CN114548302B - A non-intrusive load cluster classification method and system - Google Patents
A non-intrusive load cluster classification method and system Download PDFInfo
- Publication number
- CN114548302B CN114548302B CN202210183179.3A CN202210183179A CN114548302B CN 114548302 B CN114548302 B CN 114548302B CN 202210183179 A CN202210183179 A CN 202210183179A CN 114548302 B CN114548302 B CN 114548302B
- Authority
- CN
- China
- Prior art keywords
- cluster
- gaussian process
- model
- double
- subset
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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 OR CALCULATING; 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
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S20/00—Management or operation of end-user stationary applications or the last stages of power distribution; Controlling, monitoring or operating thereof
- Y04S20/20—End-user application control systems
- Y04S20/222—Demand response systems, e.g. load shedding, peak shaving
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Computational Mathematics (AREA)
- Mathematical Physics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Probability & Statistics with Applications (AREA)
- Artificial Intelligence (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Evolutionary Computation (AREA)
- Pure & Applied Mathematics (AREA)
- Algebra (AREA)
- Operations Research (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
本发明涉及一种非侵入式负荷集群分类方法及系统，确定各训练子集对应的双层高斯过程混合模型；求解各训练子集对应的双层高斯过程混合模型的参数；根据M组训练子集对应的双层高斯过程混合模型的参数求取均值，确定双层高斯过程混合平均模型的参数；将各测试子集中各类集群代入所述双层高斯过程混合平均模型进行计算，输出各测试子集中各类集群的概率值；将各测试子集中各类集群输出概率最大值对应的集群类别作为各测试子集对应的分类结果。本发明方案将用户侧具有相同操作特点、统一用电规律的负荷建模成一个集群，利用隶属于双层高斯过程混合平均模型的概率值来判定未知集群的所属类别，从而有效实现集群的分类，为用户提高用电效率提供指导。
The present invention relates to a non-intrusive load cluster classification method and system, which determines a double-layer Gaussian process mixture model corresponding to each training subset; solves the parameters of the double-layer Gaussian process mixture model corresponding to each training subset; takes the mean value according to the parameters of the double-layer Gaussian process mixture model corresponding to M groups of training subsets, and determines the parameters of the double-layer Gaussian process mixture average model; substitutes each type of cluster in each test subset into the double-layer Gaussian process mixture average model for calculation, and outputs the probability value of each type of cluster in each test subset; and uses the cluster category corresponding to the maximum output probability of each type of cluster in each test subset as the classification result corresponding to each test subset. The scheme of the present invention models the loads with the same operating characteristics and unified power consumption rules on the user side into a cluster, and uses the probability value belonging to the double-layer Gaussian process mixture average model to determine the category to which the unknown cluster belongs, thereby effectively realizing the classification of the cluster and providing guidance for users to improve power consumption efficiency.
Description
Technical Field
The invention relates to the technical field of load monitoring, in particular to a non-invasive load cluster classification method and system.
Background
Along with the continuous increase of the types and the quantity of electric equipment, the composition structure, the load characteristics and the like of the load side are also more complex, and the detailed analysis of the user energy information provides basic information reference for developing the management of the demand side.
Under some electricity utilization scenes (such as park electricity utilization scenes), a large amount of load often appears in the form of load clusters, the running state of the whole power system is directly influenced by the behavior of the load clusters, unified planning and running control of a power distribution network and a distributed power supply by taking the load clusters as units become a development trend, coordination among clusters can be realized through load cluster management, and source-load matching is realized in the clusters.
At present, the research on load clusters is mainly focused on the fields of distributed power supply clusters, power distribution network clusters, elastic adjustable load clusters and the like, and the division index and the operation control of the clusters are mainly researched. There are relatively few load cluster classification studies with uniform electricity usage laws with the same operating characteristics.
Disclosure of Invention
The invention aims to provide a non-invasive load cluster classification method and a non-invasive load cluster classification system, so as to classify load clusters with the same operation characteristics and unified electricity utilization rule.
To achieve the above object, the present invention provides a non-invasive load cluster classification method, the method comprising:
constructing a test set T and a training set S based on an electricity utilization data set D corresponding to a load cluster, wherein the training set S comprises M training subsets, and the test set T comprises M test subsets, wherein M is a positive integer greater than 1;
Determining a double-layer Gaussian process mixed model corresponding to each training subset, wherein parameters of the double-layer Gaussian process mixed model are to be solved;
Solving parameters of a double-layer Gaussian process mixed model corresponding to each training subset;
Solving an average value of the M groups of parameters obtained by solving, and substituting the average value into a double-layer Gaussian process mixed model to obtain the double-layer Gaussian process mixed average model;
substituting each cluster in each test subset into the double-layer Gaussian process mixed average model for calculation, and outputting probability values of each cluster in each test subset;
And taking the cluster category corresponding to the maximum value of the output probability of each cluster in each test subset as the classification result corresponding to each test subset.
Optionally, the determining the two-layer gaussian process mixing model corresponding to each training subset specifically includes:
Determining the total class number K of clusters contained in each training subset;
and determining a double-layer Gaussian process mixing model corresponding to each training subset based on the input feature sets of all clusters in each training subset.
Optionally, the determining the total class number K of the clusters included in each training subset specifically includes:
Step S211, determining load cluster characteristics based on characteristics corresponding to various clusters;
Step S212, judging whether clusters meeting the load cluster characteristics exist in the training subset S l, if so, executing a step S214, and if not, rejecting the training subset S l, enabling l=l+1, and executing a step S213;
step S213, judging whether l is larger than M, if l is larger than M, ending, and if l is smaller than or equal to M, returning to step S212;
Step S214, determining the total class number K of the clusters and the input feature set of each cluster in each training subset according to the load cluster features.
Optionally, the determining the two-layer gaussian process mixing model corresponding to each training subset based on the input feature set of each cluster in each training subset specifically includes:
s221, constructing a Gaussian process model corresponding to each input feature based on each input feature in an ith input feature set, wherein the input feature set comprises p input features;
step S222, constructing a Gaussian process mixture model corresponding to an ith cluster according to the Gaussian process models corresponding to p input features, wherein the Gaussian process mixture model is a lower model;
Step S223, judging whether i is smaller than K, if i is smaller than K, letting i=i+1, and executing step S221, if i is larger than or equal to K, constructing an upper model according to a Gaussian process mixture model corresponding to the K-class cluster;
and step 224, determining a double-layer Gaussian process mixing model based on the upper-layer model and the lower-layer model corresponding to various clusters.
Optionally, the load cluster comprises at least one of a computer cluster, an air conditioner cluster, a water heater cluster and an electric automobile cluster.
The invention also provides a non-invasive load cluster classification system, comprising:
The system comprises a data set construction module, a load cluster and a training set, wherein the data set construction module is used for constructing a test set T and the training set S based on an electricity data set D corresponding to the load cluster, and the test set T comprises M test subsets, wherein M is a positive integer greater than 1;
The double-layer Gaussian process mixed model determining module is used for determining a double-layer Gaussian process mixed model corresponding to each training subset, wherein parameters of the double-layer Gaussian process mixed model are to be solved;
The parameter solving module is used for solving parameters of the double-layer Gaussian process mixed model corresponding to each training subset;
the double-layer Gaussian process mixed average model determining module is used for solving M groups of parameters obtained by solving to obtain a mean value, and substituting the mean value into the double-layer Gaussian process mixed model to obtain the double-layer Gaussian process mixed average model;
the probability value calculation module is used for substituting various clusters in each test subset into the double-layer Gaussian process mixed average model to calculate and outputting probability values of various clusters in each test subset;
and the classification result determining module is used for taking the cluster category corresponding to the maximum value of the output probability of each cluster in each test subset as the classification result corresponding to each test subset.
Optionally, the dual-layer gaussian process mixing model determining module specifically includes:
the total class number determining unit is used for determining the total class number K of the clusters contained in each training subset;
and the double-layer Gaussian process mixed model determining unit is used for determining the double-layer Gaussian process mixed model corresponding to each training subset based on the input characteristic sets of all clusters in each training subset.
Optionally, the total class number determining unit specifically includes:
The load cluster feature determining subunit is used for determining load cluster features based on features corresponding to various clusters;
A first judging subunit, configured to judge whether a cluster satisfying the load cluster feature exists in the training subset S l, execute the "total class number and input feature set determining subunit" if a cluster satisfying the load cluster feature exists, reject the training subset S l if a cluster satisfying the load cluster feature does not exist, and let l=l+1, and execute the "second judging subunit";
the second judging subunit is used for judging whether l is larger than M, ending if l is larger than M, and returning to the first judging subunit if l is smaller than or equal to M;
and the total class number and input feature set determining subunit is used for determining the total class number K of the clusters and the input feature set of each cluster contained in each training subset according to the load cluster features.
Optionally, the dual-layer gaussian process mixing model determining unit specifically includes:
the Gaussian process model construction subunit is used for constructing a Gaussian process model corresponding to each input feature based on each input feature in the ith input feature set, wherein the input feature set comprises p input features;
The lower layer model construction subunit is used for constructing a Gaussian process mixed model corresponding to the ith class cluster according to the Gaussian process models corresponding to the p input features, wherein the Gaussian process mixed model is a lower layer model;
The third judging subunit is used for judging whether i is smaller than K, if i is smaller than K, making i=i+1, and executing a Gaussian process model construction subunit;
And the double-layer Gaussian process mixing model determining subunit is used for determining a double-layer Gaussian process mixing model based on the upper-layer model and the lower-layer model corresponding to various clusters.
Optionally, the load cluster comprises at least one of a computer cluster, an air conditioner cluster, a water heater cluster and an electric automobile cluster.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
According to the scheme, the load with the same operation characteristics and the unified electricity utilization rule on the user side is modeled into one cluster, the probability value of the mixed average model belonging to the double-layer Gaussian process is utilized to judge the category of the unknown cluster in the test subset, so that the classification of the cluster is effectively realized, guidance is provided for the user to improve the electricity utilization efficiency, the method has important significance for safe, reliable, stable and economic operation of an electric power system, and the method is favorable for promoting the realization of comprehensive electricity utilization intellectualization.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions of the prior art, the drawings that are needed in the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a non-intrusive load cluster classification method of the present invention;
FIG. 2 is a block diagram of a non-intrusive load cluster classification system of the present invention;
FIG. 3 is a schematic diagram of a non-intrusive load cluster classification system architecture according to the present invention;
FIG. 4 is a graph of a total current waveform obtained by a non-invasive acquisition device in accordance with the present invention;
FIG. 5 is a graph showing current waveforms of a computer cluster according to the present invention;
Fig. 6 is a current waveform of an air conditioning cluster according to the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The invention aims to provide a non-invasive load cluster classification method and a non-invasive load cluster classification system, so as to classify load clusters with the same operation characteristics and unified electricity utilization rule.
In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to the appended drawings and appended detailed description.
Example 1
As shown in fig. 1, the present invention discloses a non-invasive load cluster classification method, which includes:
The method comprises the step S1 of constructing a test set T and a training set S based on an electricity utilization data set D corresponding to a load cluster, wherein the training set S comprises M training subsets, the test set T comprises M test subsets, and M is a positive integer greater than 1.
And S2, determining a double-layer Gaussian process mixed model corresponding to each training subset, wherein parameters of the double-layer Gaussian process mixed model are to be solved.
Step S3, solving parameters of the double-layer Gaussian process mixed model corresponding to each training subset
And S4, calculating the mean value of the M groups of parameters obtained by solving, and substituting the mean value into the double-layer Gaussian process mixed model to obtain the double-layer Gaussian process mixed average model.
And S5, substituting each cluster in each test subset into the double-layer Gaussian process mixed average model for calculation, and outputting probability values of each cluster in each test subset.
And S6, taking the cluster category corresponding to the maximum value of the output probability of each cluster in each test subset as the classification result corresponding to each test subset.
The steps are discussed in detail below:
Step S1, a test set T and a training set S are constructed based on an electricity utilization data set D corresponding to a load cluster, and the method specifically comprises the following steps:
And step S11, collecting electricity utilization data at the power inlet of the user by using a non-invasive collecting device to construct an electricity utilization data set D, and specifically, collecting total current data or total power data at the power inlet of the user by using the non-invasive collecting device in a high-frequency collecting mode.
The method comprises the steps of S12, dividing an electricity data set D into a test set T and a training set S by adopting an M-fold cross validation method, specifically dividing the electricity data set D into M subsets, namely D= [ D 1,D2,…,DM ], taking a first subset D l (l=1, 2, the..M) as the test subset T l each time, taking the remaining M-l subsets as the training subset S l, constructing the test set T on the basis of M groups of test subsets T l, namely T= [ T 1,T2,…,TM ], and constructing the training set S on the basis of M groups of training subsets S l, namely S= [ S 1,S2,…,SM ]. In this embodiment, M is selected arbitrarily according to actual requirements, and preferably, M is 10.
Step S2, determining a double-layer Gaussian process mixing model corresponding to each training subset, wherein the method specifically comprises the following steps:
step S21, determining total class number K of clusters contained in each training subset, wherein the step S21 specifically comprises the following steps:
Step S211, determining load cluster characteristics based on characteristics corresponding to various clusters, wherein the characteristics corresponding to various clusters comprise working characteristics, operation characteristics, random characteristics and the like. The load cluster comprises at least one of a computer cluster, an air conditioner cluster, a water heater cluster and an electric automobile cluster.
Step S212, judging whether a cluster meeting the load cluster characteristics exists in the training subset S l, if so, executing a step S214, and if not, rejecting the training subset S l, and letting l=l+1, and executing a step S213.
Step S213, judging whether l is larger than M, if l is larger than M, ending, and if l is smaller than or equal to M, returning to step S212.
Step S214, determining the total class number K of the clusters and the input feature set of each cluster in each training subset according to the load cluster features, wherein the input feature set comprises p input features.
Step S22, determining a double-layer Gaussian process mixing model corresponding to each training subset based on input feature sets of various clusters in each training subset, wherein the method specifically comprises the following steps:
step S221, constructing a Gaussian process model corresponding to each input feature based on each input feature in the ith input feature set, wherein the specific formula is as follows:
yig(x)～GPFR(xg;big,θig);
where x g represents the g-th input feature in the input feature set x 'k of the i-th class of cluster, g=1,..p, x' k＝[x1,x2,…,xp],big,θig represents the mean and standard deviation of the g-th input feature in the input feature set x 'k of the i-th class cluster, GPFR () represents a gaussian process function and y ig (x) represents a gaussian process model corresponding to the g-th input feature in the input feature set x' k of the i-th cluster.
Step S222, constructing a Gaussian process mixed model corresponding to the ith class of clusters according to the Gaussian process models corresponding to the p input features, namely a lower model corresponding to the ith class of clusters.
Step S223, judging whether i is smaller than K, if i is smaller than K, letting i=i+1, executing step S221, and if i is larger than or equal to K, constructing an upper layer model according to the Gaussian process mixture model corresponding to the K-class cluster.
Step 224, determining a double-layer gaussian process mixing model based on the upper layer model and the lower layer model corresponding to various clusters, which specifically comprises the following steps:
defining an indicator variable (i.e., hidden variable) in the upper model Describing the relevance of the s-th two-layer Gaussian process mixture model relative to the underlying Gaussian process model, if the i-th cluster belongs to the s-th Gaussian process mixture model, then there isOtherwiseAt the same timeObeys the following probability distribution:
Wherein, Pi s is an indicator variable in the upper modelProbability of time.
In the lower layer model, the relevance of the s-th two-layer Gaussian process mixture model relative to the lower layer Gaussian process model can use another hidden variableDescribing that if the g input feature of the i-th class cluster belongs to the q-th Gaussian process model, and the i-th class load cluster belongs to the s-th double-layer Gaussian process mixed model, the method comprises the following steps ofOtherwiseAt the same timeObeys the following probability distribution:
Wherein, Eta q|s is an indicator variable in the lower modelProbability at time, n is the total number of gaussian process models.
Is provided withAnd assuming that the input variables obey gaussian distribution at each group, namely:
Where x ig is the input variable, h qs, Representing mean and covariance, respectively.
The specific formula for defining the double-layer Gaussian process mixture model is as follows:
Θ={(πs,ηq|s,Θqs)|s=1,...,K;q=1,...,n}
Wherein Θ qs＝{hqs,sqs,bqs,θqs},hqs,sqs is the mean and standard deviation of the upper model, b qs,θqs is the mean and standard deviation of the lower model, η q|s is the indicated variable in the lower model Probability at time, pi s is the indicator variable in the upper modelThe probability of time, n is the total number of Gaussian process models, and K is the total number of clusters contained in the load clusters.
Step S3, solving parameters of the double-layer Gaussian process mixed model corresponding to each training subsetWherein, H qs,sqs is the mean and standard deviation of the upper model, b qs,θqs is the mean and standard deviation of the lower model, η q|s is the indicated variable in the lower modelProbability at time, pi s is the indicator variable in the upper modelProbability of time.
The step S3 specifically comprises the following steps:
step S31, initializing parameters of the double-layer Gaussian process mixture model
Step S32, adopting a modified MCMC algorithm according to the following stepsAnd calculating all hidden variables Z in the upper model and all hidden variables A in the lower model.
And step S33, establishing likelihood functions of the load cluster double-layer Gaussian process mixed model according to all hidden variables Z in the upper-layer model and all hidden variables A in the lower-layer model.
Step S34, establishing the expectation function of the conditional probability distribution of the unknown data according to the likelihood functionThe following formula is calculated:
Wherein I represents the number of hidden variables Z, Z (i) represents the ith MCMC sample of Z, D represents a sample of the sample set, E A [ ] represents the desired computation function, L (Θ, Z (I), A) represents the likelihood function, A desired function representing a conditional probability distribution of unknown data.
Step S35 based on maximizationAnd update
Step S36, judging whether the convergence condition is met, if so, outputting parameters of the double-layer Gaussian process mixed model corresponding to the M groups of training subsetsIf the convergence condition is not satisfied, return to "step S32". The convergence condition in this embodiment includes but is not limited toNo further changes occur.
S5, substituting each cluster in each test subset into the double-layer Gaussian process mixed average model for calculation, and outputting probability values of each cluster in each test subset, wherein a specific calculation formula is as follows:
Wherein p (z|x) represents the probability that the z-th cluster in the test subset T l belongs to the double-layer gaussian process mixed model x, p (x) represents the probability that the x-th cluster in the double-layer gaussian process mixed average model appears, and p (zx) represents the probability that the z-th cluster in the test subset T l exists in the x-th cluster in the double-layer gaussian process mixed average model.
Example 2
As shown in fig. 2-3, the present invention discloses a non-invasive load cluster classification system, the system comprising:
The data set construction module 201 is configured to construct a test set T and a training set S based on an electricity data set D corresponding to the load cluster, where the training set S includes M training subsets, and the test set T includes M test subsets, where M is a positive integer greater than 1.
The two-layer gaussian process mixing model determining module 202 is configured to determine two-layer gaussian process mixing models corresponding to the training subsets, where parameters of the two-layer gaussian process mixing models are to be solved.
And the parameter solving module 203 is configured to solve parameters of the two-layer gaussian process mixing model corresponding to each training subset.
The two-layer gaussian process mixed average model determining module 204 is configured to determine an average value of M groups of parameters obtained by the solution, and obtain a two-layer gaussian process mixed average model by substituting the average value into the two-layer gaussian process mixed model.
The probability value calculation module 205 is configured to substitute each cluster in each test subset into the double-layer gaussian process hybrid average model for calculation, and output probability values of each cluster in each test subset.
The classification result determining module 206 is configured to use the cluster category corresponding to the maximum value of the output probabilities of the clusters in each test subset as the classification result corresponding to each test subset.
As an optional implementation manner, the dual-layer gaussian process mixing model determining module 202 of the present invention specifically includes:
and the total class number determining unit is used for determining the total class number K of the clusters contained in each training subset.
And the double-layer Gaussian process mixed model determining unit is used for determining the double-layer Gaussian process mixed model corresponding to each training subset based on the input characteristic sets of all clusters in each training subset.
As an optional implementation manner, the total class number determining unit of the present invention specifically includes:
And the load cluster characteristic determining subunit is used for determining the load cluster characteristic based on the characteristics corresponding to various clusters.
A first judging subunit, configured to judge whether a cluster satisfying the load cluster feature exists in the training subset S l, execute the "total class number and input feature set determining subunit" if a cluster satisfying the load cluster feature exists, reject the training subset S l if a cluster satisfying the load cluster feature does not exist, and let l=l+1, and execute the "second judging subunit".
And the second judging subunit is used for judging whether l is larger than M, ending if l is larger than M, and returning to the first judging subunit if l is smaller than or equal to M.
And the total class number and input feature set determining subunit is used for determining the total class number K of the clusters and the input feature set of each cluster contained in each training subset according to the load cluster features.
As an optional implementation manner, the dual-layer gaussian process mixing model determining unit of the present invention specifically includes:
and the Gaussian process model construction subunit is used for constructing a Gaussian process model corresponding to each input feature based on each input feature in the ith input feature set, wherein the input feature set comprises p input features.
The lower layer model construction subunit is used for constructing a Gaussian process mixed model corresponding to the ith class of clusters according to the Gaussian process models corresponding to the p input features, wherein the Gaussian process mixed model is a lower layer model.
And the third judging subunit is used for judging whether i is smaller than K, if i is smaller than K, letting i=i+1, executing a Gaussian process model construction subunit, and if i is larger than or equal to K, constructing an upper layer model according to a Gaussian process mixed model corresponding to the K-class cluster.
And the double-layer Gaussian process mixing model determining subunit is used for determining a double-layer Gaussian process mixing model based on the upper-layer model and the lower-layer model corresponding to various clusters.
Example 3
A non-invasive acquisition device is arranged at an electric inlet of a small office, voltage waveform and current waveform signal data are acquired, the sampling frequency is 5000Hz, and the acquisition duration is one week.
The method specifically comprises the following steps:
1) And a non-invasive acquisition device is arranged at the electric inlet of the small office, voltage waveform and current waveform signal data are acquired, and as shown in fig. 4, the electric data are divided into a test set and a training set by adopting a ten-fold cross validation method.
2) And detecting whether clusters meeting the load cluster characteristics exist in the training subset S l, and determining an input characteristic x and an output characteristic y for classifying the load clusters.
3) And establishing a Gaussian process mixture model for the ith cluster (such as a computer cluster) in the training set S l to form a lower model of the double-layer Gaussian process mixture model.
4) And establishing K Gaussian process mixed models for K clusters (such as a computer cluster and an air conditioner cluster) to form an upper model of the double-layer Gaussian process mixed model.
5) And respectively defining the indicating variables (hidden variables) of an upper layer model and a lower layer model in the load cluster double-layer Gaussian process mixed model to form the load cluster double-layer Gaussian process mixed model.
6) Solving the parameters of the load cluster double-layer Gaussian process mixed model.
7) Repeating the steps 2) to 6) to obtain 10 groups of parameters of the load cluster double-layer Gaussian process mixed model, and obtaining an average value to finally determine parameter values.
8) And calculating the probability that various load clusters in the test set T l are affiliated to a double-layer Gaussian process mixed average model, judging the class of the clusters according to the maximum value of the probability, wherein the current waveforms of the computer clusters and the air conditioner clusters are shown in fig. 5 and 6 respectively.
In the present specification, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different point from other embodiments, and identical and similar parts between the embodiments are all enough to refer to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant points refer to the description of the method section.
The principles and embodiments of the present invention have been described herein with reference to specific examples, which are intended to facilitate an understanding of the principles and concepts of the invention and are to be varied in scope and detail by persons of ordinary skill in the art based on the teachings herein. In view of the foregoing, this description should not be construed as limiting the invention.
Claims (4)
1. A method of non-intrusive load cluster classification, the method comprising:
constructing a test set T and a training set S based on an electricity utilization data set D corresponding to a load cluster, wherein the training set S comprises M training subsets, and the test set T comprises M test subsets, wherein M is a positive integer greater than 1;
The method specifically comprises the steps of determining a double-layer Gaussian process mixture model corresponding to each training subset, wherein the method specifically comprises the steps of determining a double-layer Gaussian process mixture model corresponding to each training subset based on an input feature set of each cluster in each training subset, determining the double-layer Gaussian process mixture model corresponding to each training subset, wherein the method specifically comprises the steps of determining a load cluster feature based on features corresponding to each cluster in each training subset, determining whether clusters meeting the load cluster feature exist in the training subset S l or not, executing a step S214 if clusters meeting the load cluster feature exist, eliminating the training subset S l and enabling l=l+1 if clusters meeting the load cluster feature do not exist, executing a step S213, determining whether l is larger than M, ending if l is smaller than or equal to M, returning a step S212 if l is smaller than or equal to M, determining whether the total class K of clusters contained in each training subset and the input feature set corresponding to each cluster in each training subset according to the feature set in step S l according to the feature, constructing a step i=p, inputting a step i-type Gaussian process mixture model corresponding to each training subset based on the corresponding to each cluster in each training subset is determined to be the step S221, inputting a step i=1, and inputting a step i+1 to a high-class model corresponding to each training subset is input to the step i, and inputting a step i+1, and determining a step i to be a high-class model according to the step I, and inputting a high-class model, if i is greater than 1, step S224, determining a double-layer Gaussian process mixed model based on the upper layer model and the lower layer model corresponding to various clusters, wherein parameters of the double-layer Gaussian process mixed model are to be solved;
Solving parameters of a double-layer Gaussian process mixed model corresponding to each training subset;
Solving an average value of the M groups of parameters obtained by solving, and substituting the average value into a double-layer Gaussian process mixed model to obtain the double-layer Gaussian process mixed average model;
substituting each cluster in each test subset into the double-layer Gaussian process mixed average model for calculation, and outputting probability values of each cluster in each test subset;
And taking the cluster category corresponding to the maximum value of the output probability of each cluster in each test subset as the classification result corresponding to each test subset.
2. The method of claim 1, wherein the load clusters comprise at least one of a computer cluster, an air conditioning cluster, a water heater cluster, and an electric vehicle cluster.
3. A non-intrusive load cluster classification system, the system comprising:
The system comprises a data set construction module, a load cluster and a training set, wherein the data set construction module is used for constructing a test set T and the training set S based on an electricity data set D corresponding to the load cluster, and the test set T comprises M test subsets, wherein M is a positive integer greater than 1;
The double-layer Gaussian process mixed model determining module is used for determining a double-layer Gaussian process mixed model corresponding to each training subset; the double-layer Gaussian process mixed model determining module specifically comprises a total class number determining unit, a double-layer Gaussian process mixed model determining unit, a second judging subunit and a first judging subunit, wherein the total class number determining unit is used for determining the total class number K of clusters contained in each training subset, the double-layer Gaussian process mixed model determining unit is used for determining a double-layer Gaussian process mixed model corresponding to each training subset based on an input characteristic set of each cluster in each training subset, the total class number determining unit specifically comprises a load cluster characteristic determining subunit used for determining the load cluster characteristic based on the characteristics corresponding to each cluster, the first judging subunit is used for judging whether clusters meeting the load cluster characteristic exist in the training subset S l, if clusters meeting the load cluster characteristic exist, the total class number and the input characteristic set determining subunit are executed, if clusters meeting the load cluster characteristic do not exist, the training subset S l is eliminated, l=l+1 is made, the second judging subunit is executed, the second judging subunit is used for judging whether l is larger than M, if l is smaller than or equal to M, the first judging subunit is returned, the total class number determining subunit is used for judging whether the total class number K meets the load cluster characteristic or not exists in each training subset, if clusters meeting the load cluster characteristic exists in each training subset S l, the total class number and the double-layer Gaussian process mixed model is used for constructing the input model corresponding to each training subset is input to the total class characteristic, the input characteristic is input to the input model, the total class number is input to the total class cluster is input to the training subset, and the total class characteristic is input to the total class model is input to the high, and the total class model is input to the total is 1 is input to the high, and is 1, the system comprises a Gaussian process mixing model, a third judging subunit, a double-layer Gaussian process mixing model determining subunit, a parameter solving unit and a model processing unit, wherein the Gaussian process mixing model is used for constructing a Gaussian process mixing model corresponding to an i-th cluster according to Gaussian process models corresponding to p input features, the Gaussian process mixing model is a lower-layer model, the third judging subunit is used for judging whether i is smaller than K, if i is smaller than K, the i=i+1 and executing a Gaussian process model constructing subunit, and if i is larger than or equal to K, the upper-layer model is constructed according to the Gaussian process mixing model corresponding to the K-th cluster;
The parameter solving module is used for solving parameters of the double-layer Gaussian process mixed model corresponding to each training subset;
the double-layer Gaussian process mixed average model determining module is used for solving M groups of parameters obtained by solving to obtain a mean value, and substituting the mean value into the double-layer Gaussian process mixed model to obtain the double-layer Gaussian process mixed average model;
the probability value calculation module is used for substituting various clusters in each test subset into the double-layer Gaussian process mixed average model to calculate and outputting probability values of various clusters in each test subset;
and the classification result determining module is used for taking the cluster category corresponding to the maximum value of the output probability of each cluster in each test subset as the classification result corresponding to each test subset.
4. The non-invasive load cluster classification system of claim 3, wherein the load clusters comprise at least one of a computer cluster, an air conditioning cluster, a water heater cluster, and an electric vehicle cluster.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210183179.3A CN114548302B (en) | 2022-02-28 | 2022-02-28 | A non-intrusive load cluster classification method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210183179.3A CN114548302B (en) | 2022-02-28 | 2022-02-28 | A non-intrusive load cluster classification method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114548302A CN114548302A (en) | 2022-05-27 |
| CN114548302B true CN114548302B (en) | 2025-05-30 |
Family
ID=81680036
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210183179.3A Active CN114548302B (en) | 2022-02-28 | 2022-02-28 | A non-intrusive load cluster classification method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114548302B (en) |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7697752B2 (en) * | 2005-12-20 | 2010-04-13 | General Instrument Corporation | Method and apparatus for performing object detection |
| US9275345B1 (en) * | 2011-02-11 | 2016-03-01 | Allure Security Technology, Inc. | System level user behavior biometrics using feature extraction and modeling |
| CN109978201A (en) * | 2017-12-27 | 2019-07-05 | 深圳市景程信息科技有限公司 | Probability load prediction system and method based on Gaussian process quantile estimate model |
| CN108921358B (en) * | 2018-07-16 | 2021-10-01 | 广东工业大学 | Prediction method, prediction system and related device for power load characteristics |
| CN109492093A (en) * | 2018-09-30 | 2019-03-19 | 平安科技（深圳）有限公司 | File classification method and electronic device based on gauss hybrid models and EM algorithm |
| CN109638830B (en) * | 2019-01-18 | 2022-03-22 | 广东电网有限责任公司 | Power load model construction method, device and equipment |
| CN110266010B (en) * | 2019-05-30 | 2023-01-17 | 武汉大学 | Multi-objective collaborative optimization method for grid structure of DC distribution network and distributed photovoltaic |
-
2022
- 2022-02-28 CN CN202210183179.3A patent/CN114548302B/en active Active
Non-Patent Citations (1)
| Title |
|---|
| 非侵入式负荷集群感知方法及影响评估研究;严萌;《中国优秀硕士学位论文全文数据库 (工程科技Ⅱ辑)》;20240415(第4期);C042-1334 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114548302A (en) | 2022-05-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN118484666B (en) | Energy storage power station evaluation method and system for source network load multi-element application | |
| CN111368904B (en) | Electrical equipment identification method based on electric power fingerprint | |
| CN114139940A (en) | A method for evaluating the interaction level of generalized demand-side resource network-load based on combined weighting-cloud model | |
| CN118868422B (en) | Dynamic load intelligent monitoring method and system for digital circuit of distribution network | |
| CN107274079A (en) | The division of the power supply area method clustered based on improved grey model | |
| CN112712281A (en) | Cloud model-based energy storage working condition adaptability comprehensive evaluation method and system | |
| CN113361776A (en) | Power load probability prediction method based on user power consumption behavior clustering | |
| CN112990627A (en) | Electric energy quality evaluation method | |
| CN115952404A (en) | Hydroelectric generating set fault early warning method, device and terminal | |
| CN103544546A (en) | Method for online prediction of quiescent voltage stability limit of electric system | |
| CN114358490A (en) | Power demand response-oriented multi-adjustable-resource adjusting capacity aggregation method and system | |
| CN115392133A (en) | An Equivalence Method for Optimal Clustering of Wind Farms Using Gaussian Mixture Models | |
| Justo et al. | Behavioral similarity of residential customers using a neural network based on adaptive resonance theory | |
| CN108898273B (en) | A clustering evaluation method of user-side load characteristics based on morphological analysis | |
| CN116914751A (en) | Intelligent power distribution control system | |
| CN103559556B (en) | A kind of method of on-line prediction power system load capacity limit | |
| CN107194526A (en) | A kind of sales marketization reform progress appraisal procedure based on fuzzy clustering | |
| Baherifard et al. | Improving the Effect of Electric Vehicle Charging on Imbalance Index in the Unbalanced Distribution Network Using Demand Response Considering Data Mining Techniques | |
| CN115130830B (en) | Non-intrusive load decomposition method based on cascade width learning and sparrow algorithm | |
| CN115828755A (en) | Evaluation method, device and readable storage medium for microgrid group participating in power grid service | |
| CN114548302B (en) | A non-intrusive load cluster classification method and system | |
| CN119416099A (en) | A method, system, device and storage medium for predicting line loss in an active distribution station area | |
| CN109615142A (en) | A combined prediction method of wind speed in wind farm based on wavelet analysis | |
| CN118485179A (en) | Low-voltage transformer area line loss index differentiation prediction method and system | |
| CN115622034B (en) | A method for identifying power loads in IoT meter based on improved genetic neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |