CN112926686B - BRB and LSTM model-based power consumption anomaly detection method and device for big power data - Google Patents

BRB and LSTM model-based power consumption anomaly detection method and device for big power data Download PDF

Info

Publication number
CN112926686B
CN112926686B CN202110338988.2A CN202110338988A CN112926686B CN 112926686 B CN112926686 B CN 112926686B CN 202110338988 A CN202110338988 A CN 202110338988A CN 112926686 B CN112926686 B CN 112926686B
Authority
CN
China
Prior art keywords
confidence
abnormal
power consumption
user
brb
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
Application number
CN202110338988.2A
Other languages
Chinese (zh)
Other versions
CN112926686A (en
Inventor
刘威
陈成
卢涛
万磊
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Wuhan Institute of Technology
Original Assignee
Wuhan Institute of Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Wuhan Institute of Technology filed Critical Wuhan Institute of Technology
Priority to CN202110338988.2A priority Critical patent/CN112926686B/en
Publication of CN112926686A publication Critical patent/CN112926686A/en
Application granted granted Critical
Publication of CN112926686B publication Critical patent/CN112926686B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—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
    • G06F18/21—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
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/04—Inference or reasoning models
    • 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
    • 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)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Business, Economics & Management (AREA)
  • Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Economics (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Marketing (AREA)
  • Molecular Biology (AREA)
  • Biomedical Technology (AREA)
  • Public Health (AREA)
  • Water Supply & Treatment (AREA)
  • Human Resources & Organizations (AREA)
  • Biophysics (AREA)
  • Primary Health Care (AREA)
  • General Business, Economics & Management (AREA)
  • Tourism & Hospitality (AREA)
  • Strategic Management (AREA)
  • Supply And Distribution Of Alternating Current (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Testing Electric Properties And Detecting Electric Faults (AREA)

Abstract

The invention discloses a method and a device for detecting power consumption abnormality of big electric power data based on BRB and LSTM models, wherein the method specifically comprises the following steps: extracting fluctuation characteristics of the user power consumption and abnormal characteristics of a user power consumption curve from the power consumption big data; establishing a confidence rule reasoning BRB system, and performing confidence conversion on the sum of the electric quantity fluctuation coefficient and the burr width; according to a confidence rule base in the confidence rule reasoning BRB system, comparing the converted confidence degrees by adopting a evidence reasoning ER algorithm to obtain the confidence degree of each reference value in the abnormal output result of the NTL of the non-technical loss of the user; calibrating abnormal power consumption of non-technical loss NTL of a user; and establishing a long and short memory LSTM model based on the calibration data, effectively extracting and detecting abnormal electricity utilization characteristics by using the LSTM model, and finally accurately diagnosing NTL abnormal conditions. The invention can effectively identify abnormal electricity consumption.

Description

BRB and LSTM model-based power consumption anomaly detection method and device for big power data
Technical Field
The invention relates to the technical field of metering data detection, in particular to a method and a device for detecting power consumption abnormality of big electric power data based on BRB and LSTM models.
Background
In recent years, with the rising and popularization of "smart grids", power transmission and distribution losses in operation are receiving increasing attention, and the power transmission and distribution losses can be roughly divided into two major categories, namely Technical Loss (TL) and non-technical loss (NTL). The serious non-technical loss, namely the abnormal electricity stealing behavior of the user, brings great economic loss to the power grid industry. Compared with the national power consumption ratio of NTL of China such as India, brazil and the like, NTL of China is relatively low, but the total power consumption of China is huge, and the national power consumption is also in an ascending trend. Therefore, how to efficiently and rapidly detect the abnormal electricity utilization behavior of the user from a large amount of electric power data so as to make a decision by power supply network personnel has important research significance for improving economic benefit and promoting development and progress of the power grid.
In the case of large power data, the detection of NTL anomalies has been a problem in hot spots and difficulties in this field. At present, most of the NTL anomaly detection methods for large power data are based on data-driven detection methods, including a clustering-based method, a deep learning-based method and the like. Although the abnormal detection of the NTL can be well completed based on the methods, a large number of user electricity consumption data samples are needed to be used as supports in the methods, and particularly for the method based on deep learning, calibration of positive and negative sample sets is a key of the abnormal detection accuracy. Particularly for supervised learning, good data calibration, namely correct calibration of fault data samples and normal data samples, can restrict the network to extract the characteristics of the fault samples more effectively, thereby improving the detection accuracy of the network. However, in practical applications, a manual calibration method is generally adopted for data calibration, which is time-consuming and costly.
Disclosure of Invention
The invention mainly aims to provide an accurate and effective electricity utilization abnormality detection method for an intelligent power grid system, and designs a set of device capable of accurately diagnosing abnormal electricity utilization behavior of a user from large electric power data, so that non-technical loss (NTL) is reduced, and economic loss is brought to the electric power industry.
The technical scheme adopted by the invention is as follows:
the utility model provides a method for detecting power consumption abnormality of big electric data based on BRB and LSTM models, which is characterized by comprising the following steps:
s1, extracting a fluctuation characteristic of the electricity consumption of a user and an abnormal characteristic of a curve of the electricity consumption of the user from large data of the electricity consumption, wherein the fluctuation coefficient of the electricity consumption is used as the fluctuation characteristic of the electricity consumption, and the sum of burr widths is used for representing the abnormal characteristic of the curve of the electricity consumption;
s2, establishing a confidence rule reasoning BRB system, and performing confidence degree conversion on the sum of the electric quantity fluctuation coefficient and the burr width;
s3, reasoning a confidence rule base in the BRB system according to the confidence rule, and comparing the converted confidence degrees by adopting a evidence reasoning ER algorithm to obtain the confidence degree of each reference value in the NTL abnormal output result of the user non-technical loss;
s4, calibrating abnormal power consumption of the non-technical loss NTL of the user according to the trust degree of each reference value;
s5, establishing a long and short memory LSTM model based on the calibration data, effectively extracting and detecting abnormal electricity utilization characteristics by using the LSTM model, and finally accurately diagnosing NTL abnormal conditions.
In the above technical solution, in step S1:
coefficient of electric fluctuation CV:
in the above formula, N is the cumulative days, sigma is the standard deviation, mu is the mean value,for daily average electric quantity value, q i The single-day electricity consumption of the user on the i th day;
sum of burr widths M:
in the above, d i And d j Respectively representing a lower bound node and an upper bound node of the corresponding time of the boundary value around the burr;
with the above technical solution, step S2 specifically includes:
taking the electric quantity fluctuation coefficient CV and the burr width sum M as precondition attributes of the BRB system, and converting the input values of the precondition attributes into membership degrees of corresponding precondition attribute reference values;
precondition attribute B i The input value conversion formula of (a) is:
S(B i ,α i )＝{(l im ,β im ),m＝1,...,m i },i∈{1,...,N}
wherein S is the distribution of the input values of the precondition attributes, l im Representing input precondition attribute B i Is the mth reference value of beta im Is l im Confidence beta of (2) im Not less than 0, andm i for the number of reference values, alpha i An input value representing a precondition attribute;
after the membership distribution is obtained, activating the activation weight omega of the precondition attribute under all the confidence rules in the confidence rule base k The calculation of the activation weight under the kth confidence rule is obtained with the following formula:
in the above, beta ik Representing the confidence matching degree of the individual, which is the input confidence degree of the ith precondition attribute and is used for evaluating the parameter value under the K-th confidence rule For assessing joint match between whole precondition attributes and reference values, epsilon i Representing the initial attribute weight, ++>Representing relative attribute weights, ++>The rule weight of the kth rule is represented.
In connection with the above technical solution, step S3 specifically includes: and (3) gathering the input values of the precondition attributes under all confidence rules by adopting an ER algorithm, so as to obtain the trust degree of each reference value in the NTL result attributes of the power consumption of the user, and carrying out NTL abnormal calibration on the data by comparing the confidence rules.
In the technical scheme, the reference values of the precondition attributes are set to three levels, which are respectively: HR represents that the power consumption data is high in the possibility value of abnormality, MR represents that the possibility value is general, LR represents that the possibility value is low, and the input conversion process is as follows:
HR≥α i ≥MR
MR≥α i ≥LR
α i an input value representing a precondition attribute.
In step S4, two LSTM models are adopted to respectively process the abnormal power utilization sequence and the normal power utilization sequence.
The invention also provides a device for detecting the power consumption abnormality of the big electric power data based on the BRB and LSTM models, which comprises the following components:
and the electricity consumption data characteristic extraction unit is used for extracting the user electricity consumption electric characteristic and the user electricity consumption curve abnormal characteristic.
The BRB system unit is used for performing confidence conversion on the electric quantity fluctuation coefficient and the burr width sum; according to a confidence rule base in the confidence rule reasoning BRB system, comparing the converted confidence degrees by adopting a evidence reasoning ER algorithm to obtain the confidence degree of each reference value in the abnormal output result of the NTL of the non-technical loss of the user;
s4, calibrating abnormal power consumption of the non-technical loss NTL of the user according to the trust degree of each reference value;
and the LSTM model unit is used for effectively extracting and detecting the abnormal electricity utilization characteristics on the basis of the calibration data and finally diagnosing the abnormal condition of the user electricity utilization NTL.
The device also comprises a data storage unit, wherein the data storage unit is used for storing data required by the operation of the device and generated data, and the data comprise a power consumption fluctuation coefficient, a burr width coefficient, a confidence rule base, an NTL abnormal power consumption sequence and a normal power consumption sequence.
The invention also provides a computer storage medium which can be executed by a processor, wherein a computer program is stored in the computer storage medium, and the computer program executes the electric power consumption abnormality detection method of the BRB and LSTM models according to the technical scheme.
The invention has the beneficial effects that: the invention discloses a method for detecting power consumption abnormality of big power data based on BRB and LSTM models, which is characterized in that the BRB method is introduced into abnormality detection of the big power data, the abnormality of the power consumption of a user is detected by extracting the fluctuation characteristics of the power consumption of the user and the confidence rule reasoning method of the abnormality characteristics of the power consumption curve, and a positive and negative sample set with high reliability and robustness is automatically and quickly obtained. Based on the sample set, the LSTM module is adopted to effectively extract NTL abnormal characteristics, and finally, the detection of abnormal electricity utilization behaviors of the user is efficiently and accurately completed.
Drawings
The invention will be further described with reference to the accompanying drawings and examples, in which:
FIG. 1 is a flowchart of a method for detecting power consumption abnormality of big power data based on BRB and LSTM models according to the embodiment of the invention;
FIG. 2 is a schematic diagram I of an electrical power consumption abnormality detection device for big data of electric power based on BRB and LSTM models according to the embodiment of the invention;
fig. 3 is a schematic diagram II of an electric power consumption abnormality detection device for big electric power data based on BRB and LSTM models according to an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
In order to improve the efficiency of data calibration, a confidence rule reasoning (BRB) method is introduced into anomaly detection of large electric power data, and a confidence rule reasoning method based on fluctuation characteristics of electric power consumption of a user and anomaly characteristics of an electric power consumption curve is provided for detecting the anomaly of the electric power consumption of the user, so that a positive and negative sample set with high reliability and robustness is automatically and quickly obtained. Then, based on the sample set, an LSTM module is adopted to effectively extract NTL abnormal characteristics, and finally detection of abnormal electricity utilization behaviors of the user is completed.
As shown in FIG. 1, the method for detecting the power consumption abnormality of the big electric power data based on the BRB and LSTM models according to the embodiment of the invention comprises the following steps:
s1, extracting a fluctuation characteristic of the electricity consumption of a user and an abnormal characteristic of a curve of the electricity consumption of the user from large data of the electricity consumption, wherein the fluctuation coefficient of the electricity consumption is used as the fluctuation characteristic of the electricity consumption, and the sum of burr widths is used for representing the abnormal characteristic of the curve of the electricity consumption;
s2, establishing a confidence rule reasoning BRB system, and performing confidence degree conversion on the sum of the electric quantity fluctuation coefficient and the burr width;
s3, reasoning a confidence rule base in the BRB system according to the confidence rule, and comparing the converted confidence degrees by adopting a evidence reasoning ER algorithm to obtain the confidence degree of each reference value in the NTL abnormal output result of the user non-technical loss;
s4, calibrating abnormal power consumption of the non-technical loss NTL of the user according to the trust degree of each reference value;
s5, establishing a long and short memory LSTM model based on the calibration data, effectively extracting and detecting abnormal electricity utilization characteristics by using the LSTM model, and finally accurately diagnosing NTL abnormal conditions.
According to the electric power big data electricity consumption abnormality detection method based on the BRB and LSTM models in the other specific embodiment of the invention, the effective extraction and detection of abnormal electricity consumption characteristics are realized, and the NTL abnormal condition is finally and accurately diagnosed. The method specifically comprises the following steps:
1. and analyzing the user power consumption in a certain time range from the real power consumption big data, and extracting the fluctuation characteristics of the user power consumption and the abnormal characteristics of the user power consumption curve.
1) And extracting the fluctuation characteristics of the electricity consumption of the user. The electricity consumption of the user in a certain time range is analyzed, and the traditional indexes such as average value or variance cannot reflect the electricity fluctuation condition in the time period, so that the current data abnormal information cannot be represented. Therefore, in order to more effectively reflect the characteristics of the electricity consumption data, the electricity consumption fluctuation coefficient CV is selected to analyze the abnormal fluctuation of the electricity consumption data within the statistical time range, which is defined as:
in the formula (1), N is the cumulative days, sigma is the standard deviation, mu is the mean value,for daily average electric quantity value, q i The single day power consumption for the i-th day of the user. The CV represents the degree of dispersion of the electricity consumption, and the smaller the CV, the smaller the degree of dispersion of the representative sample, i.e., the smaller the degree of fluctuation of the electricity consumption.
2) And extracting abnormal characteristics of the power consumption curve of the user. Describing abnormal characteristics of the electricity consumption curve, and referring to the burr characteristics, namely drawing the reading of the ammeter into an upward peak appearing after the curve. The width of the burrs can intuitively measure the abnormal quantity, and the invention adopts the sum (M) of the widths of the burrs to represent the abnormal characteristics of the electricity consumption curve, which is defined as:
in the formula (2), d i And d j And respectively representing a lower bound node and an upper bound node of the boundary value around the burr corresponding to the time.
2. The BRB system is constructed. The method comprises the steps of taking a fluctuation characteristic of the electricity consumption of a user and an abnormal characteristic of a curve of the electricity consumption of the user as precondition attribute input values, converting the precondition attribute input values into membership degrees of corresponding precondition attribute reference values, and collecting the precondition attribute input values under all the belief rules by a trained belief rule base (establishment of the belief rule base, firstly, carrying out input conversion on training sample data according to a formula (4) and a formula (5), then, training a BRB system by combining formulas (6) to (8), and finally, converting the abnormal characteristic of the electricity consumption of test sample data into the system to obtain a final belief rule base. And carrying out abnormal data calibration on the sample data to obtain an NTL abnormal power utilization sequence and a normal power utilization sequence.
Table one: confidence rule base after training
1) And (5) converting the condition of information input. According to the electric quantity fluctuation coefficient CV and the burr width sum M which are described in the step 1 and serve as precondition attributes of the BRB system, the input values are converted into membership degrees of corresponding precondition attribute reference values. The input conversion of a precondition attribute value refers to firstly customizing a conversion rule, then converting the value into different confidence degrees according to the rule, and finally reassigning the confidence degrees to reference values when each precondition attribute is used for making decisions. Precondition attribute B i The input value conversion formula of (a) is:
S(B i ,α i )＝{(l im ,β im ),m＝1,...,m i },i∈{1,...,N} (3)
in the formula (3), S is the distribution of the input values of the precondition attributes, l im Representing input precondition attribute B i Is the mth reference value of beta im Is l im Confidence of (beta) im Not less than 0) can be calculated according to the following formula (8), andm i is the number of reference values.
Generally, when 0 < CV < 0.2, electricity consumption in the month is considered normal. In addition, through statistics of historical electricity consumption, when M is more than or equal to 5 and less than or equal to 10, the current ammeter is considered to be close to the normal ammeter. Accordingly, the present invention is set to: CV-Big (large electric quantity fluctuation, CV)>0.2 Normal (fluctuation of electric quantity is Normal, 0 < CV < 0.2), small (fluctuation of electric quantity is Small, CV)<0) The large and small electric quantity fluctuation means abnormal electricity consumption; M-Large (a Large number of M>10 Normal (number is Normal, 5.ltoreq.M.ltoreq.10), little (number is small, M<5). Next, the evaluation levels are assigned corresponding confidence levels beta im Finally, the allocated confidence coefficient is redistributed to different reference values l of the precondition attribute im Is a kind of medium. Here, the present invention sets the reference values to three levels [ High-Reliability (HR), medium-Reliability (MR), low-Reliability (LR) ] respectively]Where HR represents a high likelihood of abnormality in the electricity consumption data, MR represents a general likelihood, and LR represents a low likelihood. The conversion process of the input is as follows:
HR≥α i ≥MR
MR≥α i ≥LR
to complete the input conversion of the precondition attribute, it is also necessary to determine the reference value. The determination of the reference value may be made by an expert from the sample data. The invention converts the precondition attribute input value into the membership degree of the corresponding precondition attribute reference value, namely after the distribution of the precondition attribute reference value is determined, the degree that the sample input value belongs to the precondition attribute reference value is obtained after a sample input value can be calculated by a left-side grade formula.
2) And activating weight calculation. After obtaining the membership distribution in 1), all of the activities are required to be activatedWeights ω of rule precondition attributes k The result thus obtained is only trusted. In general, the calculation of the activation weight under the kth rule can be obtained with the following formula:
in formula (6), beta ik Representing individual confidence matches, for evaluating parameter values under the K-th rule For assessing joint match between whole precondition attributes and reference values, epsilon i Representing attribute weights.
3) And outputting NTL abnormality results. Under the confidence rule base, the abnormal result is inferred through the ER algorithm, and the output result generated by the ER algorithm is:
Out(X)＝S(B j )＝{(C j ,χ j ),j＝1,...,N} (7)
in the formula (7), χ j Representing reference value C j Is χ, is the final confidence of jk Representing the confidence of the jth level in the kth rule, the calculation formula is as follows:
the output result distribution is that NTL abnormality and electricity consumption are calibrated by using 0 and 1 as data respectively.
3. And (3) obtaining an NTL abnormal data calibration sample sequence by utilizing the output result in the step (2), training the NTL abnormal power consumption sequence and the normal power consumption sequence to obtain an LSTM model, and effectively extracting and detecting abnormal power consumption characteristics by using the LSTM module to finally accurately diagnose the NTL abnormal condition.
1) Calibrating NTL abnormal data from a real electricity utilization data set;
2) Carrying out effective feature extraction on the calibrated NTL abnormal electricity utilization sequence and the calibrated normal electricity utilization sequence, and establishing an LSTM model;
3) And carrying out NTL detection and judgment on actual user electricity by using an LSTM model to obtain an NTL electricity abnormal result.
As shown in fig. 2, the apparatus for detecting power consumption abnormality of big data of electric power based on BRB and LSTM models according to the embodiment of the present invention includes:
and the electricity consumption data characteristic extraction unit is used for extracting the user electricity consumption electric characteristic and the user electricity consumption curve abnormal characteristic.
The BRB system unit is used for performing confidence conversion on the electric quantity fluctuation coefficient and the burr width sum; obtaining the trust degree of the user electricity consumption attribute according to the confidence conversion result; adopting an evidence reasoning ER algorithm to deduct a reasoning rule to obtain an abnormal output result of the non-technical loss NTL of the user; calibrating abnormal power consumption of non-technical loss NTL of a user;
and the LSTM model unit is used for effectively extracting and detecting the abnormal electricity utilization characteristics on the basis of the calibration data and finally diagnosing the abnormal condition of the user electricity utilization NTL.
The device also comprises a data storage unit which is used for storing data required by the operation of the device and generated data, wherein the data comprise a power consumption fluctuation coefficient, a burr width coefficient, a confidence rule base, an NTL abnormal power consumption sequence and a normal power consumption sequence.
The present invention also provides a computer storage medium executable by a processor, in which a computer program is stored, the computer program executing the electric power big data electricity consumption abnormality detection method of the BRB and LSTM models of the above-described embodiment.
As shown in fig. 3, another embodiment of the present invention provides an electric power big data electricity consumption abnormality detection apparatus based on BRB and LSTM models, including:
and the electricity consumption data characteristic extraction unit is used for extracting the user electricity consumption electric characteristic and the user electricity consumption curve abnormal characteristic.
The data storage unit is used for storing data required by the operation of the system and generated data, and mainly stores the power consumption fluctuation Coefficient (CV), the burr width coefficient (M), the confidence rule base, the NTL abnormal power utilization sequence and the normal power utilization sequence.
And the BRB system unit is used for calibrating the NTL abnormal data of the power consumption data sample, and storing the calibrated data into the data storage unit to obtain an NTL abnormal power consumption sequence and a normal power consumption sequence.
And the LSTM model unit is used for effectively extracting and detecting the abnormal electricity utilization characteristics and finally diagnosing the abnormal condition of the user electricity utilization NTL.
The LSTM model unit is specifically used for (1) carrying out data calibration on real electricity consumption data according to the BRB model, wherein the calibrated categories are NTL abnormal electricity consumption and normal electricity consumption, and grouping the abnormal electricity consumption and the normal electricity consumption to obtain a training set and a testing set; (2) Establishing two LSTM sub-networks to respectively process abnormal power utilization sequences and normal power utilization sequences, and dividing a training set and a testing set by taking a time sequence as a unit according to the input dimension of the LSTM as the input of the network so as to acquire fault characteristics; and (3) obtaining the final LSMT model through the training. And extracting data from the user electricity consumption database to perform abnormal electricity consumption detection on the LSMT model.
Wherein the BRB system unit comprises:
an algorithm module for converting the precondition attributes;
and the ER algorithm reasoning module is used for carrying out data calibration through the confidence regulation library.
The LSTM model unit includes:
the module is used for training the NTL abnormal power utilization sequence and the normal power utilization sequence to generate an LSTM model;
and the module is used for testing the power consumption of the user through the LSTM model and diagnosing the abnormal condition of the NTL of the power consumption of the user.
In summary, the method and the device for detecting the power consumption abnormality of the big electric data based on the BRB and LSTM models improve the correct calibration data sample in the prior art, and extract two power consumption abnormality characteristics of the fluctuation coefficient of the electric quantity and the burr width of the electric quantity curve from the big electric quantity data; formulating attribute conversion of abnormal feature input preconditions based on a confidence rule reasoning (BRB), establishing a confidence rule base suitable for NTL abnormal detection, and outputting final confidence degree by a Evidence Reasoning (ER) method so as to automatically acquire a labeled positive and negative sample training data set with high robustness; based on the data set, a multi-length-short memory (LSTM) model is constructed, effective extraction and detection of abnormal electricity utilization characteristics are achieved, and NTL abnormal conditions are accurately diagnosed finally.
It will be understood that modifications and variations will be apparent to those skilled in the art from the foregoing description, and it is intended that all such modifications and variations be included within the scope of the following claims.

Claims (7)

1. The electric power consumption abnormality detection method for the big electric power data based on the BRB and LSTM models is characterized by comprising the following steps of:
s1, extracting a fluctuation characteristic of the user power consumption and an abnormal characteristic of a curve of the user power consumption from large data of the power consumption, wherein the fluctuation coefficient of the power consumption is used as the fluctuation characteristic of the user power consumption, and the abnormal characteristic of the curve of the user power consumption is represented by the sum of burr widths;
s2, establishing a confidence rule reasoning BRB system, and performing confidence degree conversion on the sum of the electric quantity fluctuation coefficient and the burr width;
s3, reasoning a confidence rule base in the BRB system according to the confidence rule, and comparing the converted confidence degrees by adopting a evidence reasoning ER algorithm to obtain the confidence degree of each reference value in the NTL abnormal output result of the user non-technical loss;
s4, calibrating abnormal power consumption of the non-technical loss NTL of the user according to the trust degree of each reference value;
s5, establishing a long and short memory LSTM model on the basis of calibration data, effectively extracting and detecting abnormal electricity utilization characteristics by using the LSTM model, and finally accurately diagnosing NTL abnormal conditions;
wherein, in step S1:
coefficient of electric quantity fluctuationCV：
In the above-mentioned method, the step of,for accumulating days, & gt>Is standard deviation (S)>For mean value->For daily average electric quantity value, +.>To the user's firstiDaily electricity consumption of a day;
sum of burr widthsM：
In the above-mentioned method, the step of,and->Respectively representing a lower bound node and an upper bound node of the corresponding time of the boundary value around the burr;
the step S2 specifically comprises the following steps:
coefficient of fluctuation of electric quantityCVSum of burr widthMAs precondition attribute of BRB system, and converting these input values of precondition attribute into membership grade of corresponding reference value of precondition attribute;
precondition attributesB i The input value conversion formula of (a) is:
wherein,Sto precondition the distribution of the attribute input values,representing input preconditions attribute +.>Is the mth reference value of->Is->Confidence of->And->，/>For the number of reference values +.>An input value representing a precondition attribute;
activating the activation weight of the precondition attribute under all confidence rules in the confidence rule base after the membership distribution is obtainedFor the firstkThe calculation of activation weights under the bar confidence rules is obtained with the following formula:
in the above-mentioned method, the step of,representing confidence match of individual, is the ith precondition attributeFor evaluating the parameter values under the K-th confidence rule +.>；/>For assessing the degree of joint matching between the entire precondition attribute and the reference value, < >>Representing the initial attribute weight, ++>Representing relative attribute weights, ++>The rule weight of the kth rule is represented.
2. The method for detecting electrical power consumption abnormality of BRB and LSTM models according to claim 1, wherein step S3 specifically comprises: and (3) gathering the input values of the precondition attributes under all confidence rules by adopting an ER algorithm, so as to obtain the trust degree of each reference value in the NTL result attributes of the power consumption of the user, and carrying out NTL abnormal calibration on the data by comparing the confidence rules.
3. The electric power consumption abnormality detection method for electric power big data of BRB and LSTM models according to claim 1, wherein the reference values of the precondition attributes are set to three levels, respectively:HRthe likelihood value that the representative electricity data is abnormal is high,MRthe value of the representative likelihood is generally set,LRthe representative likelihood value is low, and the input conversion process is as follows:
an input value representing a precondition attribute.
4. The method for detecting electrical power consumption abnormality of BRB and LSTM models according to any one of claims 1 to 3, wherein two LSTM models are adopted in step S4 to process the abnormal electrical power consumption sequence and the normal electrical power consumption sequence, respectively.
5. The utility model provides a big data electricity consumption anomaly detection device of electric power based on BRB and LSTM model which characterized in that includes:
the electricity consumption data characteristic extraction unit is used for extracting user electricity consumption electric wave characteristics and user electricity consumption curve abnormal characteristics, wherein an electricity consumption fluctuation coefficient is used as the user electricity consumption electric wave characteristics, and the burr width sum is used for representing the user electricity consumption curve abnormal characteristics;
coefficient of electric quantity fluctuationCV：
In the above-mentioned method, the step of,for accumulating days, & gt>Is standard deviation (S)>For mean value->For daily average electric quantity value, +.>To the user's firstiDaily electricity consumption of a day;
sum of burr widthsM：
In the above-mentioned method, the step of,and->Respectively representing a lower bound node and an upper bound node of the corresponding time of the boundary value around the burr;
the BRB system unit is used for performing confidence conversion on the electric quantity fluctuation coefficient and the burr width sum; according to a confidence rule base in the confidence rule reasoning BRB system, comparing the converted confidence degrees by adopting a evidence reasoning ER algorithm to obtain the confidence degree of each reference value in the abnormal output result of the NTL of the non-technical loss of the user; specifically, the electric quantity fluctuation coefficientCVSum of burr widthMAs precondition attribute of BRB system, and converting these input values of precondition attribute into membership grade of corresponding reference value of precondition attribute;
precondition attributesB i The input value conversion formula of (a) is:
wherein,Sto precondition the distribution of the attribute input values,representing input preconditions attribute +.>Is the mth reference value of->Is->Confidence of->And->，/>For the number of reference values +.>An input value representing a precondition attribute;
activating the activation weight of the precondition attribute under all confidence rules in the confidence rule base after the membership distribution is obtainedFor the firstkThe calculation of activation weights under the bar confidence rules is obtained with the following formula:
in the above-mentioned method, the step of,representing the confidence match of the individual, is the input confidence of the ith precondition attribute, and is used for evaluating the parameter value under the K confidence rule +.>；/>For assessing the degree of joint matching between the entire precondition attribute and the reference value, < >>Representing the initial attribute weight, ++>Representing relative attribute weights, ++>Rule weights representing the kth rule; calibrating the abnormal power consumption of the non-technical loss NTL of the user according to the trust degree of each reference value;
and the LSTM model unit is used for effectively extracting and detecting the abnormal electricity utilization characteristics on the basis of the calibration data and finally diagnosing the abnormal condition of the user electricity utilization NTL.
6. The BRB and LSTM model-based power big data electricity consumption abnormality detection device according to claim 5, further comprising a data storage unit for storing data required for operation of the device and generated data including a power consumption fluctuation coefficient, a burr width coefficient, a confidence rule base, an NTL abnormal electricity consumption sequence and a normal electricity consumption sequence.
7. A computer storage medium executable by a processor and having stored therein a computer program for performing the electrical power big data anomaly detection method of BRB and LSTM models according to any one of claims 1 to 3.
CN202110338988.2A 2021-03-30 2021-03-30 BRB and LSTM model-based power consumption anomaly detection method and device for big power data Active CN112926686B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110338988.2A CN112926686B (en) 2021-03-30 2021-03-30 BRB and LSTM model-based power consumption anomaly detection method and device for big power data

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110338988.2A CN112926686B (en) 2021-03-30 2021-03-30 BRB and LSTM model-based power consumption anomaly detection method and device for big power data

Publications (2)

Publication Number Publication Date
CN112926686A CN112926686A (en) 2021-06-08
CN112926686B true CN112926686B (en) 2023-11-17

Family

ID=76176551

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110338988.2A Active CN112926686B (en) 2021-03-30 2021-03-30 BRB and LSTM model-based power consumption anomaly detection method and device for big power data

Country Status (1)

Country Link
CN (1) CN112926686B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114205855A (en) * 2021-10-25 2022-03-18 国网天津市电力公司电力科学研究院 Feeder automation service network anomaly detection method facing 5G slices
CN117290797B (en) * 2023-11-24 2024-02-02 国网山东省电力公司济宁供电公司 Building energy consumption prediction method, system, device and medium

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108416693A (en) * 2018-02-01 2018-08-17 东华大学 A kind of multiplexing electric abnormality Activity recognition method based on fuzzy neural network
CN111047094A (en) * 2019-12-12 2020-04-21 国网浙江省电力有限公司 Meter reading data anomaly analysis method based on deep learning algorithm
CN111612651A (en) * 2020-05-27 2020-09-01 福州大学 Abnormal electric quantity data detection method based on long-term and short-term memory network
CN111695639A (en) * 2020-06-17 2020-09-22 浙江经贸职业技术学院 Power consumer power consumption abnormity detection method based on machine learning

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108416693A (en) * 2018-02-01 2018-08-17 东华大学 A kind of multiplexing electric abnormality Activity recognition method based on fuzzy neural network
CN111047094A (en) * 2019-12-12 2020-04-21 国网浙江省电力有限公司 Meter reading data anomaly analysis method based on deep learning algorithm
CN111612651A (en) * 2020-05-27 2020-09-01 福州大学 Abnormal electric quantity data detection method based on long-term and short-term memory network
CN111695639A (en) * 2020-06-17 2020-09-22 浙江经贸职业技术学院 Power consumer power consumption abnormity detection method based on machine learning

Also Published As

Publication number Publication date
CN112926686A (en) 2021-06-08

Similar Documents

Publication Publication Date Title
CN112699913B (en) Method and device for diagnosing abnormal relationship of household transformer in transformer area
CN111460728B (en) Method and device for predicting residual life of industrial equipment, storage medium and equipment
US20120304008A1 (en) Supervised fault learning using rule-generated samples for machine condition monitoring
CN112926686B (en) BRB and LSTM model-based power consumption anomaly detection method and device for big power data
CN107807860B (en) Power failure analysis method and system based on matrix decomposition
CN112507479B (en) Oil drilling machine health state assessment method based on manifold learning and softmax
CN116432123A (en) Electric energy meter fault early warning method based on CART decision tree algorithm
CN110968703B (en) Method and system for constructing abnormal metering point knowledge base based on LSTM end-to-end extraction algorithm
CN111612149A (en) Main network line state detection method, system and medium based on decision tree
CN116011158A (en) Topology identification method, system and device for low-voltage transformer area
CN114595762A (en) Photovoltaic power station abnormal data sequence extraction method
CN114581699A (en) Transformer state evaluation method based on deep learning model in consideration of multi-source information
CN112595918A (en) Low-voltage meter reading fault detection method and device
CN116933201A (en) Method and system for identifying illegal electricity utilization behavior of low-voltage charging pile
CN116879672A (en) Intelligent identification method for total fault of area based on time convolution network
CN116662840A (en) Low-voltage station user phase identification method based on machine learning
CN114046816B (en) Sensor signal fault diagnosis method based on lightweight gradient lifting decision tree
CN114358160A (en) Data anomaly detection method in power system
CN115146715A (en) Power utilization potential safety hazard diagnosis method, device, equipment and storage medium
CN115598459A (en) Power failure prediction method for 10kV feeder line fault of power distribution network
CN113435915B (en) Method, device, equipment and storage medium for detecting electricity stealing behavior of user
Lu et al. Smart Meter Fault Diagnosis Model Based on DBN-LSSVM Feature Fusion
CN118036886A (en) Dynamic safety margin assessment method for novel electric-gas interconnection power system
CN117493923A (en) Method and system for repairing abnormal data of low-voltage distribution transformer area containing distributed photovoltaic
Zou et al. Maintenance Evaluation of Power Grid Fault Equipment Based on Random Forest

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