A non-intrusive load clustering classification method and system

Abstract

The present invention relates to a non-invasive load clustering classification method and system, which determines a two-layer Gaussian process mixture model corresponding to each training subset; solves the parameters of the two-layer Gaussian process mixture model corresponding to each training subset; obtains the mean value according to the parameters of the two-layer Gaussian process mixture model corresponding to M groups of training subsets, and determines the parameters of the two-layer Gaussian process mixture average model; substitutes various clusters in each test subset into the two-layer Gaussian process mixture average model for calculation, and outputs the probability values of various clusters in each test subset; and takes the cluster category corresponding to the maximum output probability value of various clusters in each test subset as the classification result corresponding to each test subset. The solution of the present invention models the loads with the same operation characteristics and unified power consumption rules on the user side into a cluster, and uses the probability value belonging to the two-layer Gaussian process mixture average model to determine the category of the unknown cluster, thereby effectively realizing the classification of the cluster and providing guidance for users to improve the power consumption efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of load monitoring, and particularly to a non-invasive load cluster classification method and system. Background Art

[0002] With the continuous increase in the types and quantities of electrical equipment, the composition structure and load characteristics on the load side have become increasingly complex. The detailed analysis of user energy consumption information will provide basic information reference for the implementation of demand-side management.

[0003] In some power consumption scenarios (such as campus power consumption scenarios), a large number of loads often appear in the form of load clusters. The behavior of load clusters directly affects the operating state of the entire power system. It has become a development trend to conduct unified planning and operation control of distribution networks and distributed power sources with load clusters as units. Through load cluster management, coordination and cooperation can be achieved among clusters, and source-load matching can be achieved within clusters.

[0004] Currently, the research on load clusters mainly focuses on fields such as distributed power source clusters, distribution network clusters, and flexible adjustable load clusters, and mainly studies the division indicators and operation control of clusters. There is relatively little research on the classification of load clusters with the same operation characteristics and unified power consumption rules. Summary of the Invention

[0005] The object of the present invention is to provide a non-invasive load cluster classification method and system to classify load clusters with the same operation characteristics and unified power consumption rules.

[0006] To achieve the above object, the present invention provides a non-invasive load cluster classification method, and the method includes:

[0007] Constructing a test set T and a training set S based on the power consumption data set D corresponding to the load cluster; the training set S includes M training subsets; the test set T includes M test subsets; where M is a positive integer greater than 1;

[0008] Determining the double-layer Gaussian process mixture model corresponding to each training subset; the parameters of the double-layer Gaussian process mixture model are to be solved;

[0009] Solving the parameters of the double-layer Gaussian process mixture model corresponding to each training subset;

[0010] Calculating the mean value of the M groups of solved parameters, and substituting the mean value into the double-layer Gaussian process mixture model to obtain a double-layer Gaussian process mixture average model;

[0011] Substituting various clusters in each test subset into the double-layer Gaussian process mixture average model for calculation, and outputting the probability values of various clusters in each test subset;

[0012] Take 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.

[0013] Optionally, the determination of the double-layer Gaussian process mixture model corresponding to each training subset specifically includes:

[0014] Determine the total number of classes K of clusters included in each training subset;

[0015] Based on the input feature sets of each type of cluster in each training subset, determine the double-layer Gaussian process mixture model corresponding to each training subset.

[0016] Optionally, the determination of the total number of classes K of clusters included in each training subset specifically includes:

[0017] Step S211: Determine the load cluster features based on the features corresponding to each type of cluster;

[0018] Step S212: Judge whether there is a cluster in the training subset S l that satisfies the load cluster features; if there is a cluster that satisfies the load cluster features, execute "Step S214"; if there is no cluster that satisfies the load cluster features, remove the training subset S l , and let l = l + 1, and execute "Step S213";

[0019] Step S213: Judge whether l is greater than M. If l is greater than M, end; if l is less than or equal to M, return to "Step S212";

[0020] Step S214: Determine the total number of classes K of clusters included in each training subset and the input feature sets of each type of cluster according to the load cluster features.

[0021] Optionally, the determination of the double-layer Gaussian process mixture model corresponding to each training subset based on the input feature sets of each type of cluster in each training subset specifically includes:

[0022] Step S221: Construct a Gaussian process model corresponding to each input feature based on each input feature in the i-th input feature set; the input feature set includes p input features;

[0023] Step S222: Construct a Gaussian process mixture model corresponding to the i-th type of cluster according to the Gaussian process models corresponding to p input features; the Gaussian process mixture model is the lower-layer model;

[0024] Step S223: Judge whether i is less than K. If i is less than K, let i = i + 1, and execute "Step S221"; if i is greater than or equal to K, construct the upper-layer model according to the Gaussian process mixture models corresponding to K types of clusters;

[0025] Step S224: Determine a two - layer Gaussian process mixture model based on the upper - layer model and the lower - layer models corresponding to various clusters.

[0026] Optionally, the load clusters include at least one of: computer clusters, air - conditioner clusters, water - heater clusters, and electric - vehicle clusters.

[0027] The present invention also provides a non - intrusive load cluster classification system, which includes:

[0028] A dataset construction module, configured to construct a test set T and a training set S based on the power consumption dataset D corresponding to the load clusters; the test set T includes M test subsets; where M is a positive integer greater than 1.

[0029] A two - layer Gaussian process mixture model determination module, configured to determine the two - layer Gaussian process mixture models corresponding to each training subset; the parameters of the two - layer Gaussian process mixture model are to be solved.

[0030] A parameter solving module, configured to solve the parameters of the two - layer Gaussian process mixture models corresponding to each training subset.

[0031] A two - layer Gaussian process mixture average model determination module, configured to calculate the mean of the M groups of solved parameters, and substitute the mean into the two - layer Gaussian process mixture model to obtain a two - layer Gaussian process mixture average model.

[0032] A probability value calculation module, configured to substitute various clusters in each test subset into the two - layer Gaussian process mixture average model for calculation, and output the probability values of various clusters in each test subset.

[0033] A classification result determination module, configured to use the cluster category corresponding to the maximum output probability of various clusters in each test subset as the classification result corresponding to each test subset.

[0034] Optionally, the two - layer Gaussian process mixture model determination module specifically includes:

[0035] A total class number determination unit, configured to determine the total number of classes K of clusters included in each training subset.

[0036] A two - layer Gaussian process mixture model determination unit, configured to determine the two - layer Gaussian process mixture models corresponding to each training subset based on the input feature sets of various clusters in each training subset.

[0037] Optionally, the total class number determination unit specifically includes:

[0038] A load cluster feature determination subunit, configured to determine the load cluster features based on the features corresponding to various clusters.

[0039] A first judgment subunit, configured to judge the training subset S lWhether there is a cluster that meets the load cluster characteristics; if there is a cluster that meets the load cluster characteristics, execute the "total class number and input feature set determination subunit"; if there is no cluster that meets the load cluster characteristics, eliminate the training subset S l , and let l = l + 1, and execute the "second judgment subunit";

[0040] The second judgment subunit is used to judge whether l is greater than M. If l is greater than M, end; if l is less than or equal to M, return to the "first judgment subunit";

[0041] The total class number and input feature set determination subunit is used to determine the total class number K of clusters included in each training subset and the input feature sets of each class of clusters according to the load cluster characteristics.

[0042] Optionally, the double-layer Gaussian process mixture model determination unit specifically includes:

[0043] The Gaussian process model construction subunit is used to construct a Gaussian process model corresponding to each input feature based on each input feature in the i-th input feature set; the input feature set includes p input features;

[0044] The lower-layer model construction subunit is used to construct a Gaussian process mixture model corresponding to the i-th class of clusters according to the Gaussian process models corresponding to p input features; the Gaussian process mixture model is the lower-layer model;

[0045] The third judgment subunit is used to judge whether i is less than K. If i is less than K, let i = i + 1, and execute the "Gaussian process model construction subunit"; if i is greater than or equal to K, construct an upper-layer model according to the Gaussian process mixture models corresponding to K classes of clusters;

[0046] The double-layer Gaussian process mixture model determination subunit is used to determine a double-layer Gaussian process mixture model based on the upper-layer model and the lower-layer models corresponding to each class of clusters.

[0047] Optionally, the load cluster includes at least one of a computer cluster, an air conditioner cluster, a water heater cluster, and an electric vehicle cluster.

[0048] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0049] The solution of the present invention models the loads with the same operation 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 of the unknown cluster in the test subset, so as to effectively realize the classification of the cluster, provide guidance for users to improve power consumption efficiency, and will be of great significance for the safe, reliable, stable and economic operation of the power system, and is conducive to promoting the realization of comprehensive power consumption intelligence. Description of the Drawings

[0050] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0051] Figure 1 It is a flowchart of the non-intrusive load cluster classification method of the present invention;

[0052] Figure 2 It is a structural diagram of the non-intrusive load cluster classification system of the present invention;

[0053] Figure 3 It is a schematic diagram of the architecture of the non-intrusive load cluster classification system of the present invention;

[0054] Figure 4 It is a total current waveform diagram obtained by the non-intrusive acquisition device of the present invention;

[0055] Figure 5 It is the current waveform of the computer cluster of the present invention;

[0056] Figure 6 It is the current waveform of the air conditioner cluster of the present invention. Specific embodiments

[0057] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0058] The purpose of the present invention is to provide a non-intrusive load cluster classification method and system to achieve the classification of load clusters with the same operation characteristics and unified power consumption rules.

[0059] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the drawings and specific embodiments.

[0060] Embodiment 1

[0061] As Figure 1 shown, the present invention discloses a non-intrusive load cluster classification method, and the method includes:

[0062] Step S1: Construct a test set T and a training set S based on the electricity consumption dataset D corresponding to the load clusters; the training set S includes M training subsets; the test set T includes M test subsets; where M is a positive integer greater than 1.

[0063] Step S2: Determine the double-layer Gaussian process mixture model corresponding to each training subset; the parameters of the double-layer Gaussian process mixture model need to be solved.

[0064] Step S3: Solve the parameters of the double-layer Gaussian process mixture model corresponding to each training subset

[0065] Step S4: Calculate the mean of the M groups of solved parameters, and substitute the mean into the double-layer Gaussian process mixture model to obtain the double-layer Gaussian process mixture average model.

[0066] Step S5: Substitute various clusters in each test subset into the double-layer Gaussian process mixture average model for calculation, and output the probability values of various clusters in each test subset.

[0067] Step S6: Take the cluster category corresponding to the maximum output probability of various clusters in each test subset as the classification result corresponding to each test subset.

[0068] The following is a detailed discussion of each step:

[0069] Step S1: Construct a test set T and a training set S based on the electricity consumption dataset D corresponding to the load clusters, specifically including:

[0070] Step S11: Use a non-intrusive acquisition device to collect electricity consumption data at the user's power inlet to construct the electricity consumption dataset D; specifically, in the present invention, the total current data or total power data is collected at the user's power inlet by using a non-intrusive acquisition device through high-frequency acquisition.

[0071] Step S12: Use the M-fold cross-validation method to divide the electricity consumption dataset D into a test set T and a training set S; specifically, divide the electricity consumption dataset D into M subsets, i.e., D = [D 1 , D 2 , …, D M ; each time, take the l-th subset D l (l = 1, 2, …, M) as the test subset T l , and the remaining M - l subsets as the training subset S l ; based on the M groups of test subsets T l construct the test set T, i.e., T = [T 1 , T 2 , …, T M , and based on the M groups of training subsets S l construct the training set S, i.e., S = [S 1,S 2 ,…,S M . In this embodiment, M is arbitrarily selected according to actual needs. Preferably, M is taken as 10.

[0072] Step S2: Determine the double-layer Gaussian process mixture model corresponding to each training subset, specifically including:

[0073] Step S21: Determine the total number of classes K of clusters included in each training subset, specifically including:

[0074] Step S211: Determine the load cluster characteristics based on the characteristics corresponding to each type of cluster; specifically, the characteristics corresponding to each type of cluster include: working characteristics, operating characteristics, and random characteristics, etc. The load clusters include at least one of computer clusters, air conditioner clusters, water heater clusters, and electric vehicle clusters.

[0075] Step S212: Determine whether there is a cluster in the training subset S l that satisfies the load cluster characteristics; if there is a cluster that satisfies the load cluster characteristics, execute "Step S214"; if there is no cluster that satisfies the load cluster characteristics, remove the training subset S l , and let l = l + 1, and execute "Step S213".

[0076] Step S213: Determine whether l is greater than M. If l is greater than M, end; if l is less than or equal to M, return to "Step S212".

[0077] Step S214: Determine the total number of classes K of clusters included in each training subset and the input feature set of each type of cluster according to the load cluster characteristics; the input feature set includes p input features.

[0078] Step S22: Determine the double-layer Gaussian process mixture model corresponding to each training subset based on the input feature sets of each type of cluster in each training subset, specifically including:

[0079] Step S221: Construct a Gaussian process model corresponding to each input feature based on each input feature in the i-th input feature set. The specific formula is:

[0080] y ig (x) ∼ GPFR(x g ; b ig , θ ig );

[0081] where x g represents the g-th input feature in the input feature set x' k of the i-th type of cluster, g = 1,..., p, x' k = [x 1 , x 2 ,…, xp , b ig , θ ig respectively represent the mean and standard deviation of the g-th input feature in the input feature set x' of the i-th type of cluster, GPFR() represents the Gaussian process function, and y k (x) represents the Gaussian process model corresponding to the g-th input feature in the input feature set x' of the i-th type of cluster. ig (x) represents the Gaussian process model corresponding to the g-th input feature in the input feature set x' of the i-th type of cluster. k in the i-th type of cluster.

[0082] Step S222: Construct the Gaussian process mixture model corresponding to the i-th type of cluster based on the Gaussian process models corresponding to p input features, that is, the lower-layer model corresponding to the i-th type of cluster.

[0083] Step S223: Determine whether i is less than K. If i is less than K, then set i = i + 1 and execute "Step S221"; if i is greater than or equal to K, then construct the upper-layer model based on the Gaussian process mixture models corresponding to K types of clusters.

[0084] Step S224: Determine the double-layer Gaussian process mixture model based on the upper-layer model and the lower-layer models corresponding to each type of cluster, specifically including:

[0085] Define an indicator variable (i.e., latent variable) in the upper-layer model to describe the relevance of the s-th double-layer Gaussian process mixture model relative to the lower-layer Gaussian process model. If the i-th type of cluster belongs to the s-th Gaussian process mixture model, then otherwise At the same time obeys the following probability distribution:

[0086]

[0087] where π s is the probability when the indicator variable in the upper-layer model.

[0088] In the lower-layer model, the relevance of the s-th double-layer Gaussian process mixture model relative to the lower-layer Gaussian process model can be described by another latent variable If the g-th input feature of the i-th type of cluster belongs to the q-th Gaussian process model and under the condition that the i-th type of load cluster belongs to the s-th double-layer Gaussian process mixture model, then otherwise At the same time obeys the following probability distribution:

[0089]

[0090] where ηq|s is the probability when the indicator variable in the lower-layer model , and n is the total number of Gaussian process models.

[0091] Let and assume that the input variables follow a Gaussian distribution in each group, that is:

[0092]

[0093] where x ig is the input variable, and h qs , represent the mean and covariance respectively.

[0094] Define the specific formula of the two-layer Gaussian process mixture model as:

[0095] Θ = {(π s , η q|s , Θ qs )|s = 1,..., K; q = 1,..., n}

[0096] where Θ qs = {h qs , s qs , b qs , θ qs}, h qs , s qs are the mean and standard deviation of the upper-layer model respectively, b qs , θ qs are the mean and standard deviation of the lower-layer model respectively, η q|s is the probability when the indicator variable in the lower-layer model , π s is the probability when the indicator variable in the upper-layer model , n is the total number of Gaussian process models, and K is the total number of clusters included in the load cluster.

[0097] Step S3: Solve the parameters of the two-layer Gaussian process mixture model corresponding to each training subset where h qs , s qs are the mean and standard deviation of the upper-layer model respectively, b qs , θ qs are the mean and standard deviation of the lower-layer model respectively, η q|s is the probability when the indicator variable in the lower-layer model , π s is the probability when the indicator variable in the upper-layer model .

[0098] Step S3 specifically includes:

[0099] Step S31: Initialize the parameters of the double-layer Gaussian process mixture model

[0100] Step S32: Adopt an improved MCMC algorithm to calculate all the latent variables Z in the upper-layer model and all the latent variables A in the lower-layer model according to

[0101] Step S33: Establish the likelihood function of the double-layer Gaussian process mixture model for the load clusters based on all the latent variables Z in the upper-layer model and all the latent variables A in the lower-layer model

[0102] Step S34: Establish the expected function of the conditional probability distribution of the unknown data based on the likelihood function Calculate as follows

[0103]

[0104] where I represents the number of latent variables Z, Z (i) represents the i-th MCMC sampling of Z, D represents the samples in the sample set, E A [ ] represents the expected calculation function, L(Θ, Z(i), A) represents the likelihood function, represents the expected function of the conditional probability distribution of the unknown data

[0105] Step S35: Based on maximizing and update

[0106] Step S36: Determine whether the convergence condition is satisfied; if the convergence condition is satisfied, output the parameters of the double-layer Gaussian process mixture model corresponding to the M groups of training subsets If the convergence condition is not satisfied, return to "Step S32"; the convergence conditions in this embodiment include but are not limited to no longer changes

[0107] Step S5: Substitute each type of cluster in each test subset into the double-layer Gaussian process mixture average model for calculation, and output the probability values of each type of cluster in each test subset. The specific calculation formula is

[0108]

[0109] where p(z|x) represents the probability that the z-th type of cluster in the test subset T l belongs to the double-layer Gaussian process mixture model x, p(x) represents the probability that the x-th type of cluster appears in the double-layer Gaussian process mixture average model, and p(zx) represents the probability that the z-th type of cluster in the test subset T l exists in the x-th type of cluster in the double-layer Gaussian process mixture average model

[0110] Embodiment 2​

[0111] As shown in Figures 2 - 3 the figure, the present invention discloses a non-invasive load clustering classification system, and the system includes:

[0112] A dataset construction module 201, configured to construct a test set T and a training set S based on an electricity consumption dataset D corresponding to a load cluster; the training set S includes M training subsets; the test set T includes M test subsets; where M is a positive integer greater than 1.

[0113] A double-layer Gaussian process mixture model determination module 202, configured to determine a double-layer Gaussian process mixture model corresponding to each training subset; the parameters of the double-layer Gaussian process mixture model are to be solved.

[0114] A parameter solving module 203, configured to solve the parameters of the double-layer Gaussian process mixture model corresponding to each training subset.

[0115] A double-layer Gaussian process mixture average model determination module 204, configured to calculate the mean value of the M groups of solved parameters, and substitute the mean value into the double-layer Gaussian process mixture model to obtain a double-layer Gaussian process mixture average model.

[0116] A probability value calculation module 205, configured to substitute various clusters in each test subset into the double-layer Gaussian process mixture average model for calculation, and output the probability values of various clusters in each test subset.

[0117] A classification result determination module 206, configured to use the cluster category corresponding to the maximum output probability of various clusters in each test subset as the classification result corresponding to each test subset.

[0118] As an optional implementation manner, the double-layer Gaussian process mixture model determination module 202 of the present invention specifically includes:

[0119] A total class number determination unit, configured to determine the total class number K of clusters included in each training subset.

[0120] A double-layer Gaussian process mixture model determination unit, configured to determine a double-layer Gaussian process mixture model corresponding to each training subset based on the input feature sets of various clusters in each training subset.

[0121] As an optional implementation manner, the total class number determination unit of the present invention specifically includes:

[0122] A load cluster feature determination subunit, configured to determine load cluster features based on the features corresponding to various clusters.

[0123] A first judgment subunit, configured to judge the training subset S lWhether there is a cluster that satisfies the load cluster characteristics; if there is a cluster that satisfies the load cluster characteristics, execute the "total class number and input feature set determination subunit"; if there is no cluster that satisfies the load cluster characteristics, eliminate the training subset S l , and let l = l + 1, and execute the "second judgment subunit".

[0124] The second judgment subunit is used to judge whether l is greater than M. If l is greater than M, end; if l is less than or equal to M, return to the "first judgment subunit".

[0125] The total class number and input feature set determination subunit is used to determine the total class number K of clusters included in each training subset and the input feature sets of various clusters according to the load cluster characteristics.

[0126] As an optional implementation manner, the double-layer Gaussian process mixture model determination unit of the present invention specifically includes:

[0127] The Gaussian process model construction subunit is used to construct Gaussian process models corresponding to each input feature based on each input feature in the i-th input feature set; the input feature set includes p input features.

[0128] The lower-layer model construction subunit is used to construct a Gaussian process mixture model corresponding to the i-th type of cluster according to the Gaussian process models corresponding to p input features; the Gaussian process mixture model is the lower-layer model.

[0129] The third judgment subunit is used to judge whether i is less than K. If i is less than K, let i = i + 1, and execute the "Gaussian process model construction subunit"; if i is greater than or equal to K, construct the upper-layer model according to the Gaussian process mixture models corresponding to K types of clusters.

[0130] The double-layer Gaussian process mixture model determination subunit is used to determine the double-layer Gaussian process mixture model based on the upper-layer model and the lower-layer models corresponding to various clusters.

[0131] Embodiment 3

[0132] Install a non-invasive acquisition device at the power inlet of a small office to collect voltage waveform and current waveform signal data. The sampling frequency is 5000Hz, and the acquisition duration is one week.

[0133] Specifically, the following steps of the present invention are adopted:

[0134] 1), Install a non-invasive acquisition device at the power inlet of the small office to collect voltage waveform and current waveform signal data, as Figure 4 shown, and divide the power consumption data into a test set and a training set by using the ten-fold cross-validation method.

[0135] 2), Detect the training subset S l to determine whether there is a cluster that satisfies the load cluster characteristics, and determine the input feature x and output feature y for load cluster classification.

[0136] 3), For the i-th type of cluster (such as computer cluster) in the training set S l to establish a Gaussian process mixture model, forming the lower layer model of the two-layer Gaussian process mixture model.

[0137] 4), Establish K Gaussian process mixture models for K types of clusters (such as computer clusters and air conditioner clusters), forming the upper layer model of the two-layer Gaussian process mixture model.

[0138] 5), Define the indicator variables (latent variables) of the upper layer model and the lower layer model in the two-layer Gaussian process mixture model of the load cluster respectively, forming the two-layer Gaussian process mixture model of the load cluster.

[0139] 6), Solve the parameters of the two-layer Gaussian process mixture model of the load cluster.

[0140] 7), Repeat steps 2) to 6) to obtain 10 sets of parameters of the two-layer Gaussian process mixture model of the load cluster, and take the mean value to finally determine the parameter values.

[0141] 8), Calculate the probabilities that various load clusters in the test set T l belong to the two-layer Gaussian process mixture average model, and determine the category to which the cluster belongs according to the maximum probability. Figure 5 and Figure 6 are the current waveforms of the computer cluster and the air conditioner cluster respectively.

[0142] In this specification, each embodiment is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. The same or similar parts among the embodiments can be referred 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 parts can be referred to the description of the method part.

[0143] In this article, specific examples are used to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A non-intrusive load cluster classification method, characterized in that, the method includes: Constructing a test set T and a training set S based on the electricity consumption dataset D corresponding to the load cluster; the training set S includes M training subsets; the test set T includes M test subsets; where M is a positive integer greater than 1; Determine the double-layer Gaussian process mixture model corresponding to each training subset, specifically including: determining the total number of classes K of clusters included in each training subset; determining the double-layer Gaussian process mixture model corresponding to each training subset based on the input feature sets of each type of cluster in each training subset; wherein, determining the total number of classes K of clusters included in each training subset specifically includes: Step S211: Determine the load cluster features based on the features corresponding to each type of cluster; Step S212: Judge whether there is a cluster in the training subset S l that satisfies the load cluster features; if there is a cluster that satisfies the load cluster features, then execute "Step S214"; if there is no cluster that satisfies the load cluster features, then remove the training subset S l , and let l = l + 1, and execute "Step S213"; Step S213: Judge whether l is greater than M, if l is greater than M, then end; if l is less than or equal to M, then return to "Step S212"; Step S214: Determine the total number of classes K of clusters included in each training subset and the input feature sets of each type of cluster according to the load cluster features; determining the double-layer Gaussian process mixture model corresponding to each training subset based on the input feature sets of each type of cluster in each training subset specifically includes: Step S221: Construct a Gaussian process model corresponding to each input feature based on each input feature in the i-th input feature set; the input feature set includes p input features; Step S222: Construct a Gaussian process mixture model corresponding to the i-th type of cluster according to the Gaussian process models corresponding to the p input features; the Gaussian process mixture model is the lower-layer model; Step S223: Judge whether i is less than K, if i is less than K, then let i = i + 1, and execute "Step S221"; if i is greater than or equal to K, then construct the upper-layer model according to the Gaussian process mixture models corresponding to the K types of clusters; Step S224: Determine the double-layer Gaussian process mixture model based on the upper-layer model and the lower-layer models corresponding to each type of cluster; the parameters of the double-layer Gaussian process mixture model are to be solved; Solving the parameters of the double-layer Gaussian process mixture model corresponding to each training subset; Taking the mean of the M sets of solved parameters and substituting the mean into the double-layer Gaussian process mixture model to obtain a double-layer Gaussian process mixture average model; Substituting various clusters in each test subset into the double-layer Gaussian process mixture average model for calculation, and outputting the probability values of various clusters in each test subset; Taking the cluster category corresponding to the maximum output probability of various clusters in each test subset as the classification result corresponding to each test subset.

2. The non-intrusive load cluster classification method according to claim 1, characterized in that, the load cluster includes at least one of a computer cluster, an air conditioner cluster, a water heater cluster, and an electric vehicle cluster.

3. A non-intrusive load cluster classification system, characterized in that, the system includes: A dataset construction module for constructing a test set T and a training set S based on the electricity consumption dataset D corresponding to the load cluster; the test set T includes M test subsets; where M is a positive integer greater than 1; A double-layer Gaussian process mixture model determination module is used to determine the double-layer Gaussian process mixture model corresponding to each training subset. The double-layer Gaussian process mixture model determination module specifically includes: a total class number determination unit for determining the total number of classes K of clusters included in each training subset; a double-layer Gaussian process mixture model determination unit for determining the double-layer Gaussian process mixture model corresponding to each training subset based on the input feature sets of various clusters in each training subset. Among them, the total class number determination unit specifically includes: a load cluster feature determination subunit for determining the load cluster feature based on the features corresponding to various clusters; a first judgment subunit for judging whether there is a cluster in the training subset S l that satisfies the load cluster feature. If there is a cluster that satisfies the load cluster feature, the "total class number and input feature set determination subunit" is executed; if there is no cluster that satisfies the load cluster feature, the training subset S l is removed, and let l = l + 1, and the "second judgment subunit" is executed; a second judgment subunit for judging whether l is greater than M. If l is greater than M, it ends; if l is less than or equal to M, it returns to the "first judgment subunit"; a total class number and input feature set determination subunit for determining the total number of classes K of clusters included in each training subset and the input feature sets of various clusters according to the load cluster feature. The double-layer Gaussian process mixture model determination unit specifically includes: a Gaussian process model construction subunit for constructing a Gaussian process model corresponding to each input feature based on each input feature in the i-th input feature set. The input feature set includes p input features; a lower layer model construction subunit for constructing a Gaussian process mixture model corresponding to the i-th class of clusters according to the Gaussian process models corresponding to the p input features. The Gaussian process mixture model is the lower layer model; a third judgment subunit for judging whether i is less than K. If i is less than K, let i = i + 1, and the "Gaussian process model construction subunit" is executed; if i is greater than or equal to K, an upper layer model is constructed according to the Gaussian process mixture models corresponding to the K classes of clusters; a double-layer Gaussian process mixture model determination subunit for determining the double-layer Gaussian process mixture model based on the upper layer model and the lower layer models corresponding to various clusters. The parameters of the double-layer Gaussian process mixture model are to be solved; A parameter solving module for solving the parameters of the double-layer Gaussian process mixture model corresponding to each training subset; A double-layer Gaussian process mixture average model determination module for taking the mean of the M sets of solved parameters and substituting the mean into the double-layer Gaussian process mixture model to obtain a double-layer Gaussian process mixture average model; A probability value calculation module for substituting various clusters in each test subset into the double-layer Gaussian process mixture average model for calculation, and outputting the probability values of various clusters in each test subset; A classification result determination module for taking the cluster category corresponding to the maximum output probability of various clusters in each test subset as the classification result corresponding to each test subset.

4. The non-intrusive load cluster classification system according to claim 3, characterized in that, the load cluster includes at least one of a computer cluster, an air conditioner cluster, a water heater cluster, and an electric vehicle cluster.

Images