CN116933000A - Non-invasive load monitoring method based on low-frequency load multi-feature fusion - Google Patents
Non-invasive load monitoring method based on low-frequency load multi-feature fusion Download PDFInfo
- Publication number
- CN116933000A CN116933000A CN202310757598.8A CN202310757598A CN116933000A CN 116933000 A CN116933000 A CN 116933000A CN 202310757598 A CN202310757598 A CN 202310757598A CN 116933000 A CN116933000 A CN 116933000A
- Authority
- CN
- China
- Prior art keywords
- data
- initial power
- feature
- modal
- power data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 35
- 230000004927 fusion Effects 0.000 title claims abstract description 23
- 238000012544 monitoring process Methods 0.000 title claims abstract description 22
- 238000013136 deep learning model Methods 0.000 claims abstract description 16
- 238000004140 cleaning Methods 0.000 claims abstract description 4
- 230000002708 enhancing effect Effects 0.000 claims abstract description 4
- 238000000354 decomposition reaction Methods 0.000 claims description 15
- 238000010606 normalization Methods 0.000 claims description 6
- 238000005516 engineering process Methods 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000012804 iterative process Methods 0.000 claims description 3
- 230000011218 segmentation Effects 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 2
- 238000013135 deep learning Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 230000003631 expected effect Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/003—Load forecast, e.g. methods or systems for forecasting future load demand
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2123/00—Data types
- G06F2123/02—Data types in the time domain, e.g. time-series data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Power Engineering (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a non-invasive load monitoring method based on low-frequency load multi-feature fusion, which comprises the following steps: acquiring total power data of a low-frequency load, and performing data cleaning and sliding window operation on the total power data to obtain initial power data; decomposing the initial power data by using a feedback VMD to obtain a plurality of modal components; selecting modal components by using kurtosis criteria and reconstructing signals; enhancing the pulse characteristics of the reconstructed signal by using a Teager energy operator; feature fusion is carried out on the features after feature enhancement and the initial power features, so that final feature data are obtained; and sending the final characteristic data into a deep learning model for learning, so as to realize load monitoring. The invention can make the deep learning model more sensitive to the pulse signals in the data, better distinguish the state change of the load and improve the accuracy of load monitoring.
Description
Technical Field
The invention relates to the technical field of non-invasive load monitoring, in particular to a non-invasive load monitoring method based on low-frequency load multi-feature fusion.
Background
The non-invasive load monitoring technique, also called non-invasive load decomposition technique, can decompose the house total energy consumption data into the electricity consumption conditions of individual electric equipment. By utilizing the technology, a monitor can improve the energy utilization rate, reduce the energy waste and find out whether the equipment fails.
Currently, the non-invasive load monitoring method mainly comprises a mathematical optimization method, a hidden Markov model-based method and a deep learning-based method. The existing research shows that the deep learning method is higher than the other two methods in terms of resolution precision and accuracy by virtue of the strong nonlinear mapping capability.
However, the current non-invasive load monitoring method based on deep learning belongs to a data-driven method mostly, which causes that the accuracy of the method depends on the quantity and the quality of training data, and the learned features of the model are mostly learned for some data, and the learned features are not always universal, so that the generalization capability of the model cannot achieve the expected effect.
Currently, input data for non-invasive load monitoring includes high frequency data and low frequency data. The high-frequency data has higher requirements on hardware equipment, and most of the current domestic and foreign acquisition equipment only supports the acquisition of low-frequency data. Therefore, research based on low frequency data is of greater importance.
However, the low-frequency characteristic data contains less information, so that the important characteristic information of the low-frequency characteristic data is difficult to mine by a common deep learning model, and the low-frequency data-based deep learning model is low in decomposition accuracy.
Disclosure of Invention
The invention aims to overcome the defects and shortcomings of the prior art, and provides a non-invasive load monitoring method based on low-frequency load multi-feature fusion, which can effectively extract pulse features in input initial power data, and perform feature fusion on the pulse features and the initial power data, so as to provide more comprehensive feature information and improve the decomposition performance of a deep learning model.
In order to achieve the above purpose, the technical scheme provided by the invention is as follows: a non-invasive load monitoring method based on low-frequency load multi-feature fusion is provided, which aims at the problem of less exposure features of low-frequency sampled residential load data, uses VMD and kurtosis criterion to extract pulse features in initial power data, and performs feature fusion with the initial power data, thereby improving the accuracy of decomposing load power by a deep learning model, and comprises the following steps:
s1: acquiring total power data of a low-frequency load, and performing data cleaning on the total power data to remove useless data points so as to obtain useful data;
s2: carrying out data segmentation and enhancement on useful data by utilizing a sliding window technology to obtain initial power data;
s3: aiming at the initial power data, the number K of modal components of the VMD is obtained by utilizing a feedback VMD method, and the initial power data is decomposed to obtain K modal components;
s4: selecting a required modal component by using a kurtosis criterion and reconstructing a signal;
s5: enhancing the pulse characteristics of the reconstructed signal by using a Teager energy operator;
s6: feature fusion is carried out on the signals subjected to feature enhancement and the initial power signals, feature extraction is completed, and final feature data are obtained;
s7: and sending the final characteristic data into a deep learning model for learning, so as to realize load monitoring.
Further, in step S2, the initial power data length obtained after processing by the sliding window technique is N.
Further, in step S3, the specific operation steps of confirming the number K of modal components of the VMD by the feedback VMD are as follows:
s31: determining initial power data x (t) to be decomposed, initializing VMD parameters, and setting the number K of initial decomposition modal components init =2;
S32: using VMD pairsThe initial power data x (t) is decomposed to obtain two modal components u 1 (t) and u 2 (t);
S33: selecting two modal components u by calculating similarity coefficients of the modal components and the initial power data x (t) 1 (t) and u 2 The purest modal component in (t) is the purest modal component u with a large similarity coefficient pure (t) the similarity coefficient calculation formula is as follows:
in delta n,i Similarity coefficient representing the ith modal component of the nth iteration with the initial power data, u n,i Representing an ith modal component of an nth iteration, t=1, 2, …, N, where N represents a data length of the initial power data;
s34: updating the power data x (t) to be decomposed, namely:
x(t)=x(t)-u pure (t)
s35: the iterative process of steps S31-S34 is repeated until the following equation is satisfied:
max{δ k,1 ,δ k,2 }<min{δ k-1,1 ,δ k-1,2 }
indicating that the signal decomposition is complete; where k represents the final iteration number, delta k,1 Similarity coefficient, delta, representing the 1 st modal component of the kth iteration to the initial power data k,2 Similarity coefficient, delta, representing the 2 nd modal component of the kth iteration to the initial power data k-1,1 Similarity coefficient, delta, of the 1 st modal component representing the k-1 st iteration with the initial power data k-1,2 Representing similarity coefficients of the 2 nd modal component of iteration k-1 st time and the initial power data; at this time, the final modal decomposition number can be determined to be k=k+1;
the set number K of the modal components is smaller than the number M of the modal components using EMD, namely the iteration number K also meets the following formula:
k<M-1。
further, in step S4, the step of selecting a modal component using a kurtosis criterion and performing signal reconstruction includes:
s41: respectively calculating kurtosis values K of K modal components obtained by decomposition urt The expression of (2) is as follows:
wherein mu is u Sigma, which is the mean of the modal components u (t) u E (·) represents the expectation as the standard deviation of the modal component u (t);
s42: selecting two mode components with the largest kurtosis value and the second largest kurtosis valueAnd->Reconstructing it to obtain a reconstructed signal x rec (t):
Further, in step S5, for the reconstructed signal x rec (t) its Teager energy operator ψ [ x ] rec (t)]The expression of (2) is as follows:
ψ[x rec (t)]=[x rec (t)] 2 -x rec (t+1)·x rec (t-1)
wherein x is rec (t+1) represents the value of the reconstructed signal at point t+1, x rec (t-1) represents the value of the reconstructed signal at point t-1.
Further, in step S6, the feature enhanced signal is fused with the features of the initial power signal, and the feature enhanced signal and the features of the initial power signal are spliced in the feature dimension, so as to obtain feature data with a shape of (N, 2), where N represents the total number of data points of the initial power signal.
In step S7, the data x (t) 'after feature fusion is normalized by standard deviation in feature dimension, and the data after standard deviation normalization is used as input data x (t)', and is sent to a deep learning model for supervised learning; wherein, the expression of standard deviation normalization is as follows:
where x (t) "is data obtained by normalizing x (t) ' with standard deviation, μ is the mean value of the signal x (t) ' and σ is the standard deviation of the signal x (t) '.
Compared with the prior art, the invention has the following advantages and beneficial effects:
1. according to the invention, aiming at the low-frequency load time sequence data, the pulse characteristics are extracted, so that the subsequent deep learning model is more sensitive to pulse signals in the data, the state change of the load is better resolved, and the accuracy of load monitoring is improved.
2. The invention provides a feature fusion method, which combines a plurality of features to be used as the input of a subsequent deep learning model, and provides more comprehensive and more effective feature information for model learning.
3. The invention provides a feedback type VMD method, which can well solve the problem of determining the number K of mode components in the traditional VMD.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
FIG. 2 is a diagram showing the feature extraction results according to the present invention.
Detailed Description
The present invention will be described in further detail with reference to examples and drawings, but embodiments of the present invention are not limited thereto.
As shown in fig. 1, the present embodiment provides a non-invasive load monitoring method based on low-frequency load multi-feature fusion, which includes the following steps:
and step 1, acquiring total power data under low-frequency sampling, and performing data cleaning on the total power data to remove useless data points, so as to obtain useful data.
And 2, carrying out data segmentation and enhancement on the useful data by using a sliding window technology to obtain initial power data with the sequence length of N.
Step 3, decomposing the segmented data by using a feedback VMD to obtain a plurality of modal components;
the steps of the feedback VMD are as follows:
step 3.1, determining initial power data x (t) to be decomposed, initializing VMD parameters, and setting the number K of initial decomposition modal components init =2;
Step 3.2, decomposing the initial power data x (t) by VMD to obtain two modal components u 1 (t) and u 2 (t);
Step 3.3, selecting two modal components u 1 (t) and u 2 The purest modal component in (t). The purest modal component is determined by calculating the similarity coefficient between the modal component and the initial power data x (t), and the similarity coefficient is the purest modal component u pure (t) the similarity coefficient calculation formula is as follows:
in delta n,i Similarity coefficient representing the ith modal component of the nth iteration with the initial power data, u n,i Representing an ith modal component of an nth iteration, t=1, 2, …, N, where N represents a data length of the initial power data;
step 3.4, updating the power data x (t) to be decomposed, namely:
x(t)=x(t)-u pure (t)
and 3.5, repeating the iterative process of the steps 3.1-3.4 until the signal decomposition is complete, and determining the final modal decomposition number as K=k+1 when iterating K times.
Specifically, the conditions for complete signal decomposition are as follows:
max{δ k,1 ,δ k,2 }<min{δ k-1,1 ,δ k-1,2 }
because the VMD method can effectively avoid the problems of modal aliasing, over-enveloping, under-enveloping, boundary effect and the like, the set number K of modal components is smaller than the number M of modal components using EMD, namely the iteration number K is smaller than M-1.
And 4, calculating the kurtosis value of each modal component, selecting the modal component by using a kurtosis criterion, and reconstructing the signal.
Specifically, kurtosis value K urt The expression of (2) is as follows:
wherein mu is u Sigma, which is the mean of the modal components u (t) u E (·) represents the expectation, which is the standard deviation of the modal component u (t).
Selecting two modal components with the maximum kurtosis valueAnd->Reconstructing it to obtain a reconstructed signal x rec (t):
Step 5, enhancing the pulse characteristics of the reconstructed signal by using a Teager energy operator;
for reconstructed signal x rec (t) Teager energy operator ψ [ x ] rec (t)]The expression of (2) is as follows:
ψ[x rec (t)]=[x rec (t)] 2 -x rec (t+1)·x rec (t-1)
wherein x is rec (t+1) represents the value of the reconstructed signal at point t+1, x rec (t-1) represents the value of the reconstructed signal at point t-1.
The waveform diagram is shown in figure 2.
And 6, carrying out feature fusion on the signals subjected to feature enhancement and the original power signals to obtain new feature data, and completing feature extraction operation, wherein the method comprises the following steps of:
the signal subjected to characteristic enhancement and the original power signal are both time sequences with the length of N, and the signal subjected to characteristic enhancement and the original power signal are spliced in characteristic dimensions to obtain characteristic data with the shape of (N, 2), namely, characteristic fusion is completed.
And 7, respectively carrying out standard deviation normalization on the data x (t) 'after feature fusion on feature dimensions, and sending the data after standard deviation normalization to a deep learning model as input data x (t)', so as to carry out supervised learning. Wherein, the expression of standard deviation normalization is as follows:
where x (t) "is the data normalized by the standard deviation of the signal x (t) ' and μ is the mean value of the signal x (t) ' and σ is the standard deviation of the signal x (t) ' respectively.
And 8, sending the feature fusion data extracted from the original data into a deep learning model for learning.
After the deep learning model is learned, the extracted feature fusion data can be decomposed into power data of the target electric appliance.
The above examples are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above examples, and any other changes, modifications, substitutions, combinations, and simplifications that do not depart from the spirit and principle of the present invention should be made in the equivalent manner, and the embodiments are included in the protection scope of the present invention.
Claims (7)
1. The non-invasive load monitoring method based on low-frequency load multi-feature fusion is characterized in that the method aims at the problem of small exposure features of low-frequency sampled residential load data, adopts VMD and kurtosis criterion to extract pulse features in initial power data, and performs feature fusion on the pulse features and the initial power data, so that the accuracy of decomposing load power by a deep learning model is improved, and the method comprises the following steps:
s1: acquiring total power data of a low-frequency load, and performing data cleaning on the total power data to remove useless data points so as to obtain useful data;
s2: carrying out data segmentation and enhancement on useful data by utilizing a sliding window technology to obtain initial power data;
s3: aiming at the initial power data, the number K of modal components of the VMD is obtained by utilizing a feedback VMD method, and the initial power data is decomposed to obtain K modal components;
s4: selecting a required modal component by using a kurtosis criterion and reconstructing a signal;
s5: enhancing the pulse characteristics of the reconstructed signal by using a Teager energy operator;
s6: feature fusion is carried out on the signals subjected to feature enhancement and the initial power signals, feature extraction is completed, and final feature data are obtained;
s7: and sending the final characteristic data into a deep learning model for learning, so as to realize load monitoring.
2. The method of claim 1, wherein in step S2, the initial power data length obtained after processing by sliding window technique is N.
3. The non-invasive load monitoring method according to claim 2, wherein in step S3, the specific operation steps of confirming the number K of modal components of the VMD by the feedback VMD are as follows:
s31: determining initial power data x (t) to be decomposed, initializing VMD parameters, and setting the number K of initial decomposition modal components init =2;
S32: decomposing the initial power data x (t) using VMD to obtain two modal components u 1 (t) and u 2 (t);
S33: selection by calculating similarity coefficients of modal components to initial power data x (t)Two modal components u 1 (t) and u 2 The purest modal component in (t) is the purest modal component u with a large similarity coefficient pure (t) the similarity coefficient calculation formula is as follows:
in delta n,i Similarity coefficient representing the ith modal component of the nth iteration with the initial power data, u n,i Representing an ith modal component of an nth iteration, t=1, 2, …, N, where N represents a data length of the initial power data;
s34: updating the power data x (t) to be decomposed, namely:
x(t)=x(t)-u pure (t)
s35: the iterative process of steps S31-S34 is repeated until the following equation is satisfied:
max{δ k,1 ,δ k,2 }<min{δ k-1,1 ,δ k-1,2 }
indicating that the signal decomposition is complete; where k represents the final iteration number, delta k,1 Similarity coefficient, delta, representing the 1 st modal component of the kth iteration to the initial power data k,2 Similarity coefficient, delta, representing the 2 nd modal component of the kth iteration to the initial power data k-1,1 Similarity coefficient, delta, of the 1 st modal component representing the k-1 st iteration with the initial power data k-1,2 Representing similarity coefficients of the 2 nd modal component of iteration k-1 st time and the initial power data; at this time, the final modal decomposition number can be determined to be k=k+1;
the set number K of the modal components is smaller than the number M of the modal components using EMD, namely the iteration number K also meets the following formula:
k<M-1。
4. a non-invasive load monitoring method based on low frequency load multi-feature fusion according to claim 3, characterized in that in step S4 the step of selecting modal components using kurtosis criteria and performing signal reconstruction comprises:
s41: respectively calculating kurtosis values K of K modal components obtained by decomposition urt The expression of (2) is as follows:
wherein mu is u Sigma, which is the mean of the modal components u (t) u E (·) represents the expectation as the standard deviation of the modal component u (t);
s42: selecting two mode components with the largest kurtosis value and the second largest kurtosis valueAnd->Reconstructing it to obtain a reconstructed signal x rec (t):
5. The method of non-invasive load monitoring based on low frequency load multi-feature fusion according to claim 4, wherein in step S5, for the reconstructed signal x rec (t) its Teager energy operator ψ [ x ] rec (t)]The expression of (2) is as follows:
ψ[x rec (t)]=[x rec (t)] 2 -x rec (t+1)·x rec (t-1)
wherein x is rec (t+1) represents the value of the reconstructed signal at point t+1, x rec (t-1) represents the value of the reconstructed signal at point t-1.
6. The method of claim 5, wherein in step S6, the feature-enhanced signal is fused with the feature of the initial power signal by stitching the feature-enhanced signal and the feature of the initial power signal in a feature dimension, so as to obtain feature data with a shape of (N, 2), wherein N represents the total number of data points of the initial power signal.
7. The non-invasive load monitoring method based on low-frequency load multi-feature fusion according to claim 6, wherein in step S7, the feature fused data x (t)' are normalized by standard deviation in feature dimension, and the standard deviation normalized data are sent to a deep learning model as input data x (t) ", and are subjected to supervised learning; wherein, the expression of standard deviation normalization is as follows:
where x (t) "is data obtained by normalizing x (t) ' with standard deviation, μ is the mean value of the signal x (t) ' and σ is the standard deviation of the signal x (t) '.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310757598.8A CN116933000A (en) | 2023-06-25 | 2023-06-25 | Non-invasive load monitoring method based on low-frequency load multi-feature fusion |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310757598.8A CN116933000A (en) | 2023-06-25 | 2023-06-25 | Non-invasive load monitoring method based on low-frequency load multi-feature fusion |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116933000A true CN116933000A (en) | 2023-10-24 |
Family
ID=88391675
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310757598.8A Pending CN116933000A (en) | 2023-06-25 | 2023-06-25 | Non-invasive load monitoring method based on low-frequency load multi-feature fusion |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116933000A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117559450A (en) * | 2024-01-12 | 2024-02-13 | 吉林大学 | Method for improving non-intrusive load monitoring precision based on feedback model |
-
2023
- 2023-06-25 CN CN202310757598.8A patent/CN116933000A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117559450A (en) * | 2024-01-12 | 2024-02-13 | 吉林大学 | Method for improving non-intrusive load monitoring precision based on feedback model |
CN117559450B (en) * | 2024-01-12 | 2024-03-29 | 吉林大学 | Method for improving non-intrusive load monitoring precision based on feedback model |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN116933000A (en) | Non-invasive load monitoring method based on low-frequency load multi-feature fusion | |
CN111242377A (en) | Short-term wind speed prediction method integrating deep learning and data denoising | |
CN111949707B (en) | Shadow field-based hidden Markov model non-invasive load decomposition method | |
CN113158553A (en) | Magnetotelluric signal noise suppression method and system based on CNN-LSTM | |
CN112215405A (en) | Non-invasive type residential electricity load decomposition method based on DANN domain adaptive learning | |
CN112434891A (en) | Method for predicting solar irradiance time sequence based on WCNN-ALSTM | |
CN110726898A (en) | Power distribution network fault type identification method | |
CN110569728A (en) | Kernel signal extraction method based on dictionary training and orthogonal matching pursuit | |
CN111489305B (en) | Image enhancement method based on reinforcement learning | |
CN114742097B (en) | Optimization method for automatically determining variation modal decomposition parameters based on bearing vibration signals | |
CN115062733A (en) | Transformer voiceprint fault diagnosis method based on empirical mode decomposition and butterfly algorithm | |
CN117290788B (en) | Power distribution network fault identification method and system based on improved wavelet transformation algorithm | |
CN109586728B (en) | Signal blind reconstruction method under modulation broadband converter framework based on sparse Bayes | |
CN113627685B (en) | Wind driven generator power prediction method considering wind power internet load limit | |
CN113987910A (en) | Method and device for identifying load of residents by coupling neural network and dynamic time planning | |
CN113065704A (en) | Hyper-parameter optimization and post-processing method of non-invasive load decomposition model | |
CN117054803A (en) | Method and system for identifying grounding faults of distribution network containing distributed photovoltaic | |
CN117131022A (en) | Heterogeneous data migration method of electric power information system | |
CN116298699A (en) | Power distribution network fault identification method based on wavelet optimization and graph convolution neural network | |
CN116127343A (en) | Household transformer relation and phase identification method and system based on voltage sequence similarity | |
CN116070128A (en) | Time series equipment fault diagnosis method based on comparison self-supervision learning | |
CN114818827A (en) | Non-invasive load decomposition method based on seq2point network | |
CN113205050B (en) | Ship radiation noise line spectrum extraction method based on GRU-AE network | |
CN111797732B (en) | Video motion identification anti-attack method insensitive to sampling | |
CN114235396B (en) | Gear reducer fault feature extraction method |
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 |