CN112926687A - User abnormal electricity utilization detection method based on PCANet and WNN - Google Patents

User abnormal electricity utilization detection method based on PCANet and WNN Download PDF

Info

Publication number
CN112926687A
CN112926687A CN202110339366.1A CN202110339366A CN112926687A CN 112926687 A CN112926687 A CN 112926687A CN 202110339366 A CN202110339366 A CN 202110339366A CN 112926687 A CN112926687 A CN 112926687A
Authority
CN
China
Prior art keywords
pcanet
user
mapping
stage
wnn
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
Application number
CN202110339366.1A
Other languages
Chinese (zh)
Inventor
刘威
江锐
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 CN202110339366.1A priority Critical patent/CN112926687A/en
Publication of CN112926687A publication Critical patent/CN112926687A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • 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
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
    • 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
    • G06F18/24—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
    • 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
    • G06N3/045—Combinations of networks
    • 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
    • G06N3/048—Activation functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Health & Medical Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Business, Economics & Management (AREA)
  • General Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Biomedical Technology (AREA)
  • Economics (AREA)
  • Water Supply & Treatment (AREA)
  • Public Health (AREA)
  • Human Resources & Organizations (AREA)
  • Marketing (AREA)
  • Primary Health Care (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a user abnormal electricity utilization detection method based on PCANet and WNN, which comprises the following steps: s1, acquiring power consumption data of a user within a certain time range, and performing data preprocessing on the power consumption data; s2, extracting the electric signal characteristics of the preprocessed data, and performing multiple dimensionality reduction processing on the power utilization characteristics by adopting principal component analysis to obtain final power utilization characteristic output; and S3, carrying out correlation mapping on the final electricity utilization characteristics through a wavelet neural network WNN model, and detecting the abnormal electricity utilization behavior of the user. The invention can accurately detect the abnormal electricity utilization of the user, prompts the system to warn the user with the abnormal electricity utilization, fully ensures to reduce the non-technical loss in the electricity utilization process and improves the economic benefit.

Description

User abnormal electricity utilization detection method based on PCANet and WNN
Technical Field
The invention relates to the technical field of electric power stealing detection, in particular to a method and a system for detecting abnormal electricity consumption of a user based on PCANet and WNN.
Background
In an electric power system, non-technical loss in transmission and distribution loss is always a problem expected to be solved by power grid enterprises of various countries. The method refers to economic loss caused by abnormal electricity utilization behaviors of users, such as electricity stealing behaviors of the users and the like, and the normal dispatching of regional power grids is seriously influenced. According to incomplete statistics, the growth rate of non-technical loss (11%) in other developing countries is much higher than the GDP acceleration rate (4% -5%) in the countries. Especially in brazil, india and mexico, the percentage of non-technical losses in the total electricity used nationwide reach astonishingly 16.85%, 26.2% and 15.83%, respectively. Although the non-technical loss occupies a relatively low proportion in the power grid transmission and distribution in China, the power consumption is continuously increased due to the fact that the demand for electric quantity in China is large overall and the living standard of people is continuously improved along with the continuous development of economy, and the loss caused by the continuous increase is considerable. It is reported that the economic loss caused by electricity stealing is as high as 1 billion yuan RMB every year in the province of Chinese Fujian alone. Therefore, the detection research on the abnormal electricity utilization behavior mode of the user is particularly important, and the method is one of important means for reducing non-technical loss and improving economic benefit.
Pca (principal Components analysis), which is a "projection" technique, maps data in a high-dimensional space to a low-dimensional space. So that we can see some high dimensional signals while removing some "noise" and some unwanted features. The dimensionality reduction of the power utilization characteristics is realized by adopting principal component analysis, the problem of low characteristic extraction efficiency caused by information overlapping in a network is solved, and meanwhile, the detection accuracy of the model on abnormal power utilization is greatly improved. The principal component analysis network (PCANet) of PCA has been developed and matured, and meanwhile, the PCA is an unsupervised feature extraction method which is mainly applied to a plurality of image classification tasks and obtains a good classification effect. The wavelet neural network replaces the activation function of the neural network, such as a Sigmod function, with a wavelet function, and the corresponding weights and activation thresholds from the input layer to the hidden layer are replaced by the scale expansion factor and the time shift factor of the wavelet function, which is the most widely used structure, i.e., a compact structure. The wavelet neural network has the obvious advantages that firstly, the elements and the whole structure of the wavelet neural network are determined according to the wavelet analysis theory, thereby avoiding the blindness in structural design of BP neural network and the like; and secondly, the wavelet neural network has stronger learning ability and higher precision. Because the wavelet theory is full-scale analysis, not only is a global optimal solution, but also a local detail optimal solution is kept, and in general, for the same learning task, the wavelet neural network has a simpler structure, higher convergence rate and higher precision. Compared with the SVM, the wavelet neural network has better characteristic expression capability and is more suitable for extracting and classifying the data sequence characteristics. The invention combines the advantages of the two to detect the abnormal power utilization mode of the user, which is not related to the prior art.
Disclosure of Invention
The invention mainly aims to provide a method and a system for detecting abnormal power consumption modes of users based on PCANet and WNN, which solve the problems that the feature extraction efficiency is low due to the overlapping of information and the data lack of training samples caused by the difficult-to-obtain labeled data at present, so that the users with abnormal power consumption can be effectively screened out.
The technical scheme adopted by the invention is as follows:
the method for detecting the abnormal electricity utilization of the user based on the PCANet and the WNN comprises the following steps:
s1, acquiring power consumption data of a user within a certain time range, and performing data preprocessing on the power consumption data;
s2, extracting electric signal characteristics of the preprocessed data, and performing multi-time dimensionality reduction processing on the power utilization characteristics by adopting PCANet, wherein the method is specifically divided into three-stage PCA mapping:
the first stage of PCANet mapping, namely acquiring sample fragments of preprocessed data by adopting a sliding window, and performing mean value removal processing on sample vectors in all windows and minimizing reconstruction errors;
and PCA network structure processing in the second stage: taking the output of the first stage as input, and repeating the PCANet mapping process of the first stage;
outputting mapping at the third stage, namely processing the feature vector by adopting a HervieSeld step function, binarizing the feature vector, dividing the binarized feature vector into a plurality of blocks, performing histogram statistical calculation on each block, and integrating all histograms into one vector to obtain final power utilization feature output;
and S3, carrying out correlation mapping on the final electricity utilization characteristics through a wavelet neural network WNN model, and detecting the abnormal electricity utilization behavior of the user.
The technical scheme is that the data are preprocessed by an interpolation method and a normalization method:
and (3) processing missing values in the data by adopting an interpolation method:
Figure BDA0002998936780000031
in the formula, xiRepresenting the electricity consumption of a user within a certain time range, NaN being xiIs a null value or is a non-numeric value;
the following formula is utilized to carry out normalization operation, and the detection precision of the network is improved:
Figure BDA0002998936780000032
x is the entire sample data set, min (X) is the minimum value in the data set, and max (X) is the maximum value in the data set.
In connection with the above technical solution, the first stage PCANet mapping specifically includes:
for each sample SiBy a size w1×w2Obtaining a sample fragment by sliding a window of:
Figure BDA0002998936780000041
wherein xi，jRepresenting S in jth vectorization windowiThe ith sample vector is obtained by performing mean value removal processing on the sample vectors in all windows:
Figure BDA0002998936780000042
the reconstruction error is then minimized with the following equation:
Figure BDA0002998936780000043
Figure BDA0002998936780000044
is L1×L1The solution of the above formula is UUTL of1And (4) a main feature vector.
In the third stage of output mapping, the PCANet processes the feature vector by using the hervesaide step function, and binarizes the feature vector:
Figure BDA0002998936780000045
h (-) represents a Hervesaide step function, wherein each feature vector is constrained to
Figure BDA0002998936780000046
Within the range of (1);
Figure BDA0002998936780000047
a filter that is a first stage;
Figure BDA0002998936780000048
is the output of the ith filter for the ith image. For each data sample SiOutput the corresponding output
Figure BDA0002998936780000049
Dividing the block units into B block units, performing histogram statistical calculation on each block unit, integrating all histograms into a vector, and obtaining a final feature output result:
Figure BDA00029989367800000410
here, Bhist (. cndot.) represents the statistical calculation of the histogram for each block unit, fiRepresents the sample S by PCANetiAnd extracting the final feature vector.
In the above technical solution, step S3 specifically includes:
selecting a wavelet neural network as the correlation mapping from the series of characteristics to the user electricity consumption abnormal behavior detection, wherein a network mapping equation can be expressed as follows:
Figure BDA0002998936780000051
in the above formula, M and N represent the mapping number of input and output respectively, and represent M input features and N output features; sigmak(t) represents the kth input feature, θ (-) is the weight function from the input layer to the hidden layer, and q is the number of hidden layers; omegaijRepresenting the connection rights of the hidden layer to the output layer, ajRepresenting the scale of the wavelet function, bjRepresenting the shift of the wavelet function and,
Figure BDA0002998936780000052
representing a Sigmoid activation function.
According to the technical scheme, the method further comprises the step of S4, generating early warning information according to the detected abnormal electricity utilization behavior result of the user, and reminding the abnormal user of electricity utilization.
The invention also provides a system for detecting abnormal electricity consumption of users based on PCANet and WNN, which comprises:
the data preprocessing module is used for preprocessing the acquired power consumption data of the user within a certain time range;
the feature extraction module adopts a principal component analysis network to perform dimensionality reduction processing on the electrical signal through three times of PCA mapping so as to extract electrical signal features, and the feature extraction module is specifically divided into three stages of PCA mapping: the first stage of PCANet mapping, namely acquiring sample fragments of preprocessed data by adopting a sliding window, and performing mean value removal processing on sample vectors in all windows and minimizing reconstruction errors; and PCA network structure processing in the second stage: taking the output of the first stage as input, and repeating the PCANet mapping process of the first stage; outputting mapping at the third stage, namely processing the feature vector by adopting a HervieSeld step function, binarizing the feature vector, dividing the binarized feature vector into a plurality of blocks, performing histogram statistical calculation on each block, and integrating all histograms into one vector to obtain final power utilization feature output;
and the user electricity utilization abnormal behavior detection module is used for performing correlation mapping on the final electricity utilization characteristics through a wavelet neural network WNN model to detect the abnormal electricity utilization behavior of the user.
In connection with the above technical solution, the system further comprises:
and the early warning module is used for generating early warning information according to the detected abnormal electricity utilization behavior result of the user and reminding the abnormal electricity utilization of the user.
According to the technical scheme, the data preprocessing module specifically utilizes an interpolation method and a normalization method to preprocess the power consumption data.
The invention also provides a computer storage medium which can be executed by a processor and in which a computer program is stored, wherein the computer program executes the method for detecting abnormal electricity consumption of the user based on the PCANet and the WNN.
The invention has the following beneficial effects: this patent combines PCANet and WNN to detect user's unusual power consumption mode, reduces the dimension through adopting principal component analysis PCANet to carry out many times to the power consumption characteristic and handles, and the final power consumption characteristic that will reduce the dimension and handle and obtain carries out the associative mapping through wavelet neural network WNN model, detects out user's unusual power consumption action, has higher robustness and detection precision.
Drawings
The invention will be further described with reference to the accompanying drawings and examples, in which:
FIG. 1 is a flow chart of a method for detecting abnormal electricity consumption of a user based on PCANet and WNN according to an embodiment of the present invention;
FIG. 2 is a first schematic structural diagram of a user abnormal electricity consumption detection system based on PCANet and WNN according to an embodiment of the present invention;
fig. 3 is a schematic structural diagram of a user abnormal electricity consumption detection system based on PCANet and WNN according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
As shown in fig. 1, the method for detecting abnormal electricity consumption of a user based on a PCANet (principal component analysis network) and a WNN (wavelet neural network) in the embodiment of the present invention includes the following steps:
s1, acquiring power consumption data of a user within a certain time range, and performing data preprocessing on the power consumption data;
s2, extracting the electric signal characteristics of the preprocessed data, and performing multi-time dimensionality reduction processing on the electric signal characteristics by adopting principal component analysis, wherein the method is specifically divided into three-stage PCA mapping:
the first stage of PCANet mapping, namely acquiring sample fragments of preprocessed data by adopting a sliding window, and performing mean value removal processing on sample vectors in all windows and minimizing reconstruction errors;
and PCA network structure processing in the second stage: taking the output of the first stage as input, and repeating the PCANet mapping process of the first stage;
outputting mapping at the third stage, namely processing the feature vector by adopting a HervieSeld step function, binarizing the feature vector, dividing the binarized feature vector into a plurality of blocks, performing histogram statistical calculation on each block, and integrating all histograms into one vector to obtain final power utilization feature output;
and S3, performing correlation mapping on the final electricity utilization characteristics through a model trained by a wavelet neural network WNN, and detecting the abnormal electricity utilization behavior of the user.
Further, the method also comprises a step S4 of generating early warning information according to the result of the detected abnormal electricity utilization behavior of the user to remind the abnormal user of electricity utilization.
The method of the embodiment of the invention can meet the early warning requirement of the feature extraction of the PCANet and the WNN network, and has the advantages of high precision, good reliability, higher generalization and the like; in addition, the provided user electricity utilization abnormity detection has good guiding significance for reducing non-technical loss and improving economic benefit, and has strong practicability and strong popularization and use values.
As shown in fig. 2, a preferred embodiment of the invention is based on a feature extraction method of PCANet and WNN networks, which preprocesses data by interpolation and normalization methods by establishing an electricity-related feature library. The method for reducing the influence of errors and missing data in the data set on the network comprises the following specific steps:
(1) and (3) processing missing values in the data by adopting an interpolation method:
Figure BDA0002998936780000081
in the formula, xiRepresenting the electricity consumption of a user within a certain time range, psi (x)i) Representing the results after treatment, NaN means xiIs a null value or is a non-numeric value.
(2) The following formula is utilized to carry out normalization operation, and the detection precision of the network is improved:
Figure BDA0002998936780000082
here, X is the entire sample data set, min (X) is the minimum value in the data set, and max (X) is the maximum value in the data set.
(3) Electrical correlation feature extraction, which comprises adopting principal component analysis to realize dimension reduction of electrical characteristics,
a total of three phases of PCA mapping:
1) first phase PCA mapping this phase is the input to the WNN networkAnd (4) processing the interlayer and the intermediate layer. For each sample SiBy a size w1×w2Obtaining a sample fragment by sliding a window of:
Figure BDA0002998936780000083
where x isi，jRepresenting S in jth vectorization windowiR represents the real number domain. And (3) carrying out mean value removal processing on the sample vectors in all the windows to obtain:
Figure BDA0002998936780000084
thus, for all data samples, the same matrix as described above is constructed and merged together to yield
Figure BDA0002998936780000085
L for the ith layeriThe filters, PCA, need to minimize the reconstruction error in these quadrature filters:
Figure BDA0002998936780000091
in the formula (3), the reaction mixture is,
Figure BDA0002998936780000092
is L1×L1The identity matrix of (2). The solution to the above formula is UUTL of1And (4) a main feature vector. To this end, the PCA filter is expressed as:
Figure BDA0002998936780000093
in the formula (4), the reaction mixture is,
Figure BDA0002998936780000094
representing a projection function, hereProjecting the vector onto a matrix Wl 1；pl(UUT) Representative UUTL of the principal eigenvector, L1Each column of a feature vector (each column containing k)1k2Individual elements).
2) The PCA network architecture and processing of the second stage is identical to that of the first stage, except that the inputs to the second stage are the outputs of the first stage.
Figure BDA0002998936780000095
A filter that is a first stage;
Figure BDA0002998936780000096
is the output of the ith filter for the ith image. SiIs the input of the ith filter for the ith image. The PCA output of the first phase is defined here as:
Figure BDA0002998936780000097
similar to the first stage, the output matrix after all filter dequantization is integrated and defined as:
Figure BDA0002998936780000098
wherein the content of the first and second substances,
Figure BDA0002998936780000099
in the second stage, the same process as the first stage is repeated, and the operations such as the averaging and the like are carried out on the output vector of the first stage, so that l of the second stage is obtained2A filter
Figure BDA00029989367800000910
Thus, the second stage PCA filter is expressed as:
Figure BDA00029989367800000911
in the formula (7), the reaction mixture is,
Figure BDA00029989367800000912
represents a projection function where the sample vector is projected onto a matrix W; p is a radical ofl(YYT) Represents YYTThe ith principal feature vector of (1). Y is like the X matrix of the first stage, YTIs a transposed matrix L of the matrix Y2Representing the number of filters of the PCA in the second stage, the final second stage output L1L2The number of feature vectors is determined by the number of feature vectors,
Figure BDA00029989367800000913
for the output of the second stage where the ith image corresponds to the ith output of the first stage,
Figure BDA00029989367800000914
a filter that is a first stage;
Figure BDA00029989367800000915
is the output of the ith filter for the ith image. Therein contains L2The convolution result has the expression:
Figure BDA0002998936780000101
3) in order to enhance the robustness of the network, the PCANet adopts a Hervesaide step function to process the feature vector, and binarizes the feature vector:
Figure BDA0002998936780000102
in the formula (9), H (-) represents a Hervesaide step function in which each feature vector is constrained
Figure BDA0002998936780000103
Within the range of (1). Finally, for each data sample SiOutput the corresponding output
Figure BDA0002998936780000104
Dividing the block units into B block units, performing histogram statistical calculation on each block unit, integrating all histograms into a vector, and obtaining a final feature output result:
Figure BDA0002998936780000105
here, Bhist (. cndot.) represents the statistical calculation of the histogram for each block unit, fiRepresents the sample S by PCANetiAnd extracting the final feature vector.
(4) And (3) correlation mapping, namely selecting a wavelet neural network as correlation mapping from series characteristics to user power consumption abnormal behavior detection, wherein a network mapping equation can be expressed as follows:
Figure BDA0002998936780000106
in the formula (11), M and N represent the mapping number of input and output, i.e. M input features and N output features, respectively; sigmak(t) represents the kth input feature,
Figure BDA0002998936780000107
for the final output result, theta (-) is a weight function from the input layer to the hidden layer, and q is the number of the hidden layers; in addition, ωijRepresenting the connection rights of the hidden layer to the output layer, ajRepresenting the scale of the wavelet function, bjRepresenting the shift of the wavelet function and,
Figure BDA0002998936780000108
representing a Sigmoid activation function.
As shown in fig. 3, the system for detecting abnormal electricity consumption of a user based on PCANet and WNN in the embodiment of the present invention is mainly used for implementing the method in the embodiment, and the system specifically includes:
the data preprocessing module is used for preprocessing the acquired power consumption data of the user within a certain time range; specifically, interpolation and normalization methods can be used to preprocess the power consumption data.
The feature extraction module adopts a principal component analysis network to perform dimensionality reduction processing on the electrical signal through three times of PCA mapping so as to extract electrical signal features, and the feature extraction module is specifically divided into three stages of PCA mapping: the first stage of PCANet mapping, namely acquiring sample fragments of preprocessed data by adopting a sliding window, and performing mean value removal processing on sample vectors in all windows and minimizing reconstruction errors; and PCA network structure processing in the second stage: taking the output of the first stage as input, and repeating the PCANet mapping process of the first stage; outputting mapping at the third stage, namely processing the feature vector by adopting a HervieSeld step function, binarizing the feature vector, dividing the binarized feature vector into a plurality of blocks, performing histogram statistical calculation on each block, and integrating all histograms into one vector to obtain final power utilization feature output;
and the user electricity utilization abnormal behavior detection module is used for performing correlation mapping on the final electricity utilization characteristics through a wavelet neural network WNN model to detect the abnormal electricity utilization behavior of the user.
The system further comprises:
and the early warning module is used for generating early warning information according to the detected abnormal electricity utilization behavior result of the user and reminding the abnormal electricity utilization of the user.
The present invention also provides a computer storage medium, which can be executed by a processor, and in which a computer program is stored, the computer program executing the method for detecting abnormal electricity consumption of a user based on PCANet and WNN of the above embodiments. The detailed description of the method is omitted here.
In conclusion, the method utilizes the PCANet network structure to extract effective sequence characteristics from the user electricity utilization data; and then, completing space mapping from the effective sequence characteristics to the user electricity consumption behavior mode through a wavelet neural network, and realizing the detection of abnormal electricity consumption. The invention can accurately detect the abnormal electricity utilization of the user, prompts the system to warn the user with the abnormal electricity utilization, fully ensures to reduce the non-technical loss in the electricity utilization process and improves the economic benefit.
It will be understood that modifications and variations can be made by persons skilled in the art in light of the above teachings and all such modifications and variations are intended to be included within the scope of the invention as defined in the appended claims.

Claims (10)

1. A method for detecting abnormal electricity utilization of a user based on PCANet and WNN is characterized by comprising the following steps:
s1, acquiring power consumption data of a user within a certain time range, and performing data preprocessing on the power consumption data;
s2, extracting electric signal characteristics of the preprocessed data, and performing multi-time dimensionality reduction processing on the power utilization characteristics by adopting PCANet, wherein the method is specifically divided into three-stage PCA mapping:
the first stage of PCANet mapping, namely acquiring sample fragments of preprocessed data by adopting a sliding window, and performing mean value removal processing on sample vectors in all windows and minimizing reconstruction errors;
and PCA network structure processing in the second stage: taking the output of the first stage as input, and repeating the PCANet mapping process of the first stage;
outputting mapping at the third stage, namely processing the feature vector by adopting a HervieSeld step function, binarizing the feature vector, dividing the binarized feature vector into a plurality of blocks, performing histogram statistical calculation on each block, and integrating all histograms into one vector to obtain final power utilization feature output;
and S3, carrying out correlation mapping on the final electricity utilization characteristics through a wavelet neural network WNN model, and detecting the abnormal electricity utilization behavior of the user.
2. The PCANet and WNN based user abnormal electricity usage detection method of claim 1, wherein the data is preprocessed using interpolation and normalization:
and (3) processing missing values in the data by adopting an interpolation method:
Figure FDA0002998936770000011
in the formula, xiRepresenting the electricity consumption of a user within a certain time range, NaN being xiIs a null value or is a non-numeric value;
the following formula is utilized to carry out normalization operation, and the detection precision of the network is improved:
Figure FDA0002998936770000021
x is the entire sample data set, min (X) is the minimum value in the data set, and max (X) is the maximum value in the data set.
3. The method for detecting abnormal electricity consumption of a user based on PCANet and WNN as claimed in claim 1, wherein the first stage PCANet mapping is specifically:
for each sample SiBy a size w1×w2Obtaining a sample fragment by sliding a window of:
Figure FDA0002998936770000022
wherein xi，jRepresenting S in jth vectorization windowiThe ith sample vector is obtained by performing mean value removal processing on the sample vectors in all windows:
Figure FDA0002998936770000023
the reconstruction error is then minimized with the following equation:
Figure FDA0002998936770000024
Figure FDA0002998936770000025
is L1×L1The solution of the above formula is UUTL of1And (4) a main feature vector.
4. The method for PCANet and WNN based abnormal power usage detection of users as claimed in claim 3, wherein in the third stage output mapping, PCANet processes the feature vector by using HeveSaede step function, and binarizes it:
Figure FDA0002998936770000026
h (-) represents a Hervesaide step function, wherein each feature vector is constrained to
Figure FDA0002998936770000027
Within the range of (1);
Figure FDA0002998936770000028
a filter that is a first stage;
Figure FDA0002998936770000029
the output of the ith filter for the ith image; for each data sample SiOutput the corresponding output
Figure FDA00029989367700000210
Dividing the block units into B block units, performing histogram statistical calculation on each block unit, integrating all histograms into a vector, and obtaining a final feature output result:
Figure FDA0002998936770000031
here, Bhist (. cndot.) represents the statistical calculation of the histogram for each block unit, fiRepresents the sample S by PCANetiAnd extracting the final feature vector.
5. The method for detecting abnormal electricity consumption of users based on PCANet and WNN as claimed in claim 3, wherein step S3 is specifically:
selecting a wavelet neural network as the correlation mapping from the series of characteristics to the user electricity consumption abnormal behavior detection, wherein a network mapping equation can be expressed as follows:
Figure FDA0002998936770000032
in the above formula, M and N represent the mapping number of input and output respectively, and represent M input features and N output features; sigmak(t) represents the kth input feature, θ (-) is the weight function from the input layer to the hidden layer, and q is the number of hidden layers; omegaijRepresenting the connection rights of the hidden layer to the output layer, ajRepresenting the scale of the wavelet function, bjRepresenting the shift of the wavelet function and,
Figure FDA0002998936770000033
representing a Sigmoid activation function.
6. The method for detecting abnormal electricity consumption of users based on PCANet and WNN of claim 1, further comprising step S4, generating early warning information according to the detected abnormal electricity consumption behavior of the users, and reminding the abnormal users of electricity consumption.
7. A system for detecting abnormal electricity consumption of a user based on PCANet and WNN is characterized by comprising:
the data preprocessing module is used for preprocessing the acquired power consumption data of the user within a certain time range;
the feature extraction module adopts a principal component analysis network to perform dimensionality reduction processing on the electrical signal through three times of PCA mapping so as to extract electrical signal features, and the feature extraction module is specifically divided into three stages of PCA mapping: the first stage of PCANet mapping, namely acquiring sample fragments of preprocessed data by adopting a sliding window, and performing mean value removal processing on sample vectors in all windows and minimizing reconstruction errors; and PCA network structure processing in the second stage: taking the output of the first stage as input, and repeating the PCANet mapping process of the first stage; outputting mapping at the third stage, namely processing the feature vector by adopting a HervieSeld step function, binarizing the feature vector, dividing the binarized feature vector into a plurality of blocks, performing histogram statistical calculation on each block, and integrating all histograms into one vector to obtain final power utilization feature output;
and the user electricity utilization abnormal behavior detection module is used for performing correlation mapping on the final electricity utilization characteristics through a wavelet neural network WNN model to detect the abnormal electricity utilization behavior of the user.
8. The PCANet and WNN based user abnormal electricity usage detection system of claim 7, further comprising:
and the early warning module is used for generating early warning information according to the detected abnormal electricity utilization behavior result of the user and reminding the abnormal electricity utilization of the user.
9. The PCANet and WNN based user abnormal electricity usage detection system of claim 7, wherein the data preprocessing module preprocesses the electricity usage data specifically using interpolation and normalization.
10. A computer storage medium executable by a processor and having stored therein a computer program for performing the method of PCANet and WNN based user abnormal electricity usage detection of any one of claims 1 to 5.
CN202110339366.1A 2021-03-30 2021-03-30 User abnormal electricity utilization detection method based on PCANet and WNN Pending CN112926687A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110339366.1A CN112926687A (en) 2021-03-30 2021-03-30 User abnormal electricity utilization detection method based on PCANet and WNN

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110339366.1A CN112926687A (en) 2021-03-30 2021-03-30 User abnormal electricity utilization detection method based on PCANet and WNN

Publications (1)

Publication Number Publication Date
CN112926687A true CN112926687A (en) 2021-06-08

Family

ID=76176577

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110339366.1A Pending CN112926687A (en) 2021-03-30 2021-03-30 User abnormal electricity utilization detection method based on PCANet and WNN

Country Status (1)

Country Link
CN (1) CN112926687A (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109086926A (en) * 2018-07-25 2018-12-25 南京理工大学 A kind of track traffic for passenger flow prediction technique in short-term based on combination neural net structure
CN109377409A (en) * 2018-09-29 2019-02-22 重庆大学 A kind of user power utilization anomaly detection method based on BP neural network
CN109800856A (en) * 2018-12-14 2019-05-24 广东电网有限责任公司 A kind of method of chaotic particle swarm optimization wavelet neural network
CN111507422A (en) * 2020-04-24 2020-08-07 西安工程大学 CQFPA-WNN-based transformer fault diagnosis method
US20200348438A1 (en) * 2019-05-03 2020-11-05 Guoqiang Xue Method for Suppressing Airborne Transient Electromagnetic In-Band Vibration Noise

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109086926A (en) * 2018-07-25 2018-12-25 南京理工大学 A kind of track traffic for passenger flow prediction technique in short-term based on combination neural net structure
CN109377409A (en) * 2018-09-29 2019-02-22 重庆大学 A kind of user power utilization anomaly detection method based on BP neural network
CN109800856A (en) * 2018-12-14 2019-05-24 广东电网有限责任公司 A kind of method of chaotic particle swarm optimization wavelet neural network
US20200348438A1 (en) * 2019-05-03 2020-11-05 Guoqiang Xue Method for Suppressing Airborne Transient Electromagnetic In-Band Vibration Noise
CN111507422A (en) * 2020-04-24 2020-08-07 西安工程大学 CQFPA-WNN-based transformer fault diagnosis method

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
TSUNG-HAN CHAN等: "PCANet: A Simple Deep Learning Baseline for Image Classification?", 《ARXIV:1404.3606V2》 *
王致杰等: "《大型风力发电机组状态监测与智能故障诊断》", 30 September 2013 *
石盼等: "小波神经网络在反窃电***中的应用研究", 《计算技术与自动化》 *

Similar Documents

Publication Publication Date Title
Xin et al. Complex network classification with convolutional neural network
Zheng et al. How framelets enhance graph neural networks
CN114564991B (en) Electroencephalogram signal classification method based on transducer guided convolutional neural network
CN109975673B (en) Method for identifying fault arc at direct current side of photovoltaic microgrid
CN108446589B (en) Face recognition method based on low-rank decomposition and auxiliary dictionary in complex environment
US8010296B2 (en) Apparatus and method for removing non-discriminatory indices of an indexed dataset
CN110796022B (en) Low-resolution face recognition method based on multi-manifold coupling mapping
CN111695455B (en) Low-resolution face recognition method based on coupling discrimination manifold alignment
Cord et al. Feature selection in robust clustering based on Laplace mixture
CN103164689A (en) Face recognition method and face recognition system
CN114067915A (en) scRNA-seq data dimension reduction method based on deep antithetical variational self-encoder
Singh et al. SVD-based redundancy removal in 1-D CNNs for acoustic scene classification
CN108108666B (en) Hybrid matrix estimation method based on wavelet analysis and time-frequency single-source detection
Zhang et al. Steganographer detection via a similarity accumulation graph convolutional network
CN112926687A (en) User abnormal electricity utilization detection method based on PCANet and WNN
CN115731462A (en) Hyperspectral image change detection method based on multi-head self-crossing mixed attention
Yang et al. Tensor completion for recovering multichannel audio signal with missing data
CN114373212A (en) Face recognition model construction method, face recognition method and related equipment
Wu et al. Learning domain-invariant representation for generalizing face forgery detection
CN112650877A (en) High-definition remote sensing image fast retrieval method based on improved depth residual convolution neural network and depth hash
Niveditha et al. Kernelized Deep Networks for Speech Signal Segmentation Using Clustering and Artificial Intelligence in Neural Networks
Amalia et al. The Application of Modified K-Nearest Neighbor Algorithm for Classification of Groundwater Quality Based on Image Processing and pH, TDS, and Temperature Sensors
Yu et al. Detach and enhance: Learning disentangled cross-modal latent representation for efficient face-voice association and matching
Sun et al. Microarray data classification using the spectral-feature-based TLS ensemble algorithm
Wang et al. Data mining with independent component analysis

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