Clustering extraction method and system for power utilization characteristics of power distribution network users

Abstract

The invention relates to the technical field of power utilization feature extraction, in particular to a power distribution network user power utilization feature clustering extraction method and system, and the method comprises the steps: collecting data of N loads in t time periods, constructing an improved fuzzy C-means clustering objective function, solving an initial clustering center, solving an optimal clustering center, obtaining an optimal clustering number, and completing feature extraction. According to the method, noise and abnormal sensitive values are solved through an entropy weight method in construction of an improved C-means clustering objective function, the problem that an initial clustering center falls into local optimum is avoided, the initial clustering center and an optimal clustering center are extracted through a star-sparrow optimization algorithm, namely, a multi-position and multi-region population search algorithm, the optimal clustering number is obtained, and the clustering accuracy is improved. The technical problem that the algorithm can only search singly and randomly select the clustering center is solved, the accuracy and extraction efficiency of power consumption feature extraction are improved, and the optimal clustering result is obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of electricity consumption feature extraction, and in particular to a method and system for clustering and extracting electricity consumption features of users in a distribution network. Background Art

[0002] The core of electricity consumption feature extraction of power users is cluster analysis. There are two very complex problems in the clustering research process. One is how to divide the data set to get the best clustering result, and the other is how to determine the number of categories for the data set.

[0003] To address these two problems, the existing technologies mainly use a two-layer iterative method of cluster analysis, a user load classification method of fuzzy clustering, an improved genetic algorithm method for searching cluster centers, and a gray wolf algorithm to implement cluster center search, or use an antlion clustering algorithm, a firefly optimization algorithm, a whale optimization algorithm, an arithmetic optimization algorithm, a grasshopper optimization algorithm, and other optimized clustering algorithms to reduce data dimensionality and improve clustering efficiency. However, the methods used in the existing technologies cannot solve the problem of the optimal number of clusters, and randomly set cluster centers are prone to fall into local values ​​and cannot obtain a global optimal solution. Summary of the invention

[0004] In order to solve the technical problem that the methods used in the prior art cannot solve the problem of the optimal number of clusters, and the randomly set cluster centers are easily trapped in local values ​​and cannot obtain the global optimal solution, the present invention provides a method and system for clustering and extracting electricity consumption characteristics of distribution network users.

[0005] The technical solution adopted by the present invention to achieve the above-mentioned purpose is: a method for clustering and extracting power consumption characteristics of distribution network users, comprising:

[0006] Collect data of N loads in t time periods;

[0007] Clustering is performed according to the collected data of N loads in t time periods to construct an improved fuzzy C-means clustering objective function; and the initial clustering center is solved according to the constructed improved fuzzy C-means clustering objective function;

[0008] According to the constructed improved fuzzy C-means clustering objective function, the optimal cluster center is solved;

[0009] The optimal number of clusters is obtained based on the initial cluster center and the optimal cluster center, and feature extraction is completed.

[0010] Preferably, constructing an improved fuzzy C-means clustering objective function includes:

[0011] Construct the fuzzy C-means clustering objective function, whose expression is:

[0012]

[0013] Among them, J(U,V) is the distance from each sample to the cluster center, N is the number of loads, C is the number of clusters, and x t is the ith data in the input load data sequence, V = {v 1 ,v 2 ,…v C} represents the randomly set initial cluster center, U is the membership matrix of the initial setting, u ij is the membership degree between the i-th sample and the j-th cluster center, v j is the jth cluster center, and m is the weight coefficient of membership;

[0014] The Lagrangian function is constructed for the fuzzy C-means clustering objective function and the extreme value is obtained. Its expression is:

[0015]

[0016] Among them, u ij is the degree of membership between the i-th sample and the j-th cluster center, C is the number of clusters, k is an increasing variable from 1 to C, x i is the i-th data in the input load data sequence, v j is the jth cluster center, m is the weight coefficient of membership, and N is the number of loads;

[0017] The data of N loads in t time periods are normalized to obtain the processed data set X, which is expressed as:

[0018]

[0019] Among them, x ij (i=1,2,…,N;j=1,2,…,t), is the value of the i-th load at the j-th time;

[0020] Calculate the characteristic weight of the i-th load in the j-th column of the data set X, and its expression is:

[0021]

[0022] Among them, f ij is the characteristic weight of the i-th load in the j-th column of the data set X, x ij is the value of the i-th load at the j-th time, x kj is the value of the kth load at the jth time, and N is the number of loads;

[0023] Calculate the entropy value of the jth (j=1,2,…,t) column, the expression is:

[0024]

[0025] Among them, H j is the entropy value of the jth column, N is the number of loads, f ij is the characteristic weight of the i-th load in the j-th column of the data set X;

[0026] Calculate the entropy weight of the jth (j=1,2,…,t) column, and its expression is:

[0027]

[0028] Among them, ω 0j is the entropy weight of the jth column, H j is the entropy value of the jth column, H k is the kth value from k=1 to t;

[0029] Consider eliminating the influence of noise and outliers on entropy and weights, and obtain the improved weights, whose calculation expression is:

[0030]

[0031] Among them, ω j is the improved weight, is the mean of the entropy values ​​of the jth column, p is the value of the load time t sampling points, ω 3j is the improved weight, ω 0j is the entropy weight of the jth column, H j is the entropy value of the jth column;

[0032] The improved weight ω 3j The expression is:

[0033]

[0034] Among them, ω 3j is the improved weight, is the mean entropy value of the jth column, H j is the entropy value of the jth column, H k is the kth value from k=1 to t;

[0035] The improved fuzzy C-means clustering objective function is obtained, and its expression is:

[0036]

[0037] Among them, J(U,V) is the distance from each sample to the cluster center, N is the number of loads, C is the number of clusters, and d ij is the distance between the i-th sample and the j-th cluster center, ξ i is the coefficient, V={v 1 ,v 2 ,…v C} represents the randomly set initial cluster center, U is the membership matrix of the initial setting, u ij is the membership degree between the i-th sample and the j-th cluster center, v j is the jth cluster center, m is the weight coefficient of membership, ω j is the improved weight.

[0038] Preferably, the star-sparrow algorithm is used to solve the initial cluster center, and its expression is:

[0039]

[0040] in, is the position of the ith star bird at the t+1th iteration, is the jth position of the ith star bird at the tth iteration, τ 1 , τ 2 、r 1 , r is a random number between 0 and 1, is the j-th dimension mean of the t-th iteration, are the jth positions of three randomly selected individuals exploring high-quality solutions in the starfinch population at the tth iteration, γ is the random number formed by the Levy flight function, μ is the coefficient, and U j and L j are the upper and lower limits of the j-th dimension position, T max is the limit of the number of iterations, and are the jth positions of the tth iteration of star bird A and star bird B respectively;

[0041] in:

[0042]

[0043] Among them, r 1 、r 2 、r 3 are random numbers that obey uniform distribution, τ 3 is a random number between 0 and 1, τ 4 is a random number that follows a normal distribution, τ 5 The random number formed by the Levy flight function;

[0044] Get the new position where the solution is stored after the t+1th iteration of the i-th star bird, and its expression is:

[0045]

[0046] in, is the new location where the solution is stored after the t+1th iteration of the i-th star bird, is the position of the tth iteration of the i-th star bird, μ is the coefficient, is the optimal solution for the tth iteration, λ is the coefficient, r is a random number between 0 and 1, is the t-th iteration value of star A, is the t-th iteration value of star B, r 1 、r 2 、r 3 are random numbers that obey uniform distribution, and l is the attenuation factor.

[0047] Preferably, solving the optimal cluster center includes the following steps:

[0048] If the starfinch remembers the position of the initial solution, it will find the iterative update position of the solution according to the following formula, which is expressed as:

[0049]

[0050] in, is the jth position of the ith star bird at the t+1th iteration, is the jth position of the ith star bird at the tth iteration, τ 3 is a random number between 0 and 1, τ 4 is a random number that follows a normal distribution, r 1 is a random number that follows a uniform distribution, r 2 is a random number that follows a uniform distribution, is the optimal position of the jth dimension in the tth iteration, is the position of the ith star bird at the first reference point in the tth iteration, is the jth position of the tth iteration of star C;

[0051] If the starfinch does not remember the position of the initial solution, it will use the following formula to find the iterative update position of the solution, which is expressed as:

[0052]

[0053] in, is the jth position of the ith star bird at the t+1th iteration, is the jth position of the ith star bird at the tth iteration, τ 3 is a random number between 0 and 1, τ 4 is a random number that follows a normal distribution, r 1 is a random number that follows a uniform distribution, r 2 is a random number that follows a uniform distribution, is the optimal position of the jth dimension in the tth iteration, is the position of the ith star bird at the second reference point in the tth iteration, is the jth position of the tth iteration of star C;

[0054] The optimal solution position is searched near the first reference point according to the memory. If the optimal solution is not found at the first reference point, the optimal solution is searched at the second reference point and other areas. The iterative update model for searching the optimal solution position near the first reference point according to the memory is:

[0055]

[0056] in, is the jth position of the ith star bird at the t+1th iteration, is the ith position of the tth iteration, To position Enter the value of the objective function of the fuzzy C-means clustering algorithm to find the cluster center, To position Enter the value of the objective function of the fuzzy C-means clustering algorithm to find the cluster center, is the position of the i-th starfinch at the first reference point in the t-th iteration;

[0057] Find the optimal solution at the second reference point and other areas, and its iterative update model is:

[0058]

[0059] in, is the jth position of the ith star bird at the t+1th iteration, is the ith position of the tth iteration, To position Enter the value of the objective function of the fuzzy C-means clustering algorithm to find the cluster center, To position Enter the value of the objective function of the fuzzy C-means clustering algorithm to find the cluster center, is the position of the i-th starfinch at the second reference point in the t-th iteration.

[0060] Preferably, the optimal number of clusters is obtained by establishing a Davidson-Bottlen objective function according to the degree of membership, and obtaining the number of clusters when the mean of the maximum similarity of each class is the smallest as the optimal number of clusters, which is expressed as:

[0061]

[0062] Among them, DBI is the mean of the maximum similarity of each class, C is the number of clusters, S i Cluster A i elements in to their cluster center c i The mean Euclidean distance of j Cluster A j elements in to their cluster center c j The mean Euclidean distance of iCluster A i The number of elements in N j Cluster A j The number of elements in x i is the i-th data in the input load data sequence, is the similarity between the i-th category and the j-th category, is the maximum similarity between the i-th class and the j-th class.

[0063] A distribution network user power consumption feature clustering extraction system, comprising:

[0064] A data acquisition module is used to collect data of N loads in t time periods;

[0065] A clustering module is used to cluster the collected data of N loads in t time periods and construct an improved fuzzy C-means clustering objective function;

[0066] An initial cluster center solving module is used to solve the initial cluster center according to the constructed improved fuzzy C-means clustering objective function;

[0067] The optimal cluster center solving module is used to solve the optimal cluster center according to the constructed improved fuzzy C-means clustering objective function;

[0068] The feature extraction module is used to obtain the optimal number of clusters based on the initial cluster center and the optimal cluster center to complete feature extraction.

[0069] Preferably, the step of constructing an improved fuzzy C-means clustering objective function comprises:

[0070] Construct the fuzzy C-means clustering objective function, whose expression is:

[0071]

[0072] Among them, J(U,V) is the distance from each sample to the cluster center, N is the number of loads, C is the number of clusters, and x t is the ith data in the input load data sequence, V = {v 1 ,v 2 ,…v C} represents the randomly set initial cluster center, U is the membership matrix of the initial setting, u ij is the membership degree between the i-th sample and the j-th cluster center, v j is the jth cluster center, and m is the weight coefficient of membership;

[0073] The Lagrangian function is constructed for the fuzzy C-means clustering objective function and the extreme value is obtained. Its expression is:

[0074]

[0075] Among them, u ij is the degree of membership between the i-th sample and the j-th cluster center, C is the number of clusters, k is an increasing variable from 1 to C, x i is the i-th data in the input load data sequence, v j is the jth cluster center, m is the weight coefficient of membership, and N is the number of loads;

[0076] The data of N loads in t time periods are normalized to obtain the processed data set X, which is expressed as:

[0077]

[0078] Among them, x ij (i=1,2,…,N;j=1,2,…,t), is the value of the i-th load at the j-th time;

[0079] Calculate the characteristic weight of the i-th load in the j-th column of the data set X, and its expression is:

[0080]

[0081] Among them, f ij is the characteristic weight of the i-th load in the j-th column of the data set X, x ij is the value of the i-th load at the j-th time, x kj is the value of the kth load at the jth time, and N is the number of loads;

[0082] Calculate the entropy value of the jth (j=1,2,…,t) column, the expression is:

[0083]

[0084] Among them, H j is the entropy value of the jth column, N is the number of loads, f ij is the characteristic weight of the i-th load in the j-th column of the data set X;

[0085] Calculate the entropy weight of the jth (j=1,2,…,t) column, and its expression is:

[0086]

[0087] Among them, ω 0j is the entropy weight of the jth column, H j is the entropy value of the jth column, H k is the kth value from k=1 to t;

[0088] Consider eliminating the influence of noise and outliers on entropy and weights, and obtain the improved weights, whose calculation expression is:

[0089]

[0090] Among them, ω j is the improved weight, is the mean of the entropy values ​​of the jth column, p is the value of the load time t sampling points, ω 3j is the improved weight, ω 0j is the entropy weight of the jth column, H j is the entropy value of the jth column;

[0091] The improved weight ω 3j The expression is:

[0092]

[0093] Among them, ω 3j is the improved weight, is the mean entropy value of the jth column, H j is the entropy value of the jth column, H k is the kth value from k=1 to t;

[0094] The improved fuzzy C-means clustering objective function is obtained, and its expression is:

[0095]

[0096] Among them, J(U,V) is the distance from each sample to the cluster center, N is the number of loads, C is the number of clusters, and d ij is the distance between the i-th sample and the j-th cluster center, ξ i is the coefficient, V={v 1 ,v 2 ,…v C} represents the randomly set initial cluster center, U is the membership matrix of the initial setting, u ij is the membership degree between the i-th sample and the j-th cluster center, v j is the jth cluster center, m is the weight coefficient of membership, ω j is the improved weight.

[0097] Preferably, the star-sparrow algorithm is used to solve the initial cluster center, and its expression is:

[0098]

[0099] in, is the position of the ith star bird at the t+1th iteration, is the jth position of the ith star bird at the tth iteration, τ 1 , τ 2 、r 1, r is a random number between 0 and 1, is the j-th dimension mean of the t-th iteration, are the jth positions of three randomly selected individuals exploring high-quality solutions in the starfinch population at the tth iteration, γ is the random number formed by the Levy flight function, μ is the coefficient, and U j and L j are the upper and lower limits of the j-th dimension position, T max is the limit of the number of iterations, and are the jth positions of the tth iteration of star bird A and star bird B respectively;

[0100] in:

[0101]

[0102] Among them, r 1 、r 2 、r 3 are random numbers that obey uniform distribution, τ 3 is a random number between 0 and 1, τ 4 is a random number that follows a normal distribution, τ 5 The random number formed by the Levy flight function;

[0103] Get the new position where the solution is stored after the t+1th iteration of the i-th star bird, and its expression is:

[0104]

[0105] in, is the new location where the solution is stored after the t+1th iteration of the i-th star bird, is the position of the tth iteration of the i-th star bird, μ is the coefficient, is the optimal solution for the tth iteration, λ is the coefficient, r is a random number between 0 and 1, is the t-th iteration value of star A, is the t-th iteration value of star B, r 1 、r 2 、r 3 are random numbers that obey uniform distribution, and l is the attenuation factor.

[0106] Preferably, solving the optimal cluster center includes the following steps:

[0107] If the starfinch remembers the position of the initial solution, it will find the iterative update position of the solution according to the following formula, which is expressed as:

[0108]

[0109] in, is the jth position of the ith star bird at the t+1th iteration, is the jth position of the ith star bird at the tth iteration, τ 3 is a random number between 0 and 1, τ 4 is a random number that follows a normal distribution, r 1 is a random number that follows a uniform distribution, r 2 is a random number that follows a uniform distribution, is the optimal position of the jth dimension in the tth iteration, is the position of the ith star bird at the first reference point in the tth iteration, is the jth position of the tth iteration of star C;

[0110] If the starfinch does not remember the position of the initial solution, it will use the following formula to find the iterative update position of the solution, which is expressed as:

[0111]

[0112] in, is the jth position of the ith star bird at the t+1th iteration, is the jth position of the ith star bird at the tth iteration, τ 3 is a random number between 0 and 1, τ 4 is a random number that follows a normal distribution, r 1 is a random number that follows a uniform distribution, r 2 is a random number that follows a uniform distribution, is the optimal position of the jth dimension in the tth iteration, is the position of the ith star bird at the second reference point in the tth iteration, is the jth position of the tth iteration of star C;

[0113] The optimal solution position is searched near the first reference point according to the memory. If the optimal solution is not found at the first reference point, the optimal solution is searched at the second reference point and other areas. The iterative update model for searching the optimal solution position near the first reference point according to the memory is:

[0114]

[0115] in, is the jth position of the ith star bird at the t+1th iteration, is the ith position of the tth iteration, To position Enter the value of the objective function of the fuzzy C-means clustering algorithm to find the cluster center, To position Enter the value of the objective function of the fuzzy C-means clustering algorithm to find the cluster center, is the position of the i-th starfinch at the first reference point in the t-th iteration;

[0116] Find the optimal solution at the second reference point and other areas, and its iterative update model is:

[0117]

[0118] in, is the jth position of the ith star bird at the t+1th iteration, is the ith position of the tth iteration, To position Enter the value of the objective function of the fuzzy C-means clustering algorithm to find the cluster center, To position Enter the value of the objective function of the fuzzy C-means clustering algorithm to find the cluster center, is the position of the i-th starfinch at the second reference point in the t-th iteration.

[0119] Preferably, the optimal number of clusters is obtained by establishing a Davidson-Bottlen objective function according to the degree of membership, and obtaining the number of clusters when the mean of the maximum similarity of each class is the smallest as the optimal number of clusters, which is expressed as:

[0120]

[0121] Among them, DBI is the mean of the maximum similarity of each class, C is the number of clusters, S i Cluster A i elements in to their cluster center c i The mean Euclidean distance of j Cluster A j elements in to their cluster center c j The mean Euclidean distance of i Cluster A i The number of elements in N j Cluster A j The number of elements in x i is the i-th data in the input load data sequence, is the similarity between the i-th category and the j-th category, is the maximum similarity between the i-th class and the j-th class.

[0122] A device for clustering and extracting electricity consumption features of users in a distribution network comprises a memory and a processor, wherein the memory is used to store a computer program, and the computer program is used to execute the above method when loaded by the processor.

[0123] A computer-readable storage medium stores a computer program, wherein the computer program is suitable for executing the above method when loaded by a processor.

[0124] The present invention and the prior art, in constructing an improved C-means clustering objective function, solve the noise and abnormal sensitivity values ​​through the entropy weight method to avoid the problem that the initial cluster center falls into the local optimum, and extract the initial cluster center and the optimal cluster center through the star sparrow optimization algorithm, that is, adopting a multi-location, multi-region population search algorithm, to obtain the optimal number of clusters, solve the technical problem that the algorithm can only search individually and randomly select cluster centers, improve the accuracy and extraction efficiency of electricity consumption feature extraction, and obtain the optimal clustering result. BRIEF DESCRIPTION OF THE DRAWINGS

[0125] Figure 1 It is a flow chart of a method for clustering and extracting electricity consumption characteristics of users in a distribution network according to the present invention;

[0126] Figure 2 It is a comparison chart of DBI values ​​between the verification result of the present invention and the clustering result of the present method;

[0127] Figure 3 is the clustering result 1 in the embodiment of the present invention;

[0128] Figure 4 is the clustering result 2 in the embodiment of the present invention;

[0129] Figure 5 is the clustering result 3 in the embodiment of the present invention;

[0130] Figure 6 is the clustering result 4 in the embodiment of the present invention;

[0131] Figure 7 is the clustering result 5 in the embodiment of the present invention;

[0132] Figure 8 is the clustering result 6 in the embodiment of the present invention;

[0133] Fig. 9 This is the clustering result 7 in the embodiment of the present invention. DETAILED DESCRIPTION

[0134] The specific embodiments of the present invention are as follows Figure 1 As shown, the following steps are included:

[0135] A method for clustering and extracting power consumption characteristics of distribution network users, comprising the following steps:

[0136] Collect data from N loads in t time periods, and collect electricity consumption data from 1,000 user loads for 24 hours a day from 2020 to 2023;

[0137] According to the collected data of N loads in t time periods, clustering is performed to construct an improved fuzzy C-means clustering objective function;

[0138] The construction of the improved fuzzy C-means clustering objective function includes the following steps:

[0139] Construct the fuzzy C-means clustering objective function, whose expression is:

[0140]

[0141] Among them, J(U,V) is the distance from each sample to the cluster center, N is the number of loads, C is the number of clusters, and x t is the ith data in the input load data sequence, V = {v 1 ,v 2 ,…v C} represents the randomly set initial cluster center, U is the membership matrix of the initial setting, u ij is the membership degree between the i-th sample and the j-th cluster center, v j is the jth cluster center, and m is the weight coefficient of membership;

[0142] The Lagrangian function is constructed for the fuzzy C-means clustering objective function and the extreme value is obtained. Its expression is:

[0143]

[0144] Among them, u ij is the degree of membership between the i-th sample and the j-th cluster center, C is the number of clusters, k is an increasing variable from 1 to C, x i is the i-th data in the input load data sequence, v j is the jth cluster center, m is the weight coefficient of membership, and N is the number of loads;

[0145] The data of N loads in t time periods are normalized to obtain the processed data set X, which is expressed as:

[0146]

[0147] Among them, x ij (i=1,2,…,N;j=1,2,…,t), is the value of the i-th load at the j-th time;

[0148] Calculate the characteristic weight of the i-th load in the j-th column of the data set X, and its expression is:

[0149]

[0150] Among them, f ij is the characteristic weight of the i-th load in the j-th column of the data set X, x ij is the value of the i-th load at the j-th time, x kj is the value of the kth load at the jth time, and N is the number of loads;

[0151] Calculate the entropy value of the jth (j=1,2,…,t) column, the expression is:

[0152]

[0153] Among them, H j is the entropy value of the jth column, N is the number of loads, f ij is the characteristic weight of the i-th load in the j-th column of the data set X;

[0154] Calculate the entropy weight of the jth (j=1,2,…,t) column, and its expression is:

[0155]

[0156] Among them, ω 0j is the entropy weight of the jth column, H j is the entropy value of the jth column, H k is the kth value from k=1 to t;

[0157] Consider eliminating the influence of noise and outliers on entropy and weights, and obtain the improved weights, whose calculation expression is:

[0158]

[0159] Among them, ω j is the improved weight, is the mean of the entropy values ​​of the jth column, p is the value of the load time t sampling points, ω 3j is the improved weight, ω 0j is the entropy weight of the jth column, H j is the entropy value of the jth column;

[0160] The improved weight ω 3j The expression is:

[0161]

[0162] Among them, ω 3j is the improved weight, H is the mean entropy value of the jth column, and H j is the entropy value of the jth column, H k is the kth value from k=1 to t;

[0163] The improved fuzzy C-means clustering objective function is obtained, and its expression is:

[0164]

[0165] Among them, J(U,V) is the distance from each sample to the cluster center, N is the number of loads, C is the number of clusters, and d ijis the distance between the i-th sample and the j-th cluster center, ξ i is the coefficient, V={v 1 ,v 2 ,…v C} represents the randomly set initial cluster center, U is the membership matrix of the initial setting, u ij is the membership degree between the i-th sample and the j-th cluster center, v j is the jth cluster center, m is the weight coefficient of membership, ω j is the improved weight;

[0166] According to the constructed improved fuzzy C-means clustering objective function, the initial cluster center is solved;

[0167] The star-sparrow algorithm is used to solve the initial cluster center, and its expression is:

[0168]

[0169] in, is the position of the ith star bird at the t+1th iteration, is the jth position of the ith star bird at the tth iteration, τ 1 , τ 2 、r 1 , r is a random number between 0 and 1, is the j-th dimension mean of the t-th iteration, are the jth positions of three randomly selected individuals exploring high-quality solutions in the starfinch population at the tth iteration, γ is the random number formed by the Levy flight function, μ is the coefficient, and U j and L j are the upper and lower limits of the j-th dimension position, T max is the limit of the number of iterations, and are the jth positions of the tth iteration of star bird A and star bird B respectively;

[0170] in:

[0171]

[0172] Among them, r 1 、r 2 、r 3 are random numbers that obey uniform distribution, τ 3 is a random number between 0 and 1, τ 4 is a random number that follows a normal distribution, τ 5 The random number formed by the Levy flight function;

[0173] Get the new position where the solution is stored after the t+1th iteration of the i-th star bird, and its expression is:

[0174]

[0175] in, is the new location where the solution is stored after the t+1th iteration of the i-th star bird, is the position of the tth iteration of the i-th star bird, μ is the coefficient, is the optimal solution for the tth iteration, λ is the coefficient, r is a random number between 0 and 1, is the t-th iteration value of star A, is the t-th iteration value of star B, r 1 、r 2 、r 3 are random numbers that obey uniform distribution, and l is the attenuation factor;

[0176] According to the constructed improved fuzzy C-means clustering objective function, the optimal cluster center is solved;

[0177] Solving the optimal cluster center includes the following steps:

[0178] If the starfinch remembers the position of the initial solution, it will find the iterative update position of the solution according to the following formula, which is expressed as:

[0179]

[0180] in, is the jth position of the ith star bird at the t+1th iteration, is the jth position of the ith star bird at the tth iteration, τ 3 is a random number between 0 and 1, τ 4 is a random number that follows a normal distribution, r 1 is a random number that follows a uniform distribution, r 2 is a random number that follows a uniform distribution, is the optimal position of the jth dimension in the tth iteration, is the position of the ith star bird at the first reference point in the tth iteration, is the jth position of the tth iteration of star C;

[0181] If the starfinch does not remember the position of the initial solution, it will use the following formula to find the iterative update position of the solution, which is expressed as:

[0182]

[0183] in, is the jth position of the ith star bird at the t+1th iteration, is the jth position of the ith star bird at the tth iteration, τ 3 is a random number between 0 and 1, τ 4 is a random number that follows a normal distribution, r1 A random number that follows a uniform distribution, r 2 is a random number that follows a uniform distribution, is the optimal position of the jth dimension in the tth iteration, is the position of the ith star bird at the second reference point in the tth iteration, is the jth position of the tth iteration of star C;

[0184] The optimal solution position is searched near the first reference point according to the memory. If the optimal solution is not found at the first reference point, the optimal solution is searched at the second reference point and other areas. The iterative update model for searching the optimal solution position near the first reference point according to the memory is:

[0185]

[0186] in, is the jth position of the ith star bird at the t+1th iteration, is the ith position of the tth iteration, To position Enter the value of the objective function of the fuzzy C-means clustering algorithm to find the cluster center, To position Enter the value of the objective function of the fuzzy C-means clustering algorithm to find the cluster center, is the position of the i-th starfinch at the first reference point in the t-th iteration;

[0187] Find the optimal solution at the second reference point and other areas, and its iterative update model is:

[0188]

[0189] in, is the jth position of the ith star bird at the t+1th iteration, is the ith position of the tth iteration, To position Enter the value of the objective function of the fuzzy C-means clustering algorithm to find the cluster center, To position Enter the value of the objective function of the fuzzy C-means clustering algorithm to find the cluster center, is the position of the i-th starfinch at the second reference point in the t-th iteration;

[0190] Obtain the optimal number of clusters based on the initial cluster center and the optimal cluster center, and complete feature extraction;

[0191] To obtain the optimal number of clusters, the Davidson-Bottlen objective function is established according to the degree of membership, and the number of clusters with the smallest mean of the maximum similarity of each class is determined as the optimal number of clusters, and its expression is:

[0192]

[0193] Among them, DBI is the mean of the maximum similarity of each class, C is the number of clusters, S i Cluster A i elements in to their cluster center c i The mean Euclidean distance of j Cluster A j elements in to their cluster center c j The mean Euclidean distance of i Cluster A i The number of elements in N j Cluster A j The number of elements in x i is the i-th data in the input load data sequence, is the similarity between the i-th category and the j-th category, is the maximum similarity between the i-th class and the j-th class.

[0194] A distribution network user power consumption feature clustering extraction system, comprising:

[0195] A data acquisition module is used to collect data of N loads in t time periods;

[0196] The clustering module is used to cluster the collected data of N loads in t time periods and construct an improved fuzzy C-means clustering objective function; the construction of the improved fuzzy C-means clustering objective function includes:

[0197] Construct the fuzzy C-means clustering objective function, whose expression is:

[0198]

[0199] Among them, J(U,V) is the distance from each sample to the cluster center, N is the number of loads, C is the number of clusters, and x t is the ith data in the input load data sequence, V = {v 1 ,v 2 ,…v C} represents the randomly set initial cluster center, U is the membership matrix of the initial setting, u ij is the membership degree between the i-th sample and the j-th cluster center, v j is the jth cluster center, and m is the weight coefficient of membership;

[0200] The Lagrangian function is constructed for the fuzzy C-means clustering objective function and the extreme value is obtained. Its expression is:

[0201]

[0202] Among them, u ij is the degree of membership between the i-th sample and the j-th cluster center, C is the number of clusters, k is an increasing variable from 1 to C, x i is the i-th data in the input load data sequence, v j is the jth cluster center, m is the weight coefficient of membership, and N is the number of loads;

[0203] The data of N loads in t time periods are normalized to obtain the processed data set X, which is expressed as:

[0204]

[0205] Among them, x ij (i=1,2,…,N;j=1,2,…,t), is the value of the i-th load at the j-th time;

[0206] Calculate the characteristic weight of the i-th load in the j-th column of the data set X, and its expression is:

[0207]

[0208] Among them, f ij is the characteristic weight of the i-th load in the j-th column of the data set X, x ij is the value of the i-th load at the j-th time, x kj is the value of the kth load at the jth time, and N is the number of loads;

[0209] Calculate the entropy value of the jth (j=1,2,…,t) column, the expression is:

[0210]

[0211] Among them, H j is the entropy value of the jth column, N is the number of loads, f ij is the characteristic weight of the i-th load in the j-th column of the data set X;

[0212] Calculate the entropy weight of the jth (j=1,2,…,t) column, and its expression is:

[0213]

[0214] Among them, ω 0j is the entropy weight of the jth column, H j is the entropy value of the jth column, H k is the kth value from k=1 to t;

[0215] Consider eliminating the influence of noise and outliers on entropy and weights, and obtain the improved weights, whose calculation expression is:

[0216]

[0217] Among them, ω j is the improved weight, is the mean of the entropy values ​​of the jth column, p is the value of the load time t sampling points, ω 3j is the improved weight, ω 0j is the entropy weight of the jth column, H j is the entropy value of the jth column;

[0218] The improved weight ω 3j The expression is:

[0219]

[0220] Among them, ω 3j is the improved weight, is the mean entropy value of the jth column, H j is the entropy value of the jth column, H k is the kth value from k=1 to t;

[0221] The improved fuzzy C-means clustering objective function is obtained, and its expression is:

[0222]

[0223] Among them, J(U,V) is the distance from each sample to the cluster center, N is the number of loads, C is the number of clusters, and d ij is the distance between the i-th sample and the j-th cluster center, ξ i is the coefficient, V={v 1 ,v 2 ,…v C} represents the randomly set initial cluster center, U is the membership matrix of the initial setting, u ij is the membership degree between the i-th sample and the j-th cluster center, v j is the jth cluster center, m is the weight coefficient of membership, ω j The improved weight

[0224] The initial cluster center solving module is used to solve the initial cluster center according to the constructed improved fuzzy C-means clustering objective function; the initial cluster center is solved by the star-sparrow algorithm, and its expression is:

[0225]

[0226] in, is the position of the ith star bird at the t+1th iteration, is the jth position of the ith star bird at the tth iteration, τ 1 , τ 2 、r1 , r is a random number between 0 and 1, is the j-th dimension mean of the t-th iteration, are the jth positions of three randomly selected individuals exploring high-quality solutions in the starfinch population at the tth iteration, γ is the random number formed by the Levy flight function, μ is the coefficient, and U j and L j are the upper and lower limits of the j-th dimension position, T max is the limit of the number of iterations, and are the jth positions of the tth iteration of star bird A and star bird B respectively;

[0227] in:

[0228]

[0229] Among them, r 1 、r 2 、r 3 are random numbers that obey uniform distribution, τ 3 is a random number between 0 and 1, τ 4 is a random number that follows a normal distribution, τ 5 The random number formed by the Levy flight function;

[0230] Get the new position where the solution is stored after the t+1th iteration of the i-th star bird, and its expression is:

[0231]

[0232] in, is the new location where the solution is stored after the t+1th iteration of the i-th star bird, is the position of the tth iteration of the i-th star bird, μ is the coefficient, is the optimal solution for the tth iteration, λ is the coefficient, r is a random number between 0 and 1, is the t-th iteration value of star A, is the t-th iteration value of star B, r 1 、r 2 、r 3 are random numbers that obey uniform distribution, and l is the attenuation factor;

[0233] The optimal cluster center solving module is used to solve the optimal cluster center according to the constructed improved fuzzy C-means clustering objective function; solving the optimal cluster center includes the following steps:

[0234] If the starfinch remembers the position of the initial solution, it will find the iterative update position of the solution according to the following formula, which is expressed as:

[0235]

[0236] in, is the jth position of the ith star bird at the t+1th iteration, is the jth position of the ith star bird at the tth iteration, τ 3 is a random number between 0 and 1, τ 4 is a random number that follows a normal distribution, r 1 is a random number that follows a uniform distribution, r 2 is a random number that follows a uniform distribution, is the optimal position of the jth dimension in the tth iteration, is the position of the ith star bird at the first reference point in the tth iteration, is the jth position of the tth iteration of star C;

[0237] If the starfinch does not remember the position of the initial solution, it will use the following formula to find the iterative update position of the solution, which is expressed as:

[0238]

[0239] in, is the jth position of the ith star bird at the t+1th iteration, is the jth position of the ith star bird at the tth iteration, τ 3 is a random number between 0 and 1, τ 4 is a random number that follows a normal distribution, r 1 is a random number that follows a uniform distribution, r 2 is a random number that follows a uniform distribution, is the optimal position of the jth dimension in the tth iteration, is the position of the ith star bird at the second reference point in the tth iteration, is the jth position of the tth iteration of star C;

[0240] The optimal solution position is searched near the first reference point according to the memory. If the optimal solution is not found at the first reference point, the optimal solution is searched at the second reference point and other areas. The iterative update model for searching the optimal solution position near the first reference point according to the memory is:

[0241]

[0242] in, is the jth position of the ith star bird at the t+1th iteration, is the ith position of the tth iteration, To position Enter the value of the objective function of the fuzzy C-means clustering algorithm to find the cluster center, To position Enter the value of the objective function of the fuzzy C-means clustering algorithm to find the cluster center, is the position of the i-th starfinch at the first reference point in the t-th iteration;

[0243] Find the optimal solution at the second reference point and other areas, and its iterative update model is:

[0244]

[0245] in, is the jth position of the ith star bird at the t+1th iteration, is the ith position of the tth iteration, To position Enter the value of the objective function of the fuzzy C-means clustering algorithm to find the cluster center, To position Enter the value of the objective function of the fuzzy C-means clustering algorithm to find the cluster center, is the position of the i-th starfinch at the second reference point in the t-th iteration;

[0246] The feature extraction module is used to obtain the optimal number of clusters according to the initial cluster center and the optimal cluster center to complete the feature extraction; the optimal number of clusters is obtained by establishing the Davidson-Bottlen objective function according to the membership degree, and the number of clusters when the mean of the maximum similarity of each class is the smallest is determined as the optimal number of clusters, and its expression is:

[0247]

[0248] Among them, DBI is the mean of the maximum similarity of each class, C is the number of clusters, S i Cluster A i elements in to their cluster center c i The mean Euclidean distance of j Cluster A j elements in to their cluster center c j The mean Euclidean distance of i Cluster A i The number of elements in N j Cluster A j The number of elements in x i is the i-th data in the input load data sequence, is the similarity between the i-th category and the j-th category, is the maximum similarity between the i-th class and the j-th class.

[0249] A device for clustering and extracting electricity consumption features of users in a distribution network comprises a memory and a processor, wherein the memory is used to store a computer program, and the computer program is used to execute the above method when loaded by the processor.

[0250] A computer-readable storage medium stores a computer program, wherein the computer program is suitable for executing the above method when loaded by a processor.

[0251] The silhouette coefficient method was used to verify and compare the clustering results. The DBI values ​​of the verification results and the clustering results of this method are as follows: Figure 2 As shown by Figure 2 It can be seen that this method can jump out of the local optimal value to obtain the global optimal value, and by Figure 2 It can be seen that when the number of clusters is 7, the DBI value is the smallest. DBI represents the mean of the maximum similarity calculation for each class. The smaller the DBI value, the tighter the internal clusters are and the more dispersed the samples between classes are. Therefore, 7 is selected as the optimal number of clusters. The clustering results are as follows: Figures 3 to 9 As shown in the figure, this method is compared and verified with the traditional K-means algorithm, fuzzy C-means algorithm (FCM), K-reciprocalNeighbors (KN) and kernel density estimation (KDE) algorithm in the UCI database using AMI and ARI indicators. The results are shown in Table 1.

[0252] Table 1 Comparison of results in UCI database

[0253]

[0254] As can be seen from Table 1, among the four sub-databases of the UCI database (Blood\Breast\Diabets\Ecoli), the clustering method used in this paper has the best results under the calculation of the indicators AMI\ARI, indicating that the algorithm in this paper has better clustering effect than the traditional K-means algorithm, fuzzy C-means algorithm (FCM), K-reciprocal Neighbors algorithm (KN) and kernel density estimation (KDE) algorithm.

[0255] The present invention is described by way of embodiments, and those skilled in the art will appreciate that various changes or equivalent substitutions may be made to these features and embodiments without departing from the spirit and scope of the present invention. In addition, under the teachings of the present invention, these features and embodiments may be modified to adapt to specific circumstances and materials without departing from the spirit and scope of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application belong to the protection scope of the present invention.

Claims

1. A method for clustering and extracting electricity consumption features of distribution network users, characterized in that: include: Collect data of N loads in t time periods; Clustering is performed according to the collected data of N loads in t time periods to construct an improved fuzzy C-means clustering objective function; and the initial clustering center is solved according to the constructed improved fuzzy C-means clustering objective function; According to the constructed improved fuzzy C-means clustering objective function, the optimal cluster center is solved; The optimal number of clusters is obtained based on the initial cluster center and the optimal cluster center, and feature extraction is completed.

2. A method for extracting power consumption characteristics of distribution network users according to claim 1, characterized in that: The construction of improved fuzzy C-means clustering objective function includes: Construct the fuzzy C-means clustering objective function, whose expression is: Among them, J(U,V) is the distance from each sample to the cluster center, N is the number of loads, C is the number of clusters, and x t is the i-th data in the input load data sequence, V = {v1, v2, …v C } represents the randomly set initial cluster center, U is the membership matrix of the initial setting, u ij is the membership degree between the i-th sample and the j-th cluster center, v j is the jth cluster center, and m is the weight coefficient of membership; The Lagrangian function is constructed for the fuzzy C-means clustering objective function and the extreme value is obtained. Its expression is: Among them, u ij is the degree of membership between the i-th sample and the j-th cluster center, C is the number of clusters, k is an increasing variable from 1 to C, x i is the i-th data in the input load data sequence, v j is the jth cluster center, m is the weight coefficient of membership, and N is the number of loads; The data of N loads in t time periods are normalized to obtain the processed data set X, which is expressed as: Among them, x ij (i=1,2,…,N;j=1,2,…,t), is the value of the i-th load at the j-th time; Calculate the characteristic weight of the i-th load in the j-th column of the data set X, and its expression is: Among them, f ij is the characteristic weight of the i-th load in the j-th column of the data set X, x ij is the value of the i-th load at the j-th time, x kj is the value of the kth load at the jth time, and N is the number of loads; Calculate the entropy value of the jth (j=1,2,…,t) column, the expression is: Among them, H j is the entropy value of the jth column, N is the number of loads, f ij is the characteristic weight of the i-th load in the j-th column of the data set X; Calculate the entropy weight of the jth (j=1,2,…,t) column, and its expression is: Among them, ω 0j is the entropy weight of the jth column, H j is the entropy value of the jth column, H k is the kth value from k=1 to t; Consider eliminating the influence of noise and outliers on entropy and weights, and obtain the improved weights, whose calculation expression is: Among them, ω j is the improved weight, is the mean of the entropy values ​​of the jth column, p is the value of the load time t sampling points, ω 3j is the improved weight, ω 0j is the entropy weight of the jth column, H j is the entropy value of the jth column; The improved weight ω 3j The expression is: Among them, ω 3j is the improved weight, is the mean entropy value of the jth column, H j is the entropy value of the jth column, H k is the kth value from k=1 to t; The improved fuzzy C-means clustering objective function is obtained, and its expression is: Among them, J(U,V) is the distance from each sample to the cluster center, N is the number of loads, C is the number of clusters, and d ij is the distance between the i-th sample and the j-th cluster center, ξ i are coefficients, V = {v1, v2, …v C } represents the randomly set initial cluster center, U is the membership matrix of the initial setting, u ij is the membership degree between the i-th sample and the j-th cluster center, v j is the jth cluster center, m is the weight coefficient of membership, ω j is the improved weight.

3. A method for extracting power consumption characteristics of distribution network users according to claim 2, characterized in that: The star-sparrow algorithm is used to solve the initial cluster center, and its expression is: in, is the position of the ith star bird at the t+1th iteration, is the jth position of the ith star bird in the tth iteration, τ1, τ2, r1, r are random numbers between 0 and 1, is the j-th dimension mean of the t-th iteration, are the jth positions of three randomly selected individuals exploring high-quality solutions in the starfinch population, γ is the random number formed by the Levy flight function, μ is the coefficient, and U j and L j are the upper and lower limits of the j-th dimension position, T max is the limit of the number of iterations, and are the jth positions of the tth iteration of star bird A and star bird B respectively; in: Among them, r1, r2, r3 are random numbers that obey uniform distribution, τ3 is a random number between 0 and 1, τ4 is a random number that obeys normal distribution, and τ5 is a random number formed by Levy flight function; Get the new position where the solution is stored after the t+1th iteration of the i-th star bird, and its expression is: in, is the new location where the solution is stored after the t+1th iteration of the i-th star bird, is the position of the tth iteration of the i-th star bird, μ is the coefficient, is the optimal solution for the tth iteration, λ is the coefficient, r is a random number between 0 and 1, is the t-th iteration value of star A, is the t-th iteration value of star B, r1, r2, r3 are random numbers that obey uniform distribution, and l is the attenuation factor.

4. A method for extracting power consumption characteristics of distribution network users according to claim 3, characterized in that: Solving the optimal cluster center includes the following steps: If the starfinch remembers the position of the initial solution, it will find the iterative update position of the solution according to the following formula, which is expressed as: in, is the jth position of the ith star bird at the t+1th iteration, is the jth position of the ith star bird in the tth iteration, τ3 is a random number between 0 and 1, τ4 is a random number that obeys a normal distribution, r1 is a random number that obeys a uniform distribution, and r2 is a random number that obeys a uniform distribution. is the optimal position of the jth dimension in the tth iteration, is the position of the ith star bird at the first reference point in the tth iteration, is the jth position of the tth iteration of star C; If the starfinch does not remember the position of the initial solution, it will use the following formula to find the iterative update position of the solution, which is expressed as: in, is the jth position of the ith star bird at the t+1th iteration, is the jth position of the ith star bird in the tth iteration, τ3 is a random number between 0 and 1, τ4 is a random number that obeys a normal distribution, r1 is a random number that obeys a uniform distribution, and r2 is a random number that obeys a uniform distribution. is the optimal position of the jth dimension in the tth iteration, is the position of the ith star bird at the second reference point in the tth iteration, is the jth position of the tth iteration of star C; The optimal solution position is searched near the first reference point according to the memory. If the optimal solution is not found at the first reference point, the optimal solution is searched at the second reference point and other areas. The iterative update model for searching the optimal solution position near the first reference point according to the memory is: in, is the jth position of the ith star bird at the t+1th iteration, is the ith position of the tth iteration, To position Enter the value of the objective function of the fuzzy C-means clustering algorithm to find the cluster center, To position Enter the value of the objective function of the fuzzy C-means clustering algorithm to find the cluster center, is the position of the i-th starfinch at the first reference point in the t-th iteration; Find the optimal solution at the second reference point and other areas, and its iterative update model is: in, is the jth position of the ith star bird at the t+1th iteration, is the ith position of the tth iteration, To position Enter the value of the objective function of the fuzzy C-means clustering algorithm to find the cluster center, To position Enter the value of the objective function of the fuzzy C-means clustering algorithm to find the cluster center, is the position of the i-th starfinch at the second reference point in the t-th iteration.

5. A method for extracting power consumption characteristics of distribution network users according to claim 4, characterized in that: To obtain the optimal number of clusters, the Davidson-Bottlen objective function is established according to the degree of membership, and the number of clusters with the smallest mean of the maximum similarity of each class is determined as the optimal number of clusters, and its expression is: Among them, DBI is the mean of the maximum similarity of each class, C is the number of clusters, S i Cluster A i elements in to their cluster center c i The mean Euclidean distance of j Cluster A j elements in to their cluster center c j The mean Euclidean distance of i Cluster A i The number of elements in N j Cluster A j The number of elements in x i is the i-th data in the input load data sequence, is the similarity between the i-th category and the j-th category, is the maximum similarity between the i-th class and the j-th class.

6. A distribution network user power consumption feature clustering extraction system, characterized in that: include: A data acquisition module is used to collect data of N loads in t time periods; A clustering module is used to cluster the collected data of N loads in t time periods and construct an improved fuzzy C-means clustering objective function; An initial cluster center solving module is used to solve the initial cluster center according to the constructed improved fuzzy C-means clustering objective function; The optimal cluster center solving module is used to solve the optimal cluster center according to the constructed improved fuzzy C-means clustering objective function; The feature extraction module is used to obtain the optimal number of clusters based on the initial cluster center and the optimal cluster center to complete feature extraction.

7. A distribution network user power consumption feature clustering extraction system according to claim 6, characterized in that: The construction of the improved fuzzy C-means clustering objective function includes: Construct the fuzzy C-means clustering objective function, whose expression is: Among them, J(U,V) is the distance from each sample to the cluster center, N is the number of loads, C is the number of clusters, and x t is the i-th data in the input load data sequence, V = {v1, v2, …v C } represents the randomly set initial cluster center, U is the membership matrix of the initial setting, u ij is the membership degree between the i-th sample and the j-th cluster center, v j is the jth cluster center, and m is the weight coefficient of membership; The Lagrangian function is constructed for the fuzzy C-means clustering objective function and the extreme value is obtained. Its expression is: Among them, u ij is the degree of membership between the i-th sample and the j-th cluster center, C is the number of clusters, k is an increasing variable from 1 to C, x i is the i-th data in the input load data sequence, v j is the jth cluster center, m is the weight coefficient of membership, and N is the number of loads; The data of N loads in t time periods are normalized to obtain the processed data set X, which is expressed as: Among them, x ij (i=1,2,…,N;j=1,2,…,t), is the value of the i-th load at the j-th time; Calculate the characteristic weight of the i-th load in the j-th column of the data set X, and its expression is: Among them, f ij is the characteristic weight of the i-th load in the j-th column of the data set X, x ij is the value of the i-th load at the j-th time, x kj is the value of the kth load at the jth time, and N is the number of loads; Calculate the entropy value of the jth (j=1,2,…,t) column, the expression is: Among them, H j is the entropy value of the jth column, N is the number of loads, f ij is the characteristic weight of the i-th load in the j-th column of the data set X; Calculate the entropy weight of the jth (j=1,2,…,t) column, and its expression is: Among them, ω 0j is the entropy weight of the jth column, H j is the entropy value of the jth column, H k is the kth value from k=1 to t; Consider eliminating the influence of noise and outliers on entropy and weights, and obtain the improved weights, whose calculation expression is: Among them, ω j is the improved weight, is the mean of the entropy values ​​of the jth column, p is the value of the load time t sampling points, ω 3j is the improved weight, ω 0j is the entropy weight of the jth column, H j is the entropy value of the jth column; The improved weight ω 3j The expression is: Among them, ω 3j is the improved weight, is the mean entropy value of the jth column, H j is the entropy value of the jth column, H k is the kth value from k=1 to t; The improved fuzzy C-means clustering objective function is obtained, and its expression is: Among them, J(U,V) is the distance from each sample to the cluster center, N is the number of loads, C is the number of clusters, and d ij is the distance between the i-th sample and the j-th cluster center, ξ i are coefficients, V = {v1, v2, …v C } represents the randomly set initial cluster center, U is the membership matrix of the initial setting, u ij is the membership degree between the i-th sample and the j-th cluster center, v j is the jth cluster center, m is the weight coefficient of membership, ω j is the improved weight.

8. A distribution network user power consumption feature clustering extraction system according to claim 6, characterized in that: The star-sparrow algorithm is used to solve the initial cluster center, and its expression is: in, is the position of the ith star bird at the t+1th iteration, is the jth position of the ith star bird in the tth iteration, τ1, τ2, r1, r are random numbers between 0 and 1, is the j-th dimension mean of the t-th iteration, are the jth positions of three randomly selected individuals exploring high-quality solutions in the starfinch population, γ is the random number formed by the Levy flight function, μ is the coefficient, and U j and L j are the upper and lower limits of the j-th dimension position, T max is the limit of the number of iterations, and are the jth positions of the tth iteration of star bird A and star bird B respectively; in: Among them, r1, r2, r3 are random numbers that obey uniform distribution, τ3 is a random number between 0 and 1, τ4 is a random number that obeys normal distribution, and τ5 is a random number formed by Levy flight function; Get the new position where the solution is stored after the t+1th iteration of the i-th star bird, and its expression is: in, is the new location where the solution is stored after the t+1th iteration of the i-th star bird, is the position of the tth iteration of the i-th star bird, μ is the coefficient, is the optimal solution for the tth iteration, λ is the coefficient, r is a random number between 0 and 1, is the t-th iteration value of star A, is the t-th iteration value of star B, r1, r2, r3 are random numbers that obey uniform distribution, and l is the attenuation factor.

9. A distribution network user power consumption feature clustering extraction system according to claim 6, characterized in that: Solving the optimal cluster center includes the following steps: If the starfinch remembers the position of the initial solution, it will find the iterative update position of the solution according to the following formula, which is expressed as: in, is the jth position of the ith star bird at the t+1th iteration, is the jth position of the ith star bird in the tth iteration, τ3 is a random number between 0 and 1, τ4 is a random number that obeys a normal distribution, r1 is a random number that obeys a uniform distribution, and r2 is a random number that obeys a uniform distribution. is the optimal position of the jth dimension in the tth iteration, is the position of the ith star bird at the first reference point in the tth iteration, is the jth position of the tth iteration of star C; If the starfinch does not remember the position of the initial solution, it will use the following formula to find the iterative update position of the solution, which is expressed as: in, is the jth position of the ith star bird at the t+1th iteration, is the jth position of the ith star bird in the tth iteration, τ3 is a random number between 0 and 1, τ4 is a random number that obeys a normal distribution, r1 is a random number that obeys a uniform distribution, and r2 is a random number that obeys a uniform distribution. is the optimal position of the jth dimension in the tth iteration, is the position of the ith star bird at the second reference point in the tth iteration, is the jth position of the tth iteration of star C; The optimal solution position is searched near the first reference point according to the memory. If the optimal solution is not found at the first reference point, the optimal solution is searched at the second reference point and other areas. The iterative update model for searching the optimal solution position near the first reference point according to the memory is: in, is the jth position of the ith star bird at the t+1th iteration, is the ith position of the tth iteration, To position Enter the value of the objective function of the fuzzy C-means clustering algorithm to find the cluster center, To position Enter the value of the objective function of the fuzzy C-means clustering algorithm to find the cluster center, is the position of the i-th starfinch at the first reference point in the t-th iteration; Find the optimal solution at the second reference point and other areas, and its iterative update model is: in, is the jth position of the ith star bird at the t+1th iteration, is the ith position of the tth iteration, To position Enter the value of the objective function of the fuzzy C-means clustering algorithm to find the cluster center, To position Enter the value of the objective function of the fuzzy C-means clustering algorithm to find the cluster center, is the position of the i-th starfinch at the second reference point in the t-th iteration.

10. A distribution network user power consumption feature clustering extraction system according to claim 6, characterized in that: To obtain the optimal number of clusters, the Davidson-Bottlen objective function is established according to the degree of membership, and the number of clusters with the smallest mean of the maximum similarity of each class is determined as the optimal number of clusters, and its expression is: Among them, DBI is the mean of the maximum similarity of each class, C is the number of clusters, S i Cluster A i elements in to their cluster center c i The mean Euclidean distance of j Cluster A j elements in to their cluster center c j The mean Euclidean distance of i Cluster A i The number of elements in N j Cluster A j The number of elements in x i is the i-th data in the input load data sequence, is the similarity between the i-th category and the j-th category, is the maximum similarity between the i-th class and the j-th class.

Images