CN112801123A - Small sample user electricity consumption data expansion method with frequency domain distribution consistency - Google Patents

Small sample user electricity consumption data expansion method with frequency domain distribution consistency Download PDF

Info

Publication number
CN112801123A
CN112801123A CN202010938107.6A CN202010938107A CN112801123A CN 112801123 A CN112801123 A CN 112801123A CN 202010938107 A CN202010938107 A CN 202010938107A CN 112801123 A CN112801123 A CN 112801123A
Authority
CN
China
Prior art keywords
data
frequency domain
user
small sample
electricity
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202010938107.6A
Other languages
Chinese (zh)
Inventor
陆俊
李国成
王赟
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
North China Electric Power University
Original Assignee
North China Electric Power University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by North China Electric Power University filed Critical North China Electric Power University
Priority to CN202010938107.6A priority Critical patent/CN112801123A/en
Publication of CN112801123A publication Critical patent/CN112801123A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • 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
    • 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
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/047—Probabilistic or stochastic networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Software Systems (AREA)
  • Computational Linguistics (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Health & Medical Sciences (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Algebra (AREA)
  • Databases & Information Systems (AREA)
  • Supply And Distribution Of Alternating Current (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention provides a small sample user electricity consumption data expansion method with consistent frequency domain distribution, which is characterized in that daily electricity consumption load data of small sample electricity consumption users are acquired by collection on the basis of fully considering the frequency domain characteristics of user electricity consumption data, frequency domain mapping vectors of the daily electricity consumption load data of the small sample users are calculated, a probabilistic distribution variance minimum generation confrontation network method is adopted to expand and generate frequency domain user data to be selected, frequency domain user electricity consumption data are screened according to a sample accumulative cosine divergence minimum criterion, and expanded time domain small sample user electricity consumption data are calculated based on frequency domain inverse transformation; the consistency of the power consumption expansion data of the small sample user and the original data on the frequency domain distribution is effectively improved.

Description

Small sample user electricity consumption data expansion method with frequency domain distribution consistency
Technical Field
The invention belongs to the field of intelligent power utilization user power utilization behavior identification, and particularly relates to a small sample user power utilization data expansion method with frequency domain distribution consistency.
Background
The identification of the power utilization behaviors of the users is a necessary measure for realizing intelligent energy comprehensive service under the background of the power Internet of things. When the big data are used for analyzing the electricity consumption behaviors of the users in an artificial intelligence mode, the data set samples are often unbalanced, the classification of the electricity consumption data is inaccurate, and the subsequent data analysis and processing are problematic, so that the expansion of the small sample data through the small sample user electricity consumption data generation method has important significance in balancing the data sets. The generation of the countermeasure network is a deep learning model, is good at generating the expansion data which is similar to the input data distribution, and is a new method for solving the expansion problem of the small sample data.
On one hand, the existing method for expanding the power consumption data of the small sample user is a simple copy of a sample in terms of a generation algorithm and has no diversity, so that the generalization capability of subsequent analysis is not strong; on the other hand, the classification is performed by considering the time domain, and the characteristics of the user electricity data in the frequency domain are not sufficiently considered. Aiming at the limitation of the existing small sample user electricity consumption data generation method, the invention acquires daily electricity consumption load data of the small sample user by collecting on the basis of fully considering the frequency domain characteristics of the user electricity consumption data, calculates the frequency domain mapping vector of the daily electricity consumption load data of the small sample user, adopts a probabilistic distribution variance minimum generation confrontation network method to expand and generate frequency domain user candidate data, screens the frequency domain user electricity consumption data according to the minimum sample cumulative cosine divergence criterion, and calculates the expanded time domain small sample user electricity consumption data based on frequency domain inverse transformation; the consistency of the power consumption expansion data of the small sample user and the original data on the frequency domain distribution is effectively improved.
Disclosure of Invention
The invention aims to provide a small sample user electricity consumption data expansion method with frequency domain distribution consistency, which comprises the following steps:
s1, acquiring daily electricity load data of a small sample electricity user;
s2, calculating a frequency domain mapping vector of daily electric load data of the small sample user;
s3, expanding and generating frequency domain user data to be selected by adopting a method of generating a confrontation network by using a probability distribution variance minimum;
s4, screening frequency domain user power utilization data according to a sample accumulated cosine divergence minimum criterion;
and S5, calculating the expanded time domain small sample user power utilization data based on the frequency domain inverse transformation.
Compared with the general technology, the method has the advantages that on the basis of fully considering the frequency domain characteristics of the user electricity utilization data, the daily electricity utilization load data of the small sample electricity utilization user are acquired through collection, the frequency domain mapping vector of the daily electricity utilization load data of the small sample electricity utilization user is calculated, the confrontation network method is expanded by adopting the minimum probability distribution variance to generate the frequency domain user data to be selected, the frequency domain user electricity utilization data are screened according to the minimum sample cumulative cosine divergence criterion, and the expanded time domain small sample user electricity utilization data are calculated on the basis of frequency domain inverse transformation; the consistency of the power consumption expansion data of the small sample user and the original data on the frequency domain distribution is effectively improved.
Drawings
FIG. 1 is an overall flow chart of the method of the present invention.
Fig. 2 is a diagram of a typical application scenario of the method of the present invention.
Detailed Description
The preferred embodiments will be described in detail below with reference to the accompanying drawings. It should be emphasized that the following description is merely exemplary in nature and is not intended to limit the scope of the invention or its application. Referring to FIG. 1, a flow chart of the method of the present invention is shown, and the method of the present invention is further described with reference to specific examples.
The specific implementation application scenario is shown in fig. 2, the total number of small sample power users is 2, daily power load data of the small sample users after classification is obtained through an intelligent electric meter, the sampling interval is 0.5 hour, frequency domain mapping quantity is calculated through Fourier transform, frequency domain expansion data are generated through a generation antagonistic network, high-quality frequency domain data are screened based on the minimum rule of cosine distance between distributions, and time domain small sample expansion data are obtained through frequency domain inverse transformation. The obtained expansion data can be used as the predicted load data of the power consumption of the original small sample user, and can also be combined for data analysis, so that various problems caused by the similar unbalance phenomenon are solved.
The invention provides a small sample user electricity consumption data expansion method with consistent frequency domain distribution, which comprises the following steps:
s1, acquiring daily electricity load data of small sample electricity users
The total number of the power users of the small sample is set to be K (the value range of K is 1-100), and the multiple of the expanded large sample is set to be Z (the value range is5-10), and the expansion factor is lambda (the value range is 2-5); daily electricity load data of each small sample electricity user K (the value range of K is 1-K) is acquired by using a household intelligent electricity meter, and electricity data D in the ith (the value range of i is 1-50) statistical time of the small sample electricity user are periodically acquired according to a sampling interval T (unit hour)k(i) Acquiring a daily electrical load data set { Dk(i) (} (i ═ 1.., N), where
Figure RE-GDA0003012767850000031
floor (×) denotes the round-down operation.
In the example, the total number of the small sample power users is K ═ 2, the expansion large sample multiple is Z ═ 2, and the expansion factor is λ ═ 2; acquiring daily electricity load data of each small sample electricity user K (the value range of K is 1-K ()) by using a household intelligent electricity meter, and periodically acquiring electricity data D of the small sample electricity user in the ith (the value range of i is 1-50) statistical time according to the sampling interval T of 0.5 (unit hour)k(i) Acquiring a daily electrical load data set { Dk(i) (} (i ═ 1.., N), where
Figure RE-GDA0003012767850000041
floor (×) denotes the round-down operation.
S2, calculating a frequency domain mapping vector of daily electric load data of the small sample user
For each small sample electricity user k, according to the formula
Figure RE-GDA0003012767850000042
Calculating daily electric load data frequency domain mapping quantity X of small sample electric user kk(s) (s ═ 1.. said., N), obtaining daily electricity data frequency domain mapping quantity vector of all users
Figure RE-GDA0003012767850000043
Figure RE-GDA0003012767850000044
In the example, for each small sample, the electricity user k is used according to the formula
Figure RE-GDA0003012767850000045
Calculating daily electric load data frequency domain mapping quantity X of small sample electric user kk(s) (s ═ 1.. said., N), obtaining daily electricity data frequency domain mapping quantity vector of all users
Figure RE-GDA0003012767850000046
S3, adopting a method of generating a countermeasure network by minimum probability distribution variance to expand and generate frequency domain user candidate data
The daily electricity load data frequency domain mapping vector of K small sample electricity users is used as data input, a gradient descent method is adopted, the countermeasure network is subjected to data expansion training according to the minimum probability distribution variance criterion, P lambda Z frequency domain user data to be selected are generated through expansion, wherein the P (P1
Figure RE-GDA0003012767850000051
In the example, daily electricity load data frequency domain mapping vectors of K small sample electricity users are used as data input, a gradient descent method is adopted, a countermeasure network is generated according to a probability distribution variance minimum criterion for data expansion training, P lambda Z frequency domain user data to be selected are generated through expansion, wherein the P (P1, the
Figure RE-GDA0003012767850000052
S4, screening frequency domain user electricity utilization data according to the minimum criterion of sample accumulated cosine divergence
For the p-th user day frequency domain data to be selected generated by each expansion, according to a formula
Figure RE-GDA0003012767850000053
Calculating the sample accumulated cosine divergence Q (p) of the data to be selected; according to the value of Q (p), the daily frequency of the user to be selected is increased from small to largeCarrying out sequencing operation on the domain data vectors; selecting the front Z user data to be selected as the power consumption daily frequency domain data of the extended user
Figure RE-GDA0003012767850000054
In the example, the data of the p-th frequency domain of the user to be selected generated by each expansion is processed according to a formula
Figure RE-GDA0003012767850000055
Calculating the sample accumulated cosine divergence Q (p) of the data to be selected; sequencing the day frequency domain data vectors of the users to be selected according to the value Q (p) from small to large; selecting the front Z user data to be selected as the power consumption daily frequency domain data of the extended user
Figure RE-GDA0003012767850000056
S5, time domain small sample user electricity consumption data based on frequency domain inverse transformation calculation expansion
For each z th extended user, the power consumption daily frequency domain data
Figure RE-GDA0003012767850000057
According to the formula
Figure RE-GDA0003012767850000058
Sequentially calculating q ═ 1.. and N time domain components H of the expansion dataz(q) (q ═ 1.., N), constituting extended daily electricity load data { H @z(q) } (q 1.., N), completing the operation of expanding the electricity data by the small sample user.
In the example, the electricity utilization daily frequency domain data of every z th expansion user
Figure RE-GDA0003012767850000061
According to the formula
Figure RE-GDA0003012767850000062
Sequentially calculating q ═ 1.. and N time domain components H of the expansion dataz(q) (q ═ 1.., N), constituting extended daily electricity load data { H @z(q)}(q＝1,...,N)，And finishing the electricity consumption data expansion operation of the small sample user.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims (5)

1. A small sample user electricity consumption data expansion method with frequency domain distribution consistency comprises the following steps:
s1, acquiring daily electricity load data of a small sample electricity user;
s2, calculating a frequency domain mapping vector of daily electric load data of the small sample user;
s3, expanding and generating frequency domain user data to be selected by adopting a method of generating a confrontation network by using a probability distribution variance minimum;
s4, screening frequency domain user power utilization data according to a sample accumulated cosine divergence minimum criterion;
and S5, calculating the expanded time domain small sample user power utilization data based on the frequency domain inverse transformation.
2. The method for expanding power consumption data of small sample users according to the frequency domain distribution consistency of claim 1, wherein in the step S1, the total number of the small sample power consumption users is K (the value range of K is 1 to 100), the multiple of the expanded large sample is Z (the value range is 5 to 10), and the expansion factor is λ (the value range is 2 to 5); the sampling interval is T (unit hour), the electricity utilization data in the ith (i value range is 1-50) statistical time of the daily electricity utilization load data of the small sample electricity utilization user K (K value range is 1-K) is Dk(i) And the k daily electricity load data set of the small sample electricity user is { D }k(i)}(i＝1,...,N)。
3. The method as claimed in claim 1, wherein the said steps are performed in a manner that the power consumption data of small samples are consistent in frequency domain distributionIn S2, the daily electric load data frequency domain mapping quantity of the small sample electric user k is marked as Xk(s) (s ═ 1.., N), the daily electricity data frequency domain mapping quantity vector of the small sample electricity user k is identified as
Figure RE-FDA0003012767840000011
4. The method for expanding power consumption data of small sample users according to the frequency domain distribution consistency of claim 1, wherein in step S3, the daily power load data frequency domain mapping vector of K small sample power consumption users is used as data input, a gradient descent method is used, a countermeasure network is generated according to the probability distribution variance minimum criterion for data expansion training, and P λ · Z frequency domain user candidate data are generated by expansion, wherein the P (P ═ 1.., P) th user candidate daily frequency domain data vector is
Figure RE-FDA0003012767840000021
5. The method as claimed in claim 1, wherein in step S4, the p-th time-frequency domain data of the user to be selected generated by each expansion is processed according to a formula
Figure RE-FDA0003012767840000022
Calculating the sample accumulated cosine divergence Q (p) of the data to be selected; sequencing the day frequency domain data vectors of the users to be selected according to the value Q (p) from small to large; selecting the front Z user data to be selected as the power consumption daily frequency domain data of the extended user
Figure RE-FDA0003012767840000023
CN202010938107.6A 2020-09-09 2020-09-09 Small sample user electricity consumption data expansion method with frequency domain distribution consistency Pending CN112801123A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010938107.6A CN112801123A (en) 2020-09-09 2020-09-09 Small sample user electricity consumption data expansion method with frequency domain distribution consistency

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010938107.6A CN112801123A (en) 2020-09-09 2020-09-09 Small sample user electricity consumption data expansion method with frequency domain distribution consistency

Publications (1)

Publication Number Publication Date
CN112801123A true CN112801123A (en) 2021-05-14

Family

ID=75806777

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010938107.6A Pending CN112801123A (en) 2020-09-09 2020-09-09 Small sample user electricity consumption data expansion method with frequency domain distribution consistency

Country Status (1)

Country Link
CN (1) CN112801123A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113673579A (en) * 2021-07-27 2021-11-19 国网湖北省电力有限公司营销服务中心(计量中心) Power load classification algorithm based on small samples

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113673579A (en) * 2021-07-27 2021-11-19 国网湖北省电力有限公司营销服务中心(计量中心) Power load classification algorithm based on small samples
CN113673579B (en) * 2021-07-27 2024-05-28 国网湖北省电力有限公司营销服务中心(计量中心) Small sample-based electricity load classification algorithm

Similar Documents

Publication Publication Date Title
Wang et al. Electricity consumer characteristics identification: A federated learning approach
Welikala et al. Implementation of a robust real-time non-intrusive load monitoring solution
JP5888375B2 (en) Data processing apparatus, data processing method, and program
Liu et al. High-precision identification of power quality disturbances under strong noise environment based on FastICA and random forest
CN103065052B (en) A kind of method measuring the engineering goods theoretical life-span
Ni et al. A hybrid method for short-term sensor data forecasting in internet of things
Jiang et al. Literature review of power disaggregation
Tongta et al. Long short-term memory (LSTM) neural networks applied to energy disaggregation
Rizvi et al. Synchrophasor based ZIP parameters tracking using ML with adaptive window and data anomalies
CN112801123A (en) Small sample user electricity consumption data expansion method with frequency domain distribution consistency
Ganjavi et al. A study on probability of distribution loads based on expectation maximization algorithm
CN117972384A (en) Power load online identification method combining DTW with Bi-LSTM
CN116467631A (en) Power fingerprint identification model training method, power fingerprint identification method and device
Huang et al. Discriminative model for google host load prediction with rich feature set
Kumar et al. Low sampling rate electrical load disaggregation using dictionary representation and graph signal smoothness
Ming et al. Noise Level Estimation in Energy Internet Based on Artificial Neural Network
Fan et al. Performance prediction of network-intensive systems in cloud environment: a Bayesian approach
Guo et al. Unified Principal Components Analysis of Irregularly Observed Functional Time Series
Billio et al. Bayesian inference in dynamic models with latent factors
Yake et al. Adaptive VMD Parameter Optimization: A Method Based on Improved Genetic Algorithm with Temporal and Frequency Domain Characterization
Liu et al. Holistic data‐driven framework for estimating electromechanical dynamic patterns from synchrophasor measurements in bulk power grids
Vijverberg A time deformation model and its time-varying autocorrelation: An application to US unemployment data
Mao et al. Regression transfer learning based on principal curve
Lin et al. Short-Term Electricity Price Forecasting Based on Adaptive Hybrid Model
Alpcan Information metrics for model selection in function estimation

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