CN107491412B - User electricity load characteristic extraction method based on empirical wavelet transform - Google Patents
User electricity load characteristic extraction method based on empirical wavelet transform Download PDFInfo
- Publication number
- CN107491412B CN107491412B CN201710557184.5A CN201710557184A CN107491412B CN 107491412 B CN107491412 B CN 107491412B CN 201710557184 A CN201710557184 A CN 201710557184A CN 107491412 B CN107491412 B CN 107491412B
- Authority
- CN
- China
- Prior art keywords
- user
- empirical wavelet
- empirical
- load data
- power load
- 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.)
- Expired - Fee Related
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
-
- 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
Abstract
The invention discloses a user electricity load characteristic extraction method based on empirical wavelet transform, which comprises the following steps: decomposing user electricity load data into a plurality of empirical wavelet function components by empirical wavelet transform; performing Hilbert transformation on each empirical wavelet function, and representing a Hilbert spectrum in a time-frequency joint domain; calculating the normalized correlation coefficient of each component and the original load data, setting a proper threshold according to a correlation degree selection rule, selecting an effective component with higher correlation degree with the original load data, carrying out 2-norm solution on the energy of the extracted effective component, and combining the solution results to form a feature vector capable of comprehensively representing the power utilization characteristics of the user. The method combines the advantages of wavelet analysis and empirical mode decomposition methods, constructs an orthogonal wavelet filter bank by adaptively dividing Fourier frequency spectrum, decomposes a single signal into a plurality of amplitude modulation-frequency modulation components with tightly supported Fourier frequency spectrum, and realizes the separation of different frequency characteristic information.
Description
Technical Field
The invention relates to the field of extraction of user power load characteristics of a power system, in particular to a user power load characteristic extraction method based on empirical wavelet transform.
Background
At present, more and more sensors and intelligent instruments are installed in an intelligent power distribution network to acquire real-time data of a power network. The user power load data plays an important role in making power grid operation decisions, so that feature extraction research on the power load data has important theoretical significance and practical value for supporting safe, economic and reliable operation of a power grid.
The current common methods for signal feature extraction include: fourier transform, wavelet analysis method, wavelet packet analysis method, hilbert-yellow transform, and the like. The fourier transform uses a global sinusoid, which effectively divides the signal into components of different frequencies, but results in the loss of local information. Although wavelet and wavelet packet transformation can be used for analyzing non-stationary signals, the problem of wavelet base selection needs to be solved in practical application, and the wavelet and wavelet packet transformation is not adaptive. An empirical mode decomposition algorithm related in the Hilbert-Huang transform method makes up the defect that the prior method depends on prior knowledge to select a basis function, but the algorithm lacks powerful mathematical theory support and mode aliasing occurs in the signal processing process. The empirical wavelet transform method solves the problems of insufficient theoretical basis and algorithm adaptability, aims to construct a basis function directly generated by signal characteristics, and further extracts different modes of signals by designing a proper wavelet filter group.
Disclosure of Invention
The invention aims to provide a power load characteristic extraction method based on empirical wavelet transform, and provides an effective means for analyzing the power consumption characteristics of users.
The invention relates to a user electricity load characteristic extraction method based on empirical wavelet transform, which comprises the following steps:
decomposing original power load data of a user as an original signal by adopting an empirical wavelet transform method to obtain a plurality of empirical wavelet function components;
hilbert transformation is carried out on each obtained empirical wavelet function component to obtain a Hilbert spectrum representing the instantaneous amplitude and the instantaneous frequency of each component, and the Hilbert spectrum is represented in a time-frequency combined domain, so that time-frequency representation of the original electric load data signal of the user is established;
calculating normalized correlation coefficients of the original power load data of the user and the empirical wavelet function components, and screening out effective components with high correlation degrees with the original power load data of the user;
and constructing a user power load data feature vector according to the effective component, and outputting the feature vector as the feature for identifying the original power load of the user.
Further, Fourier transform is carried out on the original signal to obtain a Fourier spectrum of the original signal, an empirical wavelet filter bank is constructed through self-adaptive segmentation of the Fourier spectrum, amplitude modulation-frequency modulation (AM-FM) signal components with tightly supported Fourier spectrum are extracted, components containing different frequency characteristic information in the signal components are separated, and a plurality of empirical wavelet function components containing different frequency characteristic information are obtained.
The screening is to set a proper threshold value according to a relevancy selection rule to select an effective component which has high relevancy and can reflect the characteristics of the user power load, so as to achieve the purpose of removing a pseudo component and an irrelevant interference item.
Further, the method for constructing the characteristic vector of the electric load data according to the effective components is to calculate the energy of each effective component, carry out 2-norm solution on the energy, combine the solution results and construct the characteristic vector capable of comprehensively representing the electric characteristics of the user.
When the normalized correlation coefficient is 0.5 or more, the correlation is regarded as high.
Drawings
FIG. 1 is a block diagram of the steps of a user electrical load feature extraction method based on empirical wavelet transform according to the present invention;
FIG. 2 is a schematic diagram of raw customer electrical load data collected by a sensor;
FIG. 3 is a diagram of empirical wavelet function components generated by adaptive decomposition of original customer electrical load data signals using empirical wavelet transform;
fig. 4 is a schematic diagram of time-frequency representation of each empirical wavelet function component after Hilbert transform.
Detailed Description
The following describes the steps of the present invention with reference to the accompanying drawings.
A user electricity load feature extraction method based on empirical wavelet transform comprises the following steps:
S1: FIG. 2 is a diagram of raw customer electrical load data collected by a sensor using a time domain discrete sequence X ═ X1,x2,…,xN]And (4) performing representation. The sequence is fourier transformed to obtain a fourier spectrum F (ω).
S2: the Fourier spectrum is adaptively segmented to construct a suitable wavelet filter bank. The spectrum division principle is as follows:
assume by ωnFor the boundary, the frequency spectrum F (omega) is set at [0, pi ]]The upper division is N continuous paragraphs, the frequency spectrum division comprises N +1 boundaries, and the boundary omega is removed0=0,ωNAnd the determination method of the rest N-1 boundaries comprises the following steps: local maxima in the spectrum are first detected and sorted in descending order (including 0 and π). Assume that there are M maxima:
if M is larger than or equal to N, the fact that enough maximum values exist to define frequency spectrum region segmentation is indicated, and the first N-1 maximum values are reserved;
if M is less than N, the decomposition mode of the actual data signal is less than the expected value, all the maximum values are reserved, and N is reset to a proper value.
Taking the intermediate frequency of two continuous maximum values as omeganAnd ω0=0,ωN=πTogether forming the final spectral partitioning result.
S3: and constructing N empirical wavelets according to the Fourier spectrum segmentation in S2, and determining the empirical wavelet function components. Establishing an empirical wavelet basis, an empirical scale functionAnd empirical wavelet mother functionThe definition is as follows:
in the formula: the function β (x) is an arbitrary Ck([0,1]) The function, needs to satisfy the following condition:
τn=γωn
Obtaining an empirical wavelet transform by an inner product of empirical wavelets:
the approximation coefficients are obtained by an inner product with a scaling function:
the empirical wavelet transform of signal f (t) is constructed as follows:
mode f of empirical decompositionkGiven by:
the wavelet filter bank constructed according to the spectrum adaptive segmentation decomposes the original user electrical load data signal into a plurality of empirical wavelet function components covering different frequency characteristic information, as shown in fig. 3.
And 2, respectively carrying out Hilbert transformation on the N +1 empirical wavelet functions, and establishing time-frequency representation of each component.
Performing Hilbert transform on each empirical wavelet function component, the transform being as follows:
where (p, υ.) is the cauchy principal value and f (t) is the empirical wavelet function component.
The analytic form of the signal is as follows:
fa(t)=f(t)+ιHf(t)
wherein the AM-FM signal form is:
the signal analytic form can be expressed as:
from this, the instantaneous amplitude F (t) and instantaneous frequency of each empirical wavelet function can be extractedFig. 4 shows a time-frequency representation of each empirical wavelet function after Hilbert transform.
And 3, extracting effective components with higher correlation degree with the original load data signals.
Calculating normalized correlation coefficient r between each empirical wavelet function and electric load data signalj,rjThe formula for (j ═ 1,2, …, n) is:
wherein xiIs the ith element of the electrical load signal sequence, x is the mean value of the electrical load signal sequence, fiFor the ith element of the jth empirical wavelet function component,is the component mean of the jth empirical wavelet function.
In general, when the normalized correlation coefficient is greater than or equal to 0.5, it is regarded as high correlation, and the normalized correlation coefficient r is selected according to the correlation selection rulejAnd taking the component more than or equal to 0.5 as an effective component for solving the feature vector. Through calculation, 4 empirical wavelet functions are selected as effective components.
And 4, constructing a feature vector.
S1, calculating the energy E of each effective component extracted in the step 3j。
Ej=∫|fj(t)|2dt
A 2-norm calculation is performed for each energy S2.
S3: and combining the calculation results to construct a feature vector V capable of comprehensively representing the electricity utilization characteristics of the user.
V=[v1,v2,...,vN]
And calculating that the characteristic vector of the user electrical load data is V ═ 21.2919.6718.5711.89.
Claims (5)
1. A user electricity load feature extraction method based on empirical wavelet transform comprises the following steps:
decomposing original power load data of a user as an original signal by adopting an empirical wavelet transform method to obtain a plurality of empirical wavelet function components;
hilbert transformation is carried out on each obtained empirical wavelet function component to obtain a Hilbert spectrum representing the instantaneous amplitude and the instantaneous frequency of each component, and the Hilbert spectrum is represented in a time-frequency combined domain, so that time-frequency representation of the original electric load data signal of the user is established;
calculating normalized correlation coefficients of the original power load data of the user and the empirical wavelet function components, and screening out effective components with high correlation degrees with the original power load data of the user;
and constructing a user power load data feature vector according to the effective component, and outputting the feature vector as the feature for identifying the original power load of the user.
2. The method for extracting the characteristics of the electric load of the user based on the empirical wavelet transform as claimed in claim 1, wherein: the specific method for obtaining a plurality of empirical wavelet function components comprises the following steps: the method comprises the steps of carrying out Fourier transform on an original signal to obtain a Fourier spectrum of the original signal, constructing an empirical wavelet filter bank through self-adaptive segmentation of the Fourier spectrum, extracting amplitude modulation-frequency modulation (AM-FM) signal components with tightly supported Fourier spectrum, and separating components containing different frequency characteristic information in the signal components to obtain a plurality of empirical wavelet function components containing different frequency characteristic information.
3. The method for extracting the characteristics of the electric load of the user based on the empirical wavelet transform as claimed in claim 2, wherein: the screening is to set a proper threshold value according to a relevancy selection rule to select an effective component which has high relevancy and can reflect the characteristics of the user power load, so as to achieve the purposes of removing a pseudo component and an irrelevant interference item.
4. The method for extracting the characteristics of the user power load based on the empirical wavelet transform as claimed in claim 3, wherein the method for constructing the characteristic vector of the power load data according to the effective components is to calculate the energy of each effective component, perform 2-norm solution on the energy, combine the solution results, and construct the characteristic vector capable of comprehensively representing the characteristics of the user power load.
5. The method as claimed in claim 3, wherein the correlation is considered to be high when a normalized correlation coefficient is greater than or equal to 0.5.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710557184.5A CN107491412B (en) | 2017-07-10 | 2017-07-10 | User electricity load characteristic extraction method based on empirical wavelet transform |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710557184.5A CN107491412B (en) | 2017-07-10 | 2017-07-10 | User electricity load characteristic extraction method based on empirical wavelet transform |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107491412A CN107491412A (en) | 2017-12-19 |
CN107491412B true CN107491412B (en) | 2021-01-12 |
Family
ID=60644352
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710557184.5A Expired - Fee Related CN107491412B (en) | 2017-07-10 | 2017-07-10 | User electricity load characteristic extraction method based on empirical wavelet transform |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107491412B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109269629A (en) * | 2018-08-03 | 2019-01-25 | 河海大学 | Paralleling reactor of extra-high voltage analysis of vibration signal method based on experience wavelet transformation |
CN110490369A (en) * | 2019-07-25 | 2019-11-22 | 国网安徽省电力有限公司 | A kind of Short-Term Load Forecasting Method based on EWT and LSSVM model |
CN112200384A (en) * | 2020-10-28 | 2021-01-08 | 宁波立新科技股份有限公司 | Power load short-time prediction method based on EWT neural network |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102855385A (en) * | 2012-07-31 | 2013-01-02 | 上海交通大学 | Wind power generation short-period load forecasting method |
CN103390117A (en) * | 2013-08-08 | 2013-11-13 | 山东大学 | Feature extracting method for power load dynamic features |
CN103822786A (en) * | 2012-11-16 | 2014-05-28 | 中国水利电力物资有限公司 | Wind turbine generator set mechanical equipment state diagnosis method based on multivariate statistical analysis |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6859764B2 (en) * | 2003-04-03 | 2005-02-22 | The United States Of America As Represented By The Secretary Of The Army | Detecting, classifying and localizing minor amounts of an element within a sample of material |
-
2017
- 2017-07-10 CN CN201710557184.5A patent/CN107491412B/en not_active Expired - Fee Related
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102855385A (en) * | 2012-07-31 | 2013-01-02 | 上海交通大学 | Wind power generation short-period load forecasting method |
CN103822786A (en) * | 2012-11-16 | 2014-05-28 | 中国水利电力物资有限公司 | Wind turbine generator set mechanical equipment state diagnosis method based on multivariate statistical analysis |
CN103390117A (en) * | 2013-08-08 | 2013-11-13 | 山东大学 | Feature extracting method for power load dynamic features |
Also Published As
Publication number | Publication date |
---|---|
CN107491412A (en) | 2017-12-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107491412B (en) | User electricity load characteristic extraction method based on empirical wavelet transform | |
De Yong et al. | An effective power quality classifier using wavelet transform and support vector machines | |
CN109633368B (en) | VMD and DFA-based power quality disturbance detection method for power distribution network with distributed power supplies | |
Qiu et al. | Power quality disturbances recognition using modified s transform and parallel stack sparse auto-encoder | |
Sirojan et al. | Sustainable deep learning at grid edge for real-time high impedance fault detection | |
CN109633431B (en) | On-load tap-changer fault identification method based on vibration signal feature extraction | |
CN108229382A (en) | Vibration signal characteristics extracting method, device, storage medium and computer equipment | |
CN105356886B (en) | Power line noise compression method and device based on compressed sensing | |
CN109975673B (en) | Method for identifying fault arc at direct current side of photovoltaic microgrid | |
Wang et al. | Application of the dual-tree complex wavelet transform in biomedical signal denoising | |
Ferreira et al. | Method based on independent component analysis for harmonic extraction from power system signals | |
Wang et al. | Noise suppression of corona current measurement from HVdc transmission lines | |
CN113537102B (en) | Feature extraction method of microseismic signals | |
CN108008187B (en) | Power grid harmonic detection method based on variational modal decomposition | |
CN110609088B (en) | Post insulator fault identification method and device | |
Taskovski et al. | Classification of power quality disturbances using wavelets and support vector machine | |
CN108710861B (en) | Power grid disturbance signal detection and identification method | |
Krishna et al. | Image pattern recognition technique for the classification of multiple power quality disturbances | |
Krishna et al. | Efficient genetic-wrapper algorithm based data mining for feature subset selection in a power quality pattern recognition applicationction in a power quality pattern recognition application. | |
Qian et al. | Network traffic forecasting by support vector machines based on empirical mode decomposition denoising | |
CN108108659B (en) | Island detection key feature extraction method based on empirical mode decomposition | |
CN107590788A (en) | A kind of remote sensing image processing method | |
CN112863536A (en) | Environmental noise extraction method and device, computer equipment and storage medium | |
Ting et al. | Eeg signal processing based on proper orthogonal decomposition | |
Bank et al. | Extraction and visualization of power system interarea oscillatory modes |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20210112 Termination date: 20210710 |
|
CF01 | Termination of patent right due to non-payment of annual fee |