity suddenly drops abnormally, and the line loss abnormality includes: the line loss variation abnormality includes: line loss surge anomaly, current anomaly includes: abnormal current imbalance and abnormal current change, the abnormal current change includes: phase current dip anomalies, voltage anomalies include: the voltage under-voltage abnormality includes: the phase voltage undervoltage abnormality, the power factor abnormality includes: power factor variation anomalies, including: power factor dip anomalies.
In a specific embodiment, in S42, the process of excluding the normal sample to form the suspected fraudulent sample set and performing the class marking is: and clustering and dividing a part of marked historical collection sample sets by adopting a semi-supervised k-NNM algorithm, removing samples clustered with marked normal samples, forming a suspected electricity stealing sample set, wherein the suspected electricity stealing sample set comprises a part of marked electricity stealing sample clustered sample sets and unmarked sample clustered sample sets, and marking suspected electricity stealing category with the marked electricity stealing sample clustered sample sets.
In a specific embodiment, in S43, the process of constructing the suspected electricity theft identification model based on the GRU algorithm for the suspected electricity theft sample set is as follows: based on a historical collection sample set, a marked electricity stealing sample set and a normal electricity using sample set are extracted, a suspected electricity stealing identification model is built by adopting a GRU algorithm according to the time sequence rule of the electricity stealing sample characteristic data, the electricity stealing behavior characteristic data of the high-risk electricity stealing sample set is mined, the time sequence characteristic is taken as input, the historical track of the electricity stealing sample electricity using data is extracted and reserved by a neural network, a loss function is built according to the identification accuracy of an actual positive sample and an actual negative sample, and the model parameters are adjusted by gradient descent.
In a specific embodiment, the process of constructing the suspected electricity larceny identification model based on the GRU algorithm for the suspected electricity larceny sample set is as follows: based on a historical collection sample set, a marked electricity stealing sample set and a normal electricity using sample set are extracted, a suspected electricity stealing identification model is built by adopting a GRU algorithm according to the time sequence rule of the electricity stealing sample characteristic data, the electricity stealing behavior characteristic data of the high-risk electricity stealing sample set is mined, the time sequence characteristic is taken as input, the historical track of the electricity stealing sample electricity using data is extracted and reserved by a neural network, a loss function is built according to the identification accuracy of an actual positive sample and an actual negative sample, and the model parameters are adjusted by gradient descent.
The beneficial effects are that: based on exploration of historical electricity larceny samples, the electricity consumption characteristics of special-purpose power-change electricity larceny users are discovered, electricity consumption images of the users are generated by summarizing and inducing the electricity consumption characteristics, classical electricity larceny case libraries are formed by merging the historical electricity larceny users, electricity larceny fingerprints of the users are generated based on conversion rules according to the electricity consumption characteristics of the users, electricity larceny fingerprint libraries are generated by different electricity larceny methods, an electricity larceny identification model of the special-purpose power-change users is built based on the electricity larceny fingerprint libraries and the electricity consumption signals (electricity consumption data characteristics or electricity consumption phenomenon characteristics) of the users, and suspected electricity larceny methods are given out; therefore, by analyzing big data of the electricity larceny sample, customers for electricity larceny and illegal electricity consumption are accurately positioned, the anti-electricity larceny resource is optimized and integrated, and the smooth and efficient operation of the workflow is ensured.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the embodiments will be briefly described below, it being understood that the following drawings only illustrate some examples of the present application and therefore should not be considered as limiting the scope, and that other related drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic structural diagram of a multi-mode database fusion calculation model for instant anti-electricity-theft detection according to an embodiment of the present application;
fig. 2 is a flowchart of S6 provided in an embodiment of the present application;
fig. 3 is a S4 flowchart provided in an embodiment of the present application;
fig. 4 is a flowchart of S5 provided in an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be described below with reference to the drawings in the embodiments of the present application.
Referring to fig. 1-4, the present application provides a multimode database fusion calculation model for instant anti-electricity-theft detection, which includes the following steps:
s1: respectively designing and realizing decision trees in a service database for acquiring electric energy and electric quantity; the invention uses C4.5 and CNN neural networks, uses LeNet5 and LSTM machine learning algorithms, and uses database user-defined function UDF to reserve, which are respectively named as UDF_DT, UDF_CNN and UDF_LSTM.
It should be noted that, the present invention selects the three algorithms, and uses decision trees to represent the branch logic characteristics of the general electricity stealing for electricity extraction (such as the electricity stealing tendency of the high-level residential district in the address is generally lower than that of the low-level residential district, the female is generally lower than that of the male, the scale enterprise is generally lower than that of the individual user, etc.), uses the CNN convolutional neural network to extract the characteristics of the similar regular images which are generally presented by the user quantity data in multiple dimensions such as time, region, energy efficiency level, etc., and uses the LSTM to extract the continuous curve abnormal predictable characteristics (similar to the time sequence predicting characteristics of stock price) which can be formed by the electricity stealing data in time sequence.
S2: preprocessing the electricity consumption sample data of the user, and storing the preprocessed data into a historical sample database.
S3: selecting relevant characteristics of user electricity consumption data of the electric energy and electric quantity collection industry score data vehicle and actual credit rating of the user as a model training data set; the invention adopts time, month, user energy efficiency grade, address, sex, user organization, month electricity consumption, meter type, station line loss, three-phase unbalance rate, unit consumption and the like, and user actual credit rating, and is divided into four stages, wherein 0 represents low electricity stealing tendency, 1 represents middle stage, 2 represents high stage, 3 represents high risk, and the three-stage electricity stealing degree is divided by the severity degree of actual electricity stealing history as a model training data set.
S4: performing marker screening on the database samples; wherein the use of subsequent data is facilitated.
S5: model training is performed by a combination method, and model training is performed based on a real-time and historical electricity utilization data set; wherein, based on the real-time and historical electricity data sets (the history span can be set in a customized way), the electricity data sets can be recorded as (D_FeatureTraning, D_LabelTraning); the combined method model training script is written in a programmable SQL sentence, namely, the data set is trained by using UDF_DT, UDF_CNN and UDF_LSTM respectively.
S6: extracting electricity stealing characteristics from corresponding user electricity data or user phenomena according to the electricity stealing information, determining corresponding electricity stealing methods according to the corresponding electricity stealing characteristics, generating electricity using images of the users according to the electricity using data or the electricity stealing phenomena of the electricity stealing users in combination with the electricity stealing information, merging the electricity stealing users to form an electricity stealing case library, generating electricity stealing fingerprints of the users according to the electricity using data characteristics or the electricity stealing characteristics of the users, arranging different electricity stealing methods according to the electricity stealing fingerprints to generate electricity stealing fingerprint libraries, and establishing a special change user electricity stealing identification prediction model. The method comprises the steps of mining the electricity consumption characteristics of a user, summarizing and summarizing the electricity consumption characteristics to generate an electricity consumption portrait of the user, merging historical electricity stealing users, sorting to form a classical electricity stealing case library, generating electricity stealing fingerprints of the user according to the electricity consumption characteristics of the user based on conversion rules, sorting different electricity stealing methods to generate an electricity stealing fingerprint library, building a special change user electricity stealing identification model based on the electricity stealing fingerprint library and the electricity consumption signals (electricity consumption data characteristics or electricity consumption phenomenon characteristics) of the user, accurately positioning the electricity stealing user, and giving out suspected electricity stealing methods; therefore, by analyzing big data of the electricity larceny sample, customers for electricity larceny and illegal electricity consumption are accurately positioned, the anti-electricity larceny resource is optimized and integrated, and the smooth and efficient operation of the workflow is ensured.
In this embodiment, S6 includes the steps of:
s61: and (3) abnormality detection, namely acquiring power consumption data or power consumption phenomena of the special transformer users, detecting power consumption abnormality, capturing the abnormal data, extracting characteristic quantities of the power consumption data of the users if the abnormal data is captured, comparing and judging the acquired abnormal data with corresponding set standard data or comparison data, and calculating the rule violation probability.
S62: and judging electricity larceny, namely if the electricity larceny probability reaches or exceeds the set rule violation probability, identifying a suspected electricity larceny user according to a special change user electricity larceny identification prediction model, matching electricity consumption data characteristics of the suspected electricity larceny user with electricity larceny fingerprints in an electricity larceny fingerprint library, and primarily judging a suspected electricity larceny mode and suspected electricity larceny time to generate a suspected electricity larceny user list.
S63: and (3) electricity larceny checking, namely generating an electricity larceny prevention checking work list according to a suspected electricity larceny user list, issuing a dispatch work list for checking, acquiring checking information, recording the checking information according to the user, extracting electricity consumption data characteristics or electricity consumption phenomenon characteristics of the user according to the checking information of the user and the electricity consumption data or electricity consumption phenomenon, and updating an electricity larceny case library or electricity larceny fingerprint library.
S64: and electricity stealing processing, namely obtaining inspection conditions according to the inspection information to perform electricity stealing processing.
The electricity utilization data features or electricity theft features include: three-phase three-wire, high-voltage power meter, line loss sudden increase, power consumption sudden drop and A, C two-phase current are balanced, but the sudden drop is obvious, and the suspicion of electricity theft with secondary side symmetrical shunting is primarily judged; preferably, the electricity usage data feature or electricity theft feature comprises: three-phase four-wire, high-voltage meter, line loss sudden increase, power consumption sudden drop, three-phase current sudden drop and balance, normal voltage, and primary judgment of suspicion of electricity theft of balanced split of the secondary side A, B, C; preferably, the electricity usage data feature or electricity theft feature comprises: three-phase three-wire, high-voltage, line loss sudden increase, user electricity consumption sudden drop, A, C two-phase current obviously unbalanced and A-phase current sudden drop, power factor sudden drop, and preliminary judgment of suspicion of electricity theft of CT primary side A-phase shunt exists for the user; preferably, the electricity usage data feature or electricity theft feature comprises: three-phase three-wire, high-voltage, line loss sudden increase, user electricity consumption sudden drop, A, C two-phase current obviously unbalanced and C-phase current sudden drop, power factor sudden drop, and preliminary judgment of electricity stealing suspicion of CT primary side C-phase shunt; preferably, the electricity usage data feature or electricity theft feature comprises: three-phase three-wire, high-voltage, line loss sudden increase, user electricity consumption sudden drop, A, C two-phase current balance and A, C-phase current sudden drop, and power factor sudden drop, and the suspicion of CT primary side A, C-phase shunt electricity larceny of the user is primarily judged.
The electricity utilization data features or electricity theft features include: three-phase four-wire, high-power supply low meter, line loss sudden increase, significant current sudden drop for users and insufficient B-phase voltage, and preliminarily judging that the suspicion of B-voltage power theft of the four-wire low meter exists; preferably, the electricity usage data feature or electricity theft feature comprises: three-phase four-wire, high-supply low-meter, line loss sudden increase, significant current sudden drop for users and insufficient voltage of A phase, and preliminarily judging that the suspicion of voltage loss and electricity theft of the four-wire low-meter A exists; preferably, the electricity usage data feature or electricity theft feature comprises: three-phase four-wire, high-power supply low-power meter, line loss sudden increase, remarkable sudden drop of current for users and insufficient C-phase voltage, and the suspicion of C-voltage power theft of the four-wire low-power meter is primarily judged; preferably, the electricity usage data feature or electricity theft feature comprises: three-phase three-wire, high-voltage meter, line loss sudden increase, user current sudden drop and A-phase voltage deficiency, and preliminary judgment of suspicion of electricity larceny of three-wire high-voltage meter A voltage loss; preferably, the electricity usage data feature or electricity theft feature comprises: three-phase three-wire, high-voltage meter, line loss sudden increase, current sudden drop for users and insufficient C-phase voltage, and the suspicion of voltage-loss electricity theft of the three-wire high-voltage meter C is primarily judged; preferably, the electricity usage data feature or electricity theft feature comprises: three-phase three-wire, high-supply low meter, line loss sudden increase, sudden drop of current for users and insufficient voltage of A phase, and preliminary judgment of suspicion of voltage loss and electricity theft of the three-wire low meter A; preferably, the electricity usage data feature or electricity theft feature comprises: three-phase three-wire, high-supply low meter, line loss sudden increase, sudden drop of current for users and insufficient C-phase voltage, and the suspicion of voltage loss and electricity theft of the three-wire low meter C is primarily judged.
In this embodiment, S4 includes the steps of:
s41: screening out an electricity stealing sample, classifying the existing electricity stealing category, and marking the existing electricity stealing sample and a normal sample by category; the method comprises the steps of screening out electricity stealing samples based on a historical collection sample set, classifying existing electricity stealing categories, and marking partial categories of the existing electricity stealing samples and normal samples.
S42: removing normal samples to form a suspected electricity larceny sample set, and marking the class; the method comprises the steps of carrying out clustering division on a part of marked historical collection sample sets by adopting a semi-supervised K-NNM algorithm, namely, firstly adopting a one-to-one mode of samples, simplifying initial classification to the greatest extent based on a K-nearest neighbor clustering concept, then adopting a pair of mean modes to rely on latest unmarked sample data to the greatest extent by referring to the concept of K mean clustering, iteratively optimizing and guaranteeing the clustering of the maximum similarity distance between samples, setting an iteration threshold value for avoiding falling into an iteration dead loop, and carrying out classification according to a square error E, so that classification processing can be carried out to generate new classification when historical unmarked electricity stealing categories exist in simulation data.
S43: and constructing a suspected electricity larceny identification model of a shop in the GRU algorithm aiming at the suspected electricity larceny sample set. And (3) extracting the electricity stealing sample set marked in the step (2) and the normal electricity using sample set based on the history collection sample set, and constructing a suspected electricity stealing identification model based on a GRU algorithm.
S44: calculating the recognition accuracy rate of the suspected electricity larceny identification model according to the output result of the suspected electricity larceny identification model and the class mark of the suspected electricity larceny collection, and carrying out feedback adjustment on model parameters; aiming at complex correlation of electricity stealing sample characteristic data and time sequence rules of electricity utilization data, a GRU algorithm is adopted to realize construction of an electricity stealing identification model, the electricity stealing behavior characteristic data of a high-risk electricity stealing sample set is mined, time sequence characteristics are taken as input, a neural network is used for extracting and retaining historical tracks of the electricity stealing sample electricity utilization data, a cross entropy loss function is constructed according to identification accuracy of actual positive and negative samples, and model parameters of the GRU are adjusted through gradient descent, so that capture of the electricity stealing behavior identification rule is completed, and comprehensive mining of the electricity stealing characteristic data is improved based on the time sequence rules.
S45: and determining an anti-electricity-stealing defense simulation model for self-adaptive class identification according to the adjusted model parameters. And (3) carrying out feedback adjustment on the model parameters according to the recognition accuracy, if the recognition accuracy is smaller than an accuracy threshold, carrying out parameter feedback adjustment, namely carrying out data feedback based on a simulation result, feeding back cluster recognition verification data to the step (3), adjusting marked sample data of the semi-supervised k-NNM, feeding back the recognition verification data of the electricity larceny sample to the suspected electricity larceny recognition model in the step (4), adjusting model parameters by adopting a cross entropy loss function, and if the recognition accuracy is larger than or equal to the accuracy threshold.
In this embodiment, S5 includes the steps of:
s51: the result election of each predicted electricity utilization user in each period is carried out; the P_DT, P_CNN and P_LSTM results of each predicted electricity utilization user in each period are selected, namely the P_DT, P_CNN and P_LSTM probabilities of the predicted users in each period are used as the most levels of the corresponding probability interval sections of the credit level (0, 1,2 and 3) to predict the credit level by a combination method of the users.
S52: predicting the model by a progressive method, namely predicting the model obtained in the step S51; the model obtained in S51 is first predicted by using m_dt to predict the user features to be predicted, namely: the selected intoP_DTfrom_DT (D_FeatureTraning) then selects to generate a new set of user characteristics above the preset probability value Gate_DT based on P_DT, namely: selecting into (d_featuretraining_dt) from (d_featuretraining) whereep_dt > =gate_dt; the model m_cnn is then used to predict (d_featuretracking_dt) progressively, namely: selectintop_cnn_dtfromm_cnn (d_featuretracking_dt); then, according to the P_CNN_DT, selecting and generating a new user feature set above the preset probability value gate_CNN, namely: selecting into (d_featuretraining_dt_cnn) from (d_featuretraining_dt) whereep_cnn_dt > =gate_cnn; the model m_lstm is then used to predict (d_featuretracking_dt_cnn) progressively, i.e.: selecting intoP_DT_CNN_LSTMfrom M_LSTM (D_FeatureTraining_DT_CNN); and then, using the P_DT_CNN_LSTM to correspondingly obtain the progressive method prediction credit rating of the D_Featuretracking_DT_CNN related user in the (0, 1) interval by the (0, 1,2, 3) value.
S53: combining and de-duplicating the predicted credit level users obtained by the combination method and the progressive method and then picking up the service system to perform early warning detection.
When the specific setting is carried out, S64 also comprises fee calculation, namely checking whether electricity is stolen according to the checking information, if so, calculating basic electricity fee and illegal fine according to electricity consumption data or electricity consumption phenomenon of a user and combining electricity stealing techniques and electricity stealing time.
Specifically, S61 includes one or more of power abnormality, electric quantity abnormality, line loss abnormality, phase abnormality, voltage abnormality, current abnormality, power factor abnormality, and the power abnormality includes: the power is super-capacity, the daily active power is changed abnormally, the power is unbalanced abnormally, and the electric quantity abnormality comprises: the daily electricity quantity change abnormality includes: the daily electricity suddenly drops abnormally, and the line loss abnormality includes: the line loss variation abnormality includes: line loss surge anomaly, current anomaly includes: abnormal current imbalance and abnormal current change, the abnormal current change includes: phase current dip anomalies, voltage anomalies include: the voltage under-voltage abnormality includes: the phase voltage undervoltage abnormality, the power factor abnormality includes: power factor variation anomalies, including: power factor dip anomalies.
In this embodiment, in S42, the process of excluding the normal sample from forming the suspected fraudulent use of electricity sample set and performing the class marking is: and clustering and dividing a part of marked historical collection sample sets by adopting a semi-supervised k-NNM algorithm, removing samples clustered with marked normal samples, forming a suspected electricity stealing sample set, wherein the suspected electricity stealing sample set comprises a part of marked electricity stealing sample clustered sample sets and unmarked sample clustered sample sets, and marking suspected electricity stealing category with the marked electricity stealing sample clustered sample sets.
In a specific setting, in S43, the process of constructing the suspected electricity larceny identification model based on the GRU algorithm for the suspected electricity larceny sample set is as follows: based on a historical collection sample set, a marked electricity stealing sample set and a normal electricity using sample set are extracted, a suspected electricity stealing identification model is built by adopting a GRU algorithm according to the time sequence rule of the electricity stealing sample characteristic data, the electricity stealing behavior characteristic data of the high-risk electricity stealing sample set is mined, the time sequence characteristic is taken as input, the historical track of the electricity stealing sample electricity using data is extracted and reserved by a neural network, a loss function is built according to the identification accuracy of an actual positive sample and an actual negative sample, and the model parameters are adjusted by gradient descent.
Further, the process of constructing the suspected electricity larceny identification model based on the GRU algorithm aiming at the suspected electricity larceny sample set comprises the following steps: based on a historical collection sample set, a marked electricity stealing sample set and a normal electricity using sample set are extracted, a suspected electricity stealing identification model is built by adopting a GRU algorithm according to the time sequence rule of the electricity stealing sample characteristic data, the electricity stealing behavior characteristic data of the high-risk electricity stealing sample set is mined, the time sequence characteristic is taken as input, the historical track of the electricity stealing sample electricity using data is extracted and reserved by a neural network, a loss function is built according to the identification accuracy of an actual positive sample and an actual negative sample, and the model parameters are adjusted by gradient descent.
The multimode database fusion calculation model working principle for instant anti-electricity-stealing detection comprises the following steps: based on exploration of historical electricity larceny samples, the electricity consumption characteristics of special-purpose power-change electricity larceny users are discovered, electricity consumption images of the users are generated by summarizing and inducing the electricity consumption characteristics, classical electricity larceny case libraries are formed by merging the historical electricity larceny users, electricity larceny fingerprints of the users are generated based on conversion rules according to the electricity consumption characteristics of the users, electricity larceny fingerprint libraries are generated by different electricity larceny methods, an electricity larceny identification model of the special-purpose power-change users is built based on the electricity larceny fingerprint libraries and the electricity consumption signals (electricity consumption data characteristics or electricity consumption phenomenon characteristics) of the users, and suspected electricity larceny methods are given out; therefore, by analyzing big data of the electricity larceny sample, customers for electricity larceny and illegal electricity consumption are accurately positioned, the anti-electricity larceny resource is optimized and integrated, and the smooth and efficient operation of the workflow is ensured. And (3) marking partial positive and negative samples on the basis of preprocessing the history acquired data, dividing the electricity stealing type and the normal type by adopting a semi-supervised k-NNM algorithm, generating a high-risk suspected electricity stealing sample type, and then constructing a two-classification suspected electricity stealing identification model by adopting a GRU (gate control circulation unit) algorithm, so that even if no mark exists in the new electricity stealing type, the identification can be carried out on the basis of the electricity stealing rule, and the identification probability of the electricity stealing is improved.
The foregoing is merely exemplary embodiments of the present application and is not intended to limit the scope of the present application, and various modifications and variations may be suggested to one skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application. It should be noted that: like reference numerals and letters denote like items in the following figures, and thus once an item is defined in one figure, no further definition or explanation thereof is necessary in the following figures.
Claims (10)
1. A multimode database fusion calculation model for instant anti-electricity-theft detection is characterized by comprising the following steps:
s1: respectively designing and realizing decision trees in a service database for acquiring electric energy and electric quantity;
s2: preprocessing the electricity consumption sample data of the user, and storing the preprocessed data into a historical sample database;
s3: selecting relevant characteristics of user electricity consumption data of the electric energy and electric quantity collection industry score data vehicle and actual credit rating of the user as a model training data set;
s4: performing marker screening on the database samples;
s5: model training is performed by a combination method, and model training is performed based on a real-time and historical electricity utilization data set;
s6: extracting electricity stealing characteristics from corresponding user electricity data or user phenomena according to the electricity stealing information, determining corresponding electricity stealing methods according to the corresponding electricity stealing characteristics, generating electricity using images of the users according to the electricity using data or the electricity stealing phenomena of the electricity stealing users in combination with the electricity stealing information, merging the electricity stealing users to form an electricity stealing case library, generating electricity stealing fingerprints of the users according to the electricity using data characteristics or the electricity stealing characteristics of the users, arranging different electricity stealing methods according to the electricity stealing fingerprints to generate electricity stealing fingerprint libraries, and establishing a special change user electricity stealing identification prediction model.
2. The multi-mode database fusion calculation model for instant anti-electricity-theft detection according to claim 1, wherein S6 comprises the steps of:
s61: the abnormal detection comprises the steps of acquiring power consumption data or power consumption phenomenon of a specially-changed user, detecting power consumption abnormality, capturing abnormal data, extracting characteristic quantity of power consumption data characteristics of the user if the abnormal data is captured, comparing and judging the acquired abnormal data with corresponding set standard data or comparison data, and calculating rule violation probability;
s62: if the electricity larceny probability reaches or exceeds the preset rule violation probability, identifying a suspected electricity larceny user according to a special change user electricity larceny identification prediction model, matching electricity consumption data characteristics of the suspected electricity larceny user with electricity larceny fingerprints in an electricity larceny fingerprint library, primarily judging a suspected electricity larceny mode and suspected electricity larceny time, and generating a suspected electricity larceny user list;
s63: the electricity larceny checking comprises the steps of generating an electricity larceny checking work list according to a suspected electricity larceny user list, issuing a dispatch work list for checking, obtaining checking information, recording the checking information according to a user, extracting electricity consumption data characteristics or electricity consumption phenomenon characteristics of the user according to the checking information of the user and electricity consumption data or electricity consumption phenomenon, and updating an electricity larceny case library or electricity larceny fingerprint library;
s64: and electricity stealing processing, namely obtaining inspection conditions according to the inspection information to perform electricity stealing processing.
3. The multi-mode database fusion calculation model for instant anti-electricity-theft detection according to claim 1, wherein S4 comprises the steps of:
s41: screening out an electricity stealing sample, classifying the existing electricity stealing category, and marking the existing electricity stealing sample and a normal sample by category;
s42: removing normal samples to form a suspected electricity larceny sample set, and marking the class;
s43: and constructing a suspected electricity larceny identification model of a shop in the GRU algorithm aiming at the suspected electricity larceny sample set.
4. The multi-mode database fusion calculation model for instant anti-electricity-theft detection of claim 1, wherein S4 further comprises the steps of:
s44: calculating the recognition accuracy rate of the suspected electricity larceny identification model according to the output result of the suspected electricity larceny identification model and the class mark of the suspected electricity larceny collection, and carrying out feedback adjustment on model parameters;
s45: and determining an anti-electricity-stealing defense simulation model for self-adaptive class identification according to the adjusted model parameters.
5. The multi-mode database fusion calculation model for instant anti-electricity-theft detection according to claim 1, wherein S5 comprises the steps of:
s51: the result election of each pre-measured electricity user in each period is carried out;
s52: predicting the model by a progressive method, namely predicting the model obtained in the step S51;
s53: combining and de-duplicating the predicted credit level users obtained by the combination method and the progressive method and then picking up the service system to perform early warning detection.
6. The multi-mode database fusion calculation model for instant anti-electricity-theft detection of claim 2, wherein S64 further comprises fee calculation, wherein whether electricity is stolen or not is verified according to the checking information, and if electricity is stolen, basic electricity fee and illegal fine are calculated according to electricity consumption data of a user or electricity consumption phenomenon and electricity stealing operation and electricity stealing time to carry out pursuit.
7. The multi-mode database fusion calculation model for instant anti-electricity-theft detection according to claim 2, wherein S61 comprises one or more of power anomaly, electric quantity anomaly, line loss anomaly, phase anomaly, voltage anomaly, current anomaly, power factor anomaly, power anomaly comprising: the power is super-capacity, the daily active power is changed abnormally, the power is unbalanced abnormally, and the electric quantity abnormality comprises: the daily electricity quantity change abnormality includes: the daily electricity suddenly drops abnormally, and the line loss abnormality includes: the line loss variation abnormality includes: line loss surge anomaly, current anomaly includes: abnormal current imbalance and abnormal current change, the abnormal current change includes: phase current dip anomalies, voltage anomalies include: the voltage under-voltage abnormality includes: the phase voltage undervoltage abnormality, the power factor abnormality includes: power factor variation anomalies, including: power factor dip anomalies.
8. A multi-mode database fusion calculation model for instant anti-theft detection according to claim 3, wherein in S42, the process of excluding normal samples from forming a suspected electricity-theft sample set and performing class marking is as follows: and clustering and dividing a part of marked historical collection sample sets by adopting a semi-supervised k-NNM algorithm, removing samples clustered with marked normal samples, forming a suspected electricity stealing sample set, wherein the suspected electricity stealing sample set comprises a part of marked electricity stealing sample clustered sample sets and unmarked sample clustered sample sets, and marking suspected electricity stealing category with the marked electricity stealing sample clustered sample sets.
9. The multi-mode database fusion calculation model for instant anti-electricity-theft detection according to claim 3, wherein in S43, the process of constructing the suspected electricity-theft identification model based on the GRU algorithm for the suspected electricity-theft sample set is as follows: based on a historical collection sample set, a marked electricity stealing sample set and a normal electricity using sample set are extracted, a suspected electricity stealing identification model is built by adopting a GRU algorithm according to the time sequence rule of the electricity stealing sample characteristic data, the electricity stealing behavior characteristic data of the high-risk electricity stealing sample set is mined, the time sequence characteristic is taken as input, the historical track of the electricity stealing sample electricity using data is extracted and reserved by a neural network, a loss function is built according to the identification accuracy of an actual positive sample and an actual negative sample, and the model parameters are adjusted by gradient descent.
10. The multi-mode database fusion calculation model for instant anti-electricity-theft detection of claim 9, wherein the process of constructing the suspected electricity-theft identification model based on the GRU algorithm for the suspected electricity-theft sample set is as follows: based on a historical collection sample set, a marked electricity stealing sample set and a normal electricity using sample set are extracted, a suspected electricity stealing identification model is built by adopting a GRU algorithm according to the time sequence rule of the electricity stealing sample characteristic data, the electricity stealing behavior characteristic data of the high-risk electricity stealing sample set is mined, the time sequence characteristic is taken as input, the historical track of the electricity stealing sample electricity using data is extracted and reserved by a neural network, a loss function is built according to the identification accuracy of an actual positive sample and an actual negative sample, and the model parameters are adjusted by gradient descent.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311362622.4A CN117390546A (en) | 2023-10-20 | 2023-10-20 | Multimode database fusion calculation model for instant anti-electricity-theft detection |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311362622.4A CN117390546A (en) | 2023-10-20 | 2023-10-20 | Multimode database fusion calculation model for instant anti-electricity-theft detection |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117390546A true CN117390546A (en) | 2024-01-12 |
Family
ID=89438645
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311362622.4A Pending CN117390546A (en) | 2023-10-20 | 2023-10-20 | Multimode database fusion calculation model for instant anti-electricity-theft detection |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117390546A (en) |
-
2023
- 2023-10-20 CN CN202311362622.4A patent/CN117390546A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110223196B (en) | Anti-electricity-stealing analysis method based on typical industry feature library and anti-electricity-stealing sample library | |
| CN107742127B (en) | Improved electricity stealing prevention intelligent early warning system and method | |
| CN106504116B (en) | Stability evaluation method based on power grid operation and transient stability margin index correlation | |
| CN110097297A (en) | A kind of various dimensions stealing situation Intellisense method, system, equipment and medium | |
| CN109583680B (en) | Power stealing identification method based on support vector machine | |
| CN105677791B (en) | For analyzing the method and system of the operation data of wind power generating set | |
| CN110458230A (en) | A kind of distribution transforming based on the fusion of more criterions is with adopting data exception discriminating method | |
| CN108765004A (en) | A method of user's electricity stealing is identified based on data mining | |
| CN110837866A (en) | XGboost-based electric power secondary equipment defect degree evaluation method | |
| CN111861211B (en) | System with double-layer anti-electricity-stealing model | |
| CN105894177A (en) | Decision-making-tree-algorithm-based analysis and evaluation method for operation risk of power equipment | |
| CN111126820A (en) | Electricity stealing prevention method and system | |
| CN110348683A (en) | The main genetic analysis method, apparatus equipment of electrical energy power quality disturbance event and storage medium | |
| CN110968703B (en) | Method and system for constructing abnormal metering point knowledge base based on LSTM end-to-end extraction algorithm | |
| CN117132025A (en) | Power consumption monitoring and early warning system based on multisource data fusion | |
| CN115034671A (en) | Secondary system information fault analysis method based on association rule and cluster | |
| CN115186935B (en) | Electromechanical device nonlinear fault prediction method and system | |
| CN117390546A (en) | Multimode database fusion calculation model for instant anti-electricity-theft detection | |
| CN112434955B (en) | Distribution network line operation risk sensing method based on multi-data fusion | |
| CN114626433A (en) | Fault prediction and classification method, device and system for intelligent electric energy meter | |
| CN114066219A (en) | Electricity stealing analysis method for intelligently identifying electricity utilization abnormal points under incidence matrix | |
| Lu et al. | Time series power anomaly detection based on Light Gradient Boosting Machine | |
| CN113191409A (en) | Method for detecting abnormal electricity consumption behaviors of residents through tag data expansion and deep learning | |
| CN113435494A (en) | Low-voltage resident user abnormal electricity utilization identification method and simulation system | |
| Hou et al. | Research on real power grid security and stability based on big data analysis method and online data of power grid |
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 |