CN118051722A - Multi-scale characteristic decomposition reconfigurable power line noise analysis method - Google Patents
Multi-scale characteristic decomposition reconfigurable power line noise analysis method Download PDFInfo
- Publication number
- CN118051722A CN118051722A CN202410086580.4A CN202410086580A CN118051722A CN 118051722 A CN118051722 A CN 118051722A CN 202410086580 A CN202410086580 A CN 202410086580A CN 118051722 A CN118051722 A CN 118051722A
- Authority
- CN
- China
- Prior art keywords
- noise
- power line
- data
- line noise
- feature
- 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
- 238000004458 analytical method Methods 0.000 title claims abstract description 25
- 238000000354 decomposition reaction Methods 0.000 title claims abstract description 25
- 238000013528 artificial neural network Methods 0.000 claims abstract description 32
- 125000004122 cyclic group Chemical group 0.000 claims abstract description 31
- 238000000034 method Methods 0.000 claims abstract description 24
- 238000013144 data compression Methods 0.000 claims abstract description 6
- 238000007906 compression Methods 0.000 claims abstract description 4
- 230000006835 compression Effects 0.000 claims abstract description 4
- 238000004891 communication Methods 0.000 claims description 13
- 230000002776 aggregation Effects 0.000 claims description 4
- 238000004220 aggregation Methods 0.000 claims description 4
- 238000012549 training Methods 0.000 claims description 4
- 239000011159 matrix material Substances 0.000 abstract description 5
- 239000000284 extract Substances 0.000 abstract description 2
- 238000005516 engineering process Methods 0.000 description 5
- 210000002569 neuron Anatomy 0.000 description 5
- 230000000737 periodic effect Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 3
- 230000001360 synchronised effect Effects 0.000 description 3
- 238000010276 construction Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 1
- 230000006866 deterioration Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Landscapes
- Cable Transmission Systems, Equalization Of Radio And Reduction Of Echo (AREA)
Abstract
The invention relates to a multi-scale feature decomposition reconfigurable power line noise analysis method, which aims at the field of power line noise analysis, and is characterized in that after power line mixed noise data are collected, noise data compression is carried out based on a Gaussian random matrix, effective noise feature information is extracted, similarity results of a power line noise database and compression noise feature data are obtained through least square comparison, and then a power line noise feature prediction result is optimized and output through constructing an aggregated LSTM wavelet cyclic neural network, so that the prediction accuracy of power line noise characteristics is improved. In addition, through carrying out multi-scale feature decomposition on the noise feature data, adopting a residual method to iteratively solve the best approximation sparse solution of the estimated value and the true value of the noise signal to be reconstructed, and reconstructing the original noise signal of the power line. Compared with the prior art, the method and the device can effectively extract the effective characteristic information of the power line noise data in the complex environment, restore the power line noise original signal to the greatest extent, and realize high-precision predictive analysis of the power line noise characteristics.
Description
Technical Field
The invention belongs to the technical field of power line noise analysis, and particularly relates to a multi-scale feature decomposition reconfigurable power line noise analysis method.
The background technology is as follows:
The development of smart power grids and the construction pace of novel power systems are continuously accelerated, higher requirements are put forward on the bearing capacity of power communication networks, and power line carrier communication plays an increasingly important role in the construction of the power communication networks with the remarkable advantages of wide coverage, low cost and the like. However, the continuous access of a large number of power electronic devices causes continuous deterioration of the environment of power line communication, and the complex noise generated by the continuous access seriously affects the communication quality, thereby greatly reducing the accuracy of data transmission.
Analyzing the power line noise characteristics has important significance in reducing power line noise interference and improving the reliability of communication. However, the existing power line noise analysis technology does not consider the similarity comparison result of the noise characteristic information, so that the prediction accuracy of the power line noise characteristic is not high, and meanwhile, the influence of multi-scale characteristic decomposition of the noise characteristic data on the reconstruction of the power line carrier channel noise and the prediction accuracy of the noise characteristic is not considered, so that the prediction accuracy of the noise characteristic is poor. There is therefore a strong need for a multi-scale feature decomposition reconfigurable power line noise analysis method.
However, the existing power line noise analysis technology is difficult to meet the accuracy and high precision requirements of power line noise prediction in a complex noise environment:
1) The traditional power line noise analysis technology does not consider the similarity comparison result of noise characteristic information, and is difficult to adapt to multi-type differentiated power line carrier channel noise prediction under the current complex power line communication noise environment, so that the prediction accuracy of the power line noise characteristics is not high, and therefore, how to compare the similarity result ratio to the accuracy of improving the power line noise prediction in the power line noise analysis is a problem to be solved urgently.
2) The conventional power line noise analysis technology does not consider the influence of multi-scale feature decomposition of noise feature data on reconstruction of power line carrier channel noise and noise feature prediction precision, and is difficult to realize continuous approximation of an estimated value and a true value of a signal to be reconstructed of noise, so that the prediction precision of the power line noise characteristic is not high enough. Therefore, how to perform multi-scale feature decomposition on the power line noise feature data to reconstruct the power line noise original signal further, and improving the prediction accuracy of the power line noise characteristics is another urgent problem to be solved.
The invention comprises the following steps:
In view of the above, the present invention aims to provide a multi-scale feature decomposition reconfigurable power line noise analysis method, so as to realize reliable power line noise characteristic high-precision prediction analysis.
In order to achieve the above purpose, the present invention provides the following technical solutions:
A multi-scale characteristic decomposition reconfigurable power line noise analysis method comprises the steps of obtaining a similarity result of a power line noise database and compression noise characteristic data by means of least square comparison, optimizing a power line noise characteristic prediction result by means of constructing an aggregated LSTM wavelet cyclic neural network, performing multi-scale characteristic decomposition on the noise characteristic data, and iteratively solving optimal approximation sparse solution of an estimated value and a true value of a noise signal to be reconstructed by means of a residual method to achieve high-precision prediction analysis of power line noise characteristics. The scheme specifically comprises the following steps:
s1: collecting power line mixed noise data, communication frequency bands and power line impedance data, and establishing a power line mixed noise database;
s2: and carrying out data compression on the power line mixed noise data, and extracting effective characteristic information in the power line mixed noise data.
S3: the compressed noise characteristic data, the communication frequency band and the power line impedance 3 data are assembled into a data set.
S4: and 5 standard power line noise characteristics in a power line noise database are extracted, and a least square method is utilized to compare the center distance with the noise characteristic data extracted in the step S2, so that a similarity result is obtained.
S5: 5 LSTM wavelet cyclic neural networks are respectively constructed based on the noise types, and historical data in a database are input into the LSTM wavelet cyclic neural networks for training.
S6: and constructing an aggregation LSTM wavelet cyclic neural network, and outputting optimized power line noise characteristic prediction data through the aggregation LSTM wavelet cyclic neural network.
S7: and carrying out multi-scale characteristic decomposition on the power line noise characteristic data.
S8: and solving the approximation sparse solution of the estimated value and the true value of the noise signal to be reconstructed by a residual error method, stopping iteration when the iteration precision requirement is met, and reconstructing the power line noise original signal.
Further, in the step S1, the power line hybrid noise database includes colored background noise, narrowband noise, power frequency asynchronous periodic impulse noise, power frequency synchronous periodic impulse noise, and random impulse noise.
Further, in step S2, the power line hybrid noise data compression method uses a gaussian random matrix to perform data compression on the power line hybrid noise data, extracts effective feature information therein, and the compressed noise feature data is a one-dimensional array.
Compared with the prior art, the invention has the following advantages:
1) Based on the similarity result, the method further obtains the aggregated LSTM wavelet cyclic neural network by constructing and training the 5 LSTM wavelet cyclic neural networks, outputs the optimized power line noise characteristic prediction data, and improves the prediction reliability of the power line noise characteristic.
2) According to the method, the power line noise characteristic data is subjected to multi-scale characteristic decomposition, the influence among characteristic components is reduced, the difficulty of a noise prediction process is reduced, and meanwhile, the estimated value and the true value of the noise signal to be reconstructed are continuously approximated by a residual error method until the error precision is met, so that the high-precision prediction analysis of the power line noise characteristic is realized.
Description of the drawings:
Fig. 1 is a flow chart of a multi-scale feature decomposition reconfigurable power line noise analysis method according to an embodiment of the invention.
Detailed Description
The invention will now be described in detail with reference to the drawings and specific examples. The embodiment is implemented on the premise of the technical scheme of the invention, and detailed implementation modes and specific operation processes are given. It should be noted that, without conflict, the embodiments of the present invention and features of the embodiments may be combined with each other.
Fig. 1 is a schematic flow chart of a multi-scale feature decomposition reconfigurable power line noise analysis method according to the embodiment, which specifically includes the following steps:
1) And collecting power line mixed noise data, communication frequency band and power line impedance data, and establishing a power line noise database comprising 5 kinds of noise such as colored background noise, narrowband noise, power frequency synchronous period impulse noise, power frequency asynchronous period impulse noise, random impulse noise and the like.
2) Based on Gaussian random matrix, carrying out data compression on power line mixed noise data, extracting effective characteristic information therein, reducing interference of useless information, wherein the compressed noise characteristic data is a one-dimensional array, and the number of data in the array is m, which is expressed as
Wherein [ U 1 U2 … Um-1 Um]T ] is power line compression noise characteristic data extracted through Gaussian random matrix, ψ i,j is measurement matrix element obeying Gaussian random distribution, [ P 1 P2 … Pn-1 Pn]T ] is collected power line mixed noise data, and m is less than n.
3) The compressed noise characteristic data, the communication frequency band and the power line impedance 3 data are assembled into a data set.
4) Extracting 5 standard power line noise characteristics in a power line noise database, defining the similarity theta of the standard power line noise characteristics and the compressed noise characteristic data, and comparing the center distance with the noise characteristic data extracted in the step S2 by using a least square method to obtain a similarity result theta, wherein the similarity result theta is expressed as
Wherein X i is the standard power line noise characteristic data extracted from the database, q i is the random error between the standard power line noise characteristic and the compressed noise characteristic data, and the smaller θ is, the closer the standard power line noise characteristic and the compressed noise characteristic data are, the higher the similarity is.
5) 5 LSTM wavelet cyclic neural networks (each input node is N+2) are constructed based on the noise type, and the LSTM wavelet cyclic neural networks comprise an input layer, a hidden layer and an output layer, wherein the input layer, the hidden layer and the output layer are respectively a background noise LSTM wavelet cyclic neural network, a narrow-band noise LSTM wavelet cyclic neural network, a power frequency asynchronous periodic impulse noise LSTM wavelet cyclic neural network, a power frequency synchronous periodic impulse noise LSTM wavelet cyclic neural network and a random impulse noise LSTM wavelet cyclic neural network. The number of neurons of the input layer is R, the number of neurons of each layer of the hidden layer is S, the number of neurons of the output layer is T, and then the historical data in the database is input into the LSTM wavelet cyclic neural network for training.
6) Based on the similarity result, the trained 5 LSTM wavelet cyclic neural networks are aggregated to obtain an aggregated LSTM wavelet cyclic neural network, a data set is input into the aggregated LSTM wavelet cyclic neural network, and optimized power line noise characteristic prediction data is output, which is expressed as
Wherein,An input of an LSTM wavelet recurrent neural network input layer of an i-th noise signature type,For the weight between the LSTM wavelet cyclic neural network input layer and the hidden layer of the ith noise characteristic type, lambda s,t (i) is the weight between the LSTM wavelet cyclic neural network hidden layer neuron and the output layer neuron of the ith noise characteristic type, omega is the wavelet transformation basis function, g s,t (i) is the expansion factor of the hidden layer of the ith noise characteristic type,/>For the output layer threshold of the ith noise feature type, u c (i) is power line noise feature prediction data optimized by the aggregated LSTM wavelet cyclic neural network.
7) Decomposing the power line noise characteristic data output by the aggregated LSTM wavelet cyclic neural network into a plurality of subcomponents with different characteristic scales so as to reduce the influence among the components, reduce the difficulty of a prediction process and reduce errors, scoring the importance degree of each characteristic on the power line noise analysis, constructing a characteristic subset, and carrying out multi-scale characteristic decomposition on the power line noise characteristic data according to the characteristic importance degree of each characteristic, wherein the expression formula of the multi-scale characteristic decomposition is as follows
Wherein,Zeta k (i) is a weight coefficient of the kth feature decomposition vector, which is the kth feature decomposition vector of the power line noise feature data, and represents the importance degree of the feature.
8) And performing descending iterative matching on each feature vector in the feature subset according to the importance degree by a residual error method, sorting and preferentially restoring important features of noise signals according to the importance degree of the power line noise features, then restoring secondary features of the noise signals, solving the approximate sparse solution of the estimated value and the true value of the noise signals to be reconstructed, judging whether the new residual error value reaches the reconstruction accuracy requirement U' i+1-ui I epsilon or not, stopping loop iteration if the new residual error value reaches the accuracy requirement, returning to the previous step to continue backtracking iteration if the accuracy requirement is not reached, and re-estimating until the original power line noise signals are reconstructed.
Although specific implementations of the invention and the accompanying drawings are disclosed for illustrative purposes only, and are presented to aid in understanding the invention and its implementation, it will be appreciated by those skilled in the art that: various alternatives, variations and modifications are possible without departing from the spirit and scope of the invention and the appended claims. Therefore, the present invention should not be limited to the preferred embodiments and the disclosure of the drawings, but the scope of the invention is defined by the appended claims.
Claims (4)
1. A multi-scale characteristic decomposition reconfigurable power line noise analysis method is characterized in that the method utilizes least square method comparison to obtain similarity results of a power line noise database and compression noise characteristic data, optimizes power line noise characteristic prediction results by constructing an aggregated LSTM wavelet cyclic neural network, then carries out multi-scale characteristic decomposition on the noise characteristic data, and utilizes a residual method to iteratively solve optimal approximation sparse solution of an estimated value and a true value of a noise signal to be reconstructed to realize high-precision prediction analysis of power line noise characteristics.
2. The method for analyzing the noise of the reconfigurable power line through multi-scale feature decomposition according to claim 1, wherein the method specifically comprises the following steps:
s1: collecting power line mixed noise data, communication frequency bands and power line impedance data, and establishing a power line mixed noise database;
s2: and carrying out data compression on the power line mixed noise data, and extracting effective characteristic information in the power line mixed noise data.
S3: the compressed noise characteristic data, the communication frequency band and the power line impedance 3 data are assembled into a data set.
S4: and 5 standard power line noise characteristics in a power line noise database are extracted, and a least square method is utilized to compare the center distance with the noise characteristic data extracted in the step S2, so that a similarity result is obtained.
S5: 5 LSTM wavelet cyclic neural networks are respectively constructed based on the noise types, and historical data in a database are input into the LSTM wavelet cyclic neural networks for training.
S6: and constructing an aggregation LSTM wavelet cyclic neural network, and outputting optimized power line noise characteristic prediction data through the aggregation LSTM wavelet cyclic neural network.
S7: and carrying out multi-scale characteristic decomposition on the power line noise characteristic data.
S8: and solving the approximation sparse solution of the estimated value and the true value of the noise signal to be reconstructed by a residual error method, stopping iteration when the iteration precision requirement is met, and reconstructing the power line noise original signal.
3. The method for constructing the aggregated LSTM wavelet cyclic neural network according to claim 2, wherein 5 kinds of LSTM wavelet cyclic neural networks are constructed based on noise types, similarity results are obtained by comparing standard power line noise characteristics with compressed noise characteristics through a least square method so as to adapt to analysis requirements of different types of noise, and the 5 kinds of LSTM wavelet cyclic neural networks are aggregated based on the similarity results so as to realize optimized output of power line noise characteristic prediction data.
4. The multi-scale feature decomposition of power line noise feature data according to claim 2, wherein the power line noise feature data is decomposed into a plurality of subcomponents with different feature scales to reduce the reconstruction difficulty of power line noise signals, feature importance scoring is carried out according to the importance degree of each feature on power line noise analysis, a feature subset is constructed, each feature vector in the feature subset is subjected to descending iterative matching according to the importance degree through a residual method, and the optimal approximation thinning solution of the estimated value and the true value of the noise signal to be reconstructed is solved.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410086580.4A CN118051722A (en) | 2024-01-22 | 2024-01-22 | Multi-scale characteristic decomposition reconfigurable power line noise analysis method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410086580.4A CN118051722A (en) | 2024-01-22 | 2024-01-22 | Multi-scale characteristic decomposition reconfigurable power line noise analysis method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN118051722A true CN118051722A (en) | 2024-05-17 |
Family
ID=91052970
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410086580.4A Pending CN118051722A (en) | 2024-01-22 | 2024-01-22 | Multi-scale characteristic decomposition reconfigurable power line noise analysis method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118051722A (en) |
-
2024
- 2024-01-22 CN CN202410086580.4A patent/CN118051722A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112818891B (en) | Intelligent identification method for communication interference signal type | |
| Yuan et al. | DeepMorse: A deep convolutional learning method for blind morse signal detection in wideband wireless spectrum | |
| CN111310700A (en) | Intermediate frequency sampling sequence processing method for radiation source fingerprint feature identification | |
| CN114881093B (en) | Signal classification and identification method | |
| CN112132266A (en) | Signal modulation identification system and modulation identification method based on convolution cycle network | |
| Yang et al. | One-dimensional deep attention convolution network (ODACN) for signals classification | |
| CN115496100A (en) | Perimeter security disturbance identification algorithm based on GAF-ConvNeXt-TF | |
| CN114757224A (en) | Specific radiation source identification method based on continuous learning and combined feature extraction | |
| CN118051722A (en) | Multi-scale characteristic decomposition reconfigurable power line noise analysis method | |
| Huang et al. | FM broadcast monitoring using artificial intelligence | |
| Liu et al. | Hierarchical agglomerative clustering and LSTM-based load prediction for dynamic spectrum allocation | |
| CN113435263B (en) | CGAN data enhancement-based frequency spectrum sensing method and system | |
| Shen et al. | Deep learning based source number estimation with single-channel mixtures | |
| CN114662523A (en) | Spread spectrum signal detection method integrating cognitive enhanced compressed sensing and artificial neural network | |
| CN113095215A (en) | Solar radio filtering method and system based on improved LSTM network | |
| CN116761223B (en) | Method for realizing 4G radio frequency communication by using 5G baseband chip and vehicle-mounted radio frequency system | |
| Hao et al. | Evaluation of interference effect of speech communication system | |
| Qi et al. | Radar signal recognition based on deep convolutional neural network in complex electromagnetic environment | |
| Yang et al. | Radar Emitter Recognition Method Based on Bispectrum and Improved Multi-channel Convolutional Neural Network | |
| Li et al. | FETTrans: Analysis of Compound Interference Identification Based on Bidirectional Dynamic Feature Adaptation of Improved Transformer | |
| Hossen et al. | A two-stage network for modulation classification based on entropy of wavelet transform | |
| Ying et al. | Channel Attention Mechanism-based Multi-Feature Fusion Network for Specific Emitter Identification | |
| CN118070116A (en) | Intelligent sensing method, system, equipment and medium for abnormal electromagnetic spectrum under non-Gaussian noise | |
| CN115238768A (en) | Wireless protocol identification method based on deep learning | |
| CN112529035A (en) | Intelligent identification method for identifying individual types of different radio stations |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination |