CN113920375B - Fusion characteristic typical load identification method and system based on combination of Faster R-CNN and SVM - Google Patents
Fusion characteristic typical load identification method and system based on combination of Faster R-CNN and SVM Download PDFInfo
- Publication number
- CN113920375B CN113920375B CN202111284397.8A CN202111284397A CN113920375B CN 113920375 B CN113920375 B CN 113920375B CN 202111284397 A CN202111284397 A CN 202111284397A CN 113920375 B CN113920375 B CN 113920375B
- Authority
- CN
- China
- Prior art keywords
- characteristic
- load
- wavelet
- signal
- network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 29
- 230000004927 fusion Effects 0.000 title claims abstract description 18
- 238000000605 extraction Methods 0.000 claims abstract description 20
- 238000012545 processing Methods 0.000 claims abstract description 14
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 12
- 238000012706 support-vector machine Methods 0.000 claims description 35
- 230000007246 mechanism Effects 0.000 claims description 15
- 230000003203 everyday effect Effects 0.000 claims description 9
- 238000001514 detection method Methods 0.000 claims description 8
- 239000013598 vector Substances 0.000 claims description 8
- 238000010586 diagram Methods 0.000 claims description 7
- 238000010606 normalization Methods 0.000 claims description 5
- 230000009021 linear effect Effects 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 2
- 238000011176 pooling Methods 0.000 claims description 2
- 238000004134 energy conservation Methods 0.000 abstract description 6
- 230000005611 electricity Effects 0.000 abstract description 5
- 238000013527 convolutional neural network Methods 0.000 description 31
- 230000006870 function Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 3
- 238000009434 installation Methods 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
- Complex Calculations (AREA)
Abstract
The invention relates to a fusion characteristic typical load identification method and system based on the combination of a fast R-CNN and an SVM, wherein the method comprises the following steps: s1: the method comprises the steps of collecting loads of electrical equipment, constructing signal data samples, selecting corresponding wavelet bases, and determining the number of decomposition layers; decomposing the noise-containing signal to obtain a group of wavelet coefficients; s2: performing soft threshold function processing on the wavelet coefficients to obtain estimated wavelet coefficients; s3: reconstructing the wavelet by using the estimated wavelet coefficient to obtain a denoised reconstructed signal; s4: normalizing the reconstructed signal to obtain a normalized reconstructed signal; s5: based on the normalized reconstruction signals, constructing a load characteristic curve image, performing characteristic extraction on the load characteristic curve image by using a fast R-CNN network, and classifying by using SVM to obtain a final load identification result. The method provided by the invention improves the accuracy of typical load characteristic identification, guides energy conservation, reduces the electricity cost of users and improves the utilization rate of electric energy.
Description
Technical Field
The invention relates to the field of intelligent electric meters, in particular to a fusion characteristic typical load identification method and system based on the combination of a Faster R-CNN and an SVM.
Background
The fee-controlled intelligent ammeter is used as an intelligent metering instrument and is directly arranged on a total current inlet bus of a user to collect power information parameters of a plurality of electric equipment in the user's home. The system can accurately measure and record electric energy in multiple ways, supports real-time electricity price multiple rates, and is widely applied. Load identification is an important technology of a cost control intelligent ammeter, and the total power can be thinned into the load. The power utilization list of the main electric appliance is provided for the user, and guidance is provided for energy conservation, so that the power utilization cost of the user can be reduced, and the utilization rate of electric energy can be improved.
Load identification is largely divided into two types, invasive and non-invasive. The non-invasive load identification is to identify the electrical characteristics of the load such as current, power and the like when the current is out. Non-invasive load identification refers to the installation of identification means at each electrical interface. Since the operation such as installation and maintenance of the invasive load recognition apparatus is difficult to achieve, the noninvasive load recognition is widely accepted. The non-invasive load identification technology of the intelligent ammeter is an important electric energy classification metering technology in the intelligent power grid and is also one of important measures for realizing energy conservation and emission reduction. At present, the defect that the precision is lost and similar load characteristic electrical appliances are easy to be confused still exists when the convolutional neural network is used for identifying the load by utilizing the identification capability of the picture, so that how to improve the identification accuracy is needed to be solved.
Disclosure of Invention
In order to solve the technical problems, the invention provides a fusion characteristic typical load identification method and system based on the combination of a Faster R-CNN and an SVM.
The technical scheme of the invention is as follows: a fusion characteristic typical load identification method based on the combination of Faster R-CNN and SVM comprises the following steps:
step S1: selecting n days of a day to be detected, collecting M point loads of electrical equipment every day, constructing a signal data sample, obtaining a load characteristic set, selecting a corresponding wavelet base according to the load characteristic, and determining the number of decomposition layers; decomposing the noise-containing signal to obtain a group of wavelet coefficients;
Step S2: determining a threshold value of each layer of wavelet coefficient in stages, and performing soft threshold function processing on the wavelet coefficients to obtain estimated wavelet coefficients;
Step S3: reconstructing the wavelet by using the estimated wavelet coefficient to obtain a denoised reconstructed signal;
Step S4: normalizing the reconstructed signal to obtain a normalized reconstructed signal;
step S5: constructing a load characteristic curve image based on the normalized reconstruction signal, carrying out characteristic extraction on the load characteristic curve image by using a fast R-CNN network to obtain a characteristic image, calculating to obtain a region of interest, generating a region of interest characteristic image, generating a characteristic sequence based on the region of interest characteristic image, obtaining a weighted characteristic sequence by using an attention mechanism, and classifying the weighted characteristic sequence by using an SVM to obtain a final load identification result; wherein the fast R-CNN network comprises: a feature extraction network, a regional suggestion network, a regional feature map generation network, and a classification and regression detection network.
Compared with the prior art, the invention has the following advantages:
based on Fast R-CNN, support vector machine SVM and attention mechanism, the invention provides a fusion characteristic typical load recognition method based on Fast R-CNN and SVM, the Fast R-CNN has super strong learning ability and can realize characteristic extraction by using a small amount of sample data, and the network is more suitable for capturing local characteristics; the introduced attention mechanism can pay more clearly attention to the relation between elements in the global sequence, and the two methods are combined to learn and identify the local features and the global relation in the feature sequence more efficiently. In addition, the support vector machine is used for classifying, and the generalization capability of the SVM on the small sample data set and the feature extraction capability of the Faster R-CNN are combined, so that the applicability of the whole model on the small sample data set is improved, the accuracy of typical load feature identification is improved, the energy conservation is guided, the electricity consumption cost of a user is reduced, and the utilization rate of electric energy is improved.
Drawings
FIG. 1 is a flowchart of a fusion feature typical load identification method based on the combination of Faster R-CNN and SVM in an embodiment of the invention;
FIG. 2 is a schematic diagram of the attention mechanism in a Faster R-CNN network in accordance with an embodiment of the present invention;
FIG. 3 is a schematic diagram of a Faster R-CNN network in accordance with an embodiment of the present invention;
FIG. 4 is a block diagram of a fusion feature representative load recognition system based on the combination of Faster R-CNN and SVM in an embodiment of the invention.
Detailed Description
The invention provides a fusion characteristic typical load identification method based on the combination of a fast R-CNN and an SVM, which improves the accuracy of typical load characteristic identification, guides energy conservation, reduces the electricity cost of a user and improves the utilization rate of electric energy.
The present invention will be further described in detail below with reference to the accompanying drawings by way of specific embodiments in order to make the objects, technical solutions and advantages of the present invention more apparent.
Example 1
As shown in FIG. 1, the fusion characteristic typical load identification method based on the combination of the Faster R-CNN and the SVM provided by the embodiment of the invention comprises the following steps:
Step S1: selecting n days of a day to be detected, collecting M point loads of electrical equipment every day, constructing a signal data sample, obtaining a load characteristic set, selecting a corresponding wavelet base according to the load characteristic, and determining the number of decomposition layers; decomposing the noise-containing signal to obtain a group of wavelet coefficients;
Step S2: for each layer of wavelet coefficient in stages, determining a threshold value of the wavelet coefficient, and performing soft threshold function processing on the wavelet coefficient to obtain an estimated wavelet coefficient;
step S3: reconstructing the wavelet by using the estimated wavelet coefficient to obtain a denoised reconstructed signal;
Step S4: normalizing the reconstructed signal to obtain a normalized reconstructed signal;
Step S5: constructing a load characteristic curve image based on a normalized reconstruction signal, performing characteristic extraction on the load characteristic curve image by using a fast R-CNN network to obtain a characteristic image, calculating to obtain a region of interest, generating the region of interest characteristic image, generating a characteristic sequence based on the region of interest characteristic image, obtaining a weighted characteristic sequence by using an attention mechanism, and classifying the weighted characteristic sequence by using an SVM to obtain a final load identification result; wherein the fast R-CNN network comprises: a feature extraction network, a regional suggestion network, a regional feature map generation network, and a classification and regression detection network.
In one embodiment, step S1 described above: selecting n days of a day to be detected, constructing a signal data sample by M points of load every day to obtain a load characteristic set, selecting a corresponding wavelet base according to the load characteristic, and determining the decomposition layer number j; decomposing the noise-containing signal to obtain a group of wavelet coefficients, specifically comprising:
Step S11: selecting N days of a day to be detected, collecting M point loads of electrical equipment every day, and constructing a signal data sample to obtain a load characteristic set of X= { X 1,x2,...,xi,...,xN }, wherein N is the number of load characteristics of the electrical equipment, and X i is the ith characteristic in the load characteristic set X;
step S12: the noisy signal is calculated according to the following formula (1):
g(k)=f(k)+s(k)k=1,2,3,…,M-1 (1)
Wherein k is the moment, f (k) is the real signal, s (k) is the noise signal, and g (k) is the noise-containing signal;
Step S13: selecting a corresponding wavelet base based on the noise-containing signal g (k), and selecting a smooth continuous wavelet base when the noise-containing signal is smooth; when the linear property of the noise-containing signal is strong, selecting a linear wavelet base;
Step S14: and decomposing the noise-containing signal g (k) according to a preset decomposition layer number j to obtain a group of wavelet coefficients omega j,k.
In one embodiment, step S2 above: for each layer of wavelet coefficient in stages, determining a threshold value of the wavelet coefficient, and performing soft threshold function processing on the wavelet coefficient to obtain an estimated wavelet coefficient, wherein the method specifically comprises the following steps:
performing soft threshold function processing on wavelet coefficients according to a formula (2) to obtain estimated wavelet coefficients
Wherein T is a threshold; when the absolute value of the wavelet coefficient omega j,k is larger than T, the wavelet coefficient is estimatedThe absolute value of (a) is |omega j,k | -T, and the sign is kept unchanged; when the absolute value of the wavelet coefficient ω j,k is less than or equal to T, the wavelet coefficient/>, is estimatedZero.
In one embodiment, the step S3: using estimated wavelet coefficientsReconstructing the wavelet to obtain a denoised reconstructed signal/>
In one embodiment, step S4 above: reconstructed signalAnd (3) carrying out normalization processing, and normalizing the value to a (0, 1) interval to obtain a normalized reconstruction signal.
And (3) preprocessing the data samples through the steps S1 to S4, and carrying out noise processing and normalization processing on the images through wavelet denoising, so as to weaken the influence of bad data on the subsequent neural network training.
In one embodiment, the step S5 is as follows: constructing a load characteristic curve image based on a normalized reconstruction signal, performing characteristic extraction on the load characteristic curve image by using a fast R-CNN network to obtain a characteristic image, calculating to obtain a region of interest, generating the region of interest characteristic image, generating a characteristic sequence based on the region of interest characteristic image, obtaining a weighted characteristic sequence by using an attention mechanism, and classifying the weighted characteristic sequence by using an SVM to obtain a final load identification result; wherein the fast R-CNN network comprises: the method specifically comprises a feature extraction network, a regional suggestion network, a regional feature map generation network and a classification and regression detection network, and comprises the following steps:
Step S51: constructing a load characteristic curve image based on the normalized reconstructed signal: inputting the load characteristic curve image into a characteristic extraction network to obtain a characteristic diagram;
According to the embodiment of the invention, the VGG network is adopted to realize the feature extraction of the load feature curve image;
Step S52: inputting the feature map into an RPN (regional recommendation network), and calculating to obtain a target region to be identified on the feature map to obtain a region of interest;
step S53: inputting the region of interest into a region feature map generation network, generating a region of interest feature map by using ROI pooling, and obtaining a feature sequence through convolution layer calculation Calculating feature sequences/>, using an attention mechanismThe specific steps are as follows:
According to the following formula (3), a query vector q= [ q 1,q2,...,qi,...,qN ], a key vector m= [ m 1,m2,...,mi,...,mN ] and a value vector v= [ v 1,v2,...,vi,...,vN ] are calculated:
Wherein, W q、Wm and W v are preset weight matrix coefficients;
each is calculated according to formula (4) Is c t,i, i.e. current/>, of the current/>Correlation between the power value at time t and the power values of other vectors in the feature sequence;
ct,i=qt·mi (4)
Normalizing the attention score according to equation (5):
The normalized attention scores are weighted and summed according to equation (6):
wherein y t is the value of the weighted sequence y= [ y 1,y2,...,yt,...,yN ] of the output after the attention mechanism calculation at time t.
As shown in fig. 2, a schematic diagram of the attention mechanism in a fast R-CNN network is shown.
In the training process of the fast R-CNN network, the embodiment of the invention distributes a weight coefficient for each element in the feature sequence through introducing an attention mechanism, can focus the network on details such as peak Gu Changduan with more discrimination information, and weakens the influence of other fluctuation on classification results.
Step S54: the weighted sequence is input into a classification and regression detection network, and is classified by using a one-to-one Support Vector Machine (SVM), and the method specifically comprises the following steps:
Classifying the three classes according to a weighted sequence y= [ y 1,y2,...,yt,...,yN ] by using an SVM to obtain z classes, judging again by using z (z-1)/2 bi-classifiers, voting for the corresponding classes, and taking the class with the most votes as a final load recognition result.
According to the embodiment of the invention, the generalization capability of a small sample can be improved by selecting the SVM as the feature classifier.
As shown in FIG. 3, a schematic diagram of the structure of a Faster R-CNN network is shown.
Based on Fast R-CNN, support vector machine SVM and attention mechanism, the invention provides a fusion characteristic typical load recognition method based on Fast R-CNN and SVM, the Fast R-CNN has super strong learning ability and can realize characteristic extraction by using a small amount of sample data, and the network is more suitable for capturing local characteristics; the introduced attention mechanism can pay more clearly attention to the relation between elements in the global sequence, and the two methods are combined to learn and identify the local features and the global relation in the feature sequence more efficiently. In addition, the support vector machine is used for classifying, and the generalization capability of the SVM on the small sample data set and the feature extraction capability of the Faster R-CNN are combined, so that the applicability of the whole model on the small sample data set is improved, the accuracy of typical load feature identification is improved, the energy conservation is guided, the electricity consumption cost of a user is reduced, and the utilization rate of electric energy is improved.
Example two
As shown in FIG. 4, the embodiment of the invention provides a fusion characteristic typical load identification system based on the combination of a Faster R-CNN and an SVM, which comprises the following modules:
The wavelet coefficient acquisition module 61 is used for selecting n days of a day to be detected, collecting M points of load of electrical equipment every day, constructing a signal data sample, obtaining a load characteristic set, selecting a corresponding wavelet base according to the load characteristic, and determining the number of decomposition layers; decomposing the noise-containing signal to obtain a group of wavelet coefficients;
the module 62 for obtaining estimated wavelet coefficients is configured to determine a threshold value of each layer of wavelet coefficients in stages, and perform soft threshold function processing on the wavelet coefficients to obtain estimated wavelet coefficients;
a reconstruction signal module 63, configured to reconstruct the wavelet by using the estimated wavelet coefficient to obtain a denoised reconstruction signal;
The normalization module 64 is configured to normalize the reconstructed signal to obtain a normalized reconstructed signal;
The load recognition module 65 is configured to construct a load characteristic curve image based on the normalized reconstruction signal, perform characteristic extraction on the load characteristic curve image by using a fast R-CNN network to obtain a characteristic map, calculate to obtain a region of interest, generate a region of interest characteristic map, generate a characteristic sequence based on the region of interest characteristic map, obtain a weighted characteristic sequence by using an attention mechanism, and classify the weighted characteristic sequence to obtain a final load recognition result; wherein the fast R-CNN network comprises: a feature extraction network, a regional suggestion network, a regional feature map generation network, and a classification and regression detection network.
The above examples are provided for the purpose of describing the present invention only and are not intended to limit the scope of the present invention. The scope of the invention is defined by the appended claims. Various equivalents and modifications that do not depart from the spirit and principles of the invention are intended to be included within the scope of the invention.
Claims (4)
1. A fusion characteristic typical load identification method based on the combination of Faster R-CNN and SVM is characterized by comprising the following steps:
step S1: selecting n days of a day to be detected, collecting M point loads of electrical equipment every day, constructing a signal data sample, obtaining a load characteristic set, selecting a corresponding wavelet base according to the load characteristic, and determining the number of decomposition layers; decomposing the noise-containing signal to obtain a group of wavelet coefficients;
Step S2: determining a threshold value of each layer of wavelet coefficient in stages, and performing soft threshold function processing on the wavelet coefficients to obtain estimated wavelet coefficients;
Step S3: reconstructing the wavelet by using the estimated wavelet coefficient to obtain a denoised reconstructed signal;
Step S4: normalizing the reconstructed signal to obtain a normalized reconstructed signal;
Step S5: constructing a load characteristic curve image based on the normalized reconstruction signal, carrying out characteristic extraction on the load characteristic curve image by using a fast R-CNN network to obtain a characteristic image, calculating to obtain a region of interest, generating a region of interest characteristic image, generating a characteristic sequence based on the region of interest characteristic image, obtaining a weighted characteristic sequence by using an attention mechanism, and classifying the weighted characteristic sequence by using an SVM to obtain a final load identification result; wherein the fast R-CNN network comprises: the method specifically comprises a feature extraction network, a regional suggestion network, a regional feature map generation network and a classification and regression detection network, and comprises the following steps:
Step S51: constructing a load characteristic curve image based on the normalized reconstructed signal: inputting the load characteristic curve image into the characteristic extraction network to obtain a characteristic diagram;
Step S52: inputting the feature map into an area suggestion network (RPN) network, and calculating a target area to be identified on the feature map to obtain a region of interest (ROI);
step S53: inputting the region of interest into a region feature map generation network, generating a region of interest feature map by using ROI pooling, and calculating by using a convolution layer to obtain a feature sequence ; Calculating the feature sequence/>, using an attention mechanismWeights of (2) to obtain weighted feature sequences/>The method comprises the following specific steps:
According to the following formula (3), a query vector is calculated Key vectorSum vector/>:
(3)
Wherein,、/>And/>The method comprises the steps of setting a preset weight matrix coefficient;
each is calculated according to formula (4) Attention score/>I.e. current/>Correlation between the power value at time t and the power values of other vectors in the feature sequence;
(4)
normalizing the attention score according to equation (5):
(5)
The normalized attention scores are weighted and summed according to equation (6):
,
Wherein, Weighted sequence/>, for output after attention mechanism calculationA value at time t;
Step S54: inputting the weighted sequence into a classification and regression detection network, and classifying the sequence by using a one-to-one Support Vector Machine (SVM), wherein the method specifically comprises the following steps:
According to the weighted sequence Classifying the load with SVM to obtain z categories, discriminating the load with z (z-1)/2 classifiers, voting for the corresponding categories, and taking the category with the most votes as the final load recognition result.
2. The fusion feature typical load recognition method based on the combination of fast R-CNN and SVM according to claim 1, wherein the step S1: selecting n days of a day to be detected, constructing a signal data sample by M points of load every day to obtain a load characteristic set, selecting a corresponding wavelet base according to the load characteristic, and determining the number j of decomposition layers; decomposing the noise-containing signal to obtain a group of wavelet coefficients, specifically comprising:
step S11: selecting n days of a day to be detected, collecting M point loads of electrical equipment every day, constructing a signal data sample, and obtaining a load characteristic set as Wherein N is the load characteristic number of the electrical equipment,/>Is the ith feature in the load feature set X;
step S12: the noisy signal is calculated according to the following formula (1):
(1)
wherein k is the time of day, As a true signal,/>Is a noise signal,/>Is a noise-containing signal;
Step S13: based on the noisy signal Selecting a corresponding wavelet basis, and selecting a smooth continuous wavelet basis when the noise-containing signal is smooth; when the linear property of the noise-containing signal is strong, selecting a linear wavelet basis;
step S14: according to the preset decomposition layer number j, the noise-containing signal Decomposing to obtain a group of wavelet coefficients。
3. The fusion feature typical load recognition method based on the combination of fast R-CNN and SVM according to claim 1, wherein the step S2: for each wavelet coefficient of each layer in stages, determining a threshold value of the wavelet coefficient, and performing soft threshold function processing on the wavelet coefficient to obtain an estimated wavelet coefficient, wherein the method specifically comprises the following steps:
performing soft threshold function processing on the wavelet coefficients according to a formula (2) to obtain estimated wavelet coefficients :
(2)
Wherein T is a threshold; when the wavelet coefficientWhen the absolute value of (a) is greater than T, the estimated wavelet coefficient/>The absolute value of (2) is/>The sign remains unchanged; when the wavelet coefficient/>When the absolute value of (2) is less than or equal to T, the estimated wavelet coefficient/>Zero.
4. A fusion feature typical load recognition system based on the combination of a fast R-CNN and an SVM for implementing the fusion feature typical load recognition method based on the combination of a fast R-CNN and an SVM according to claim 1, comprising the following modules:
The system comprises a wavelet coefficient acquisition module, a decomposition layer number determination module and a decomposition layer number determination module, wherein the wavelet coefficient acquisition module is used for selecting n days of a day to be detected, acquiring M point loads of electrical equipment every day, constructing a signal data sample, obtaining a load characteristic set, selecting a corresponding wavelet base according to the load characteristic, and determining the decomposition layer number; decomposing the noise-containing signal to obtain a group of wavelet coefficients;
The module is used for obtaining estimated wavelet coefficients, determining a threshold value of each layer of wavelet coefficients in a staged manner, and carrying out soft threshold function processing on the wavelet coefficients to obtain estimated wavelet coefficients;
The reconstruction signal module is used for reconstructing the wavelet by utilizing the estimated wavelet coefficient to obtain a denoised reconstruction signal;
the normalization module is used for carrying out normalization processing on the reconstruction signals to obtain normalized reconstruction signals;
the load identification module is used for constructing a load characteristic curve image based on the normalized reconstruction signal, carrying out characteristic extraction on the load characteristic curve image by using a fast R-CNN network to obtain a characteristic image, calculating to obtain a region of interest, generating a region of interest characteristic image, generating a characteristic sequence based on the region of interest characteristic image, utilizing an attention mechanism to obtain a weighted characteristic sequence, and classifying the weighted characteristic sequence to obtain a final load identification result; wherein the fast R-CNN network comprises: a feature extraction network, a regional suggestion network, a regional feature map generation network, and a classification and regression detection network.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111284397.8A CN113920375B (en) | 2021-11-01 | 2021-11-01 | Fusion characteristic typical load identification method and system based on combination of Faster R-CNN and SVM |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111284397.8A CN113920375B (en) | 2021-11-01 | 2021-11-01 | Fusion characteristic typical load identification method and system based on combination of Faster R-CNN and SVM |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113920375A CN113920375A (en) | 2022-01-11 |
CN113920375B true CN113920375B (en) | 2024-06-25 |
Family
ID=79244809
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111284397.8A Active CN113920375B (en) | 2021-11-01 | 2021-11-01 | Fusion characteristic typical load identification method and system based on combination of Faster R-CNN and SVM |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113920375B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114332443B (en) * | 2022-03-16 | 2022-06-24 | 北京智芯微电子科技有限公司 | Non-intrusive load recognition device, method, storage medium, and electronic apparatus |
CN116722557B (en) * | 2023-05-30 | 2024-06-21 | 国网北京市电力公司 | Demand response rebound time length analysis method and system based on wavelet decomposition |
CN117496256B (en) * | 2023-11-14 | 2024-06-18 | 中国人民解放军军事科学院系统工程研究院 | Small sample electromagnetic signal identification method and device |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106908671A (en) * | 2017-01-13 | 2017-06-30 | 中国海洋大学 | A kind of non-intrusion type household loads intelligent detecting method and system |
CN110070229A (en) * | 2019-04-26 | 2019-07-30 | 中国计量大学 | The short term prediction method of home electrical load |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11183304B2 (en) * | 2019-01-08 | 2021-11-23 | International Business Machines Corporation | Personalized smart home recommendations through cognitive load analysis |
CN110188826A (en) * | 2019-05-31 | 2019-08-30 | 湖南科技大学 | Household electrical appliance operating status non-invasive inspection methods based on intelligent electric meter data |
CN110514889A (en) * | 2019-07-19 | 2019-11-29 | 浙江万胜智能科技股份有限公司 | A kind of method and system of non-intrusion type household electricity remained capacity |
CN111242423B (en) * | 2019-12-31 | 2022-07-12 | 深圳能源售电有限公司 | Active demand response potential analysis method, device, equipment and storage medium |
CN111553212B (en) * | 2020-04-16 | 2022-02-22 | 中国科学院深圳先进技术研究院 | Remote sensing image target detection method based on smooth frame regression function |
-
2021
- 2021-11-01 CN CN202111284397.8A patent/CN113920375B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106908671A (en) * | 2017-01-13 | 2017-06-30 | 中国海洋大学 | A kind of non-intrusion type household loads intelligent detecting method and system |
CN110070229A (en) * | 2019-04-26 | 2019-07-30 | 中国计量大学 | The short term prediction method of home electrical load |
Also Published As
Publication number | Publication date |
---|---|
CN113920375A (en) | 2022-01-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113920375B (en) | Fusion characteristic typical load identification method and system based on combination of Faster R-CNN and SVM | |
CN110823574B (en) | Fault diagnosis method based on semi-supervised learning deep countermeasure network | |
Qiu et al. | An automatic identification framework for complex power quality disturbances based on multifusion convolutional neural network | |
CN109522857B (en) | People number estimation method based on generation type confrontation network model | |
Othman et al. | Probabilistic neural network for brain tumor classification | |
CN113221641B (en) | Video pedestrian re-identification method based on generation of antagonism network and attention mechanism | |
CN113191215A (en) | Rolling bearing fault diagnosis method integrating attention mechanism and twin network structure | |
CN105956570B (en) | Smiling face's recognition methods based on lip feature and deep learning | |
CN112446242A (en) | Acoustic scene classification method and device and corresponding equipment | |
CN110161388A (en) | A kind of the fault type recognition method and its system of high-tension apparatus | |
CN116520140A (en) | Permanent magnet direct drive motor fault detection method integrating feature extraction and small sample classification | |
CN114565594A (en) | Image anomaly detection method based on soft mask contrast loss | |
CN115410258A (en) | Human face expression recognition method based on attention image | |
CN111289251A (en) | Rolling bearing fine-grained fault identification method | |
Chang et al. | Intelligent fault diagnosis scheme via multi-module supervised-learning network with essential features capture-regulation strategy | |
CN113987910A (en) | Method and device for identifying load of residents by coupling neural network and dynamic time planning | |
CN113758709A (en) | Rolling bearing fault diagnosis method and system combining edge calculation and deep learning | |
Xin et al. | Random part localization model for fine grained image classification | |
CN117611830A (en) | Random class target positioning and counting method based on few sample labeling | |
CN114360058B (en) | Cross-view gait recognition method based on walking view prediction | |
CN113963427B (en) | Method and system for rapid in-vivo detection | |
CN113111774B (en) | Radar signal modulation mode identification method based on active incremental fine adjustment | |
CN112241922B (en) | Power grid asset comprehensive value assessment method based on improved naive Bayesian classification | |
CN114021424A (en) | PCA-CNN-LVQ-based voltage sag source identification method | |
Ma et al. | Complex PQD Classification Using Time–Frequency Analysis and Multiscale Parallel Attention Residual Network |
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 |