Method and device for analyzing different power utilization modes of industrial load based on curve clustering

Abstract

The invention discloses a curve clustering-based method and device for analyzing different power utilization modes of industrial loads, a terminal and a medium. The method comprises the following steps of: acquiring power utilization load curve data of each industrial user; preprocessing the load curve data to obtain a complete and available load curve data set; an improved WOA-FCM algorithm is adopted to carry out clustering on the load curve data set; according to a clustering result, distinguishing each industrial power consumption mode; wherein the improved WOA-FCM algorithm comprises the following steps: optimizing the position of each clustering center of a load curve through a WOA whale algorithm, determining the optimal clustering number of the load curve in combination with a set effectiveness index, and optimizing the membership matrix and the clustering center of the current load curve cluster by adopting a fuzzy C-means FCM clustering algorithm, so as to obtain the optimal clustering number of the load curve. And finally obtaining a load curve clustering result of each industrial user. According to the method, more accurate and efficient clustering analysis of different power utilization modes of the industrial load is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a method and device for analyzing different power consumption modes of industrial loads based on curve clustering. Background Art

[0002] With the development of industrial production and the growth of electricity demand, industrial users play a vital role in the power system. Clustering and analyzing the load curves of industrial power users will help understand and predict their electricity consumption behavior and provide important support for the operation and management of the power system. However, due to the complex time series characteristics and diversity of industrial electricity load, traditional clustering algorithms often find it difficult to explore its load characteristics; when deeply exploring the industrial energy consumption pattern, we face the challenges of many uncertain factors, including but not limited to the impact of sudden social events, the impact of extreme meteorological events, and the changes in electricity consumption habits caused by holidays and seasonal changes; for example, although the high temperature and severe cold in summer and winter respectively lead to increased cooling and heating demand, forming peak electricity consumption, the overall load in summer still usually exceeds that in winter; in this complex situation, it is particularly important to clearly distinguish and understand the energy consumption characteristics of different fields such as industry, commerce, agriculture and civil use and their characteristics of evolution over time to improve the scientific nature of energy management and planning.

[0003] With the rapid development of smart grid technology and the widespread deployment of smart metering equipment, the volume of energy data has expanded dramatically, showing distinct big data characteristics. In the face of this trend, in some related technologies, based on advanced data analysis technology, the DTW-Kmedoids algorithm is used to perform cluster analysis of industrial load curves, and the singular value decomposition method and the improved adaptive threshold method are used to achieve the fitting and similarity evaluation of load curves from multiple angles; although this method is outstanding in improving the accuracy of time series clustering, due to the complexity and diversity of industrial loads, the computational amount and computational cost of this method are greatly increased during the application process. In other related technologies, the K-means algorithm is used to cluster industrial load features. Although the K-means algorithm is widely used because of its simple implementation and high efficiency, it requires the number of clusters to be determined in advance, which limits its effectiveness in processing large-scale industrial load data sets. In addition, the traditional fuzzy C-means (FCM) clustering algorithm is very sensitive to the selection of initial cluster centers; since its optimization objective is a nonlinear convex function when applied to the industrial load clustering problem, if the selected cluster center does not conform to the actual situation, the algorithm may not be able to find the optimal solution in the entire industrial load search space, thus affecting its global performance. Summary of the invention

[0004] In order to address the deficiencies in the prior art, the present invention provides a method and device for analyzing different electricity consumption patterns of industrial loads based on curve clustering, aiming to solve the problems of the current industrial load clustering analysis that the number of clusters needs to be manually set and is sensitive to the initial cluster centers, thereby achieving more accurate and efficient clustering analysis of different electricity consumption patterns of industrial loads.

[0005] The present invention adopts the following technical solution.

[0006] In a first aspect, the present invention provides a method for analyzing different power consumption modes of industrial loads based on curve clustering, the method comprising:

[0007] Obtain the load curve data of electricity consumption of each industrial user;

[0008] Preprocessing the load curve data to obtain a complete and available load curve data set;

[0009] Using an improved WOA-FCM algorithm to cluster the load curve data set;

[0010] Based on the clustering results, the industrial electricity consumption patterns are distinguished;

[0011] Among them, the improved WOA-FCM algorithm includes: optimizing the positions of each cluster center of the load curve through the WOA whale algorithm, and determining the optimal number of clusters of the load curve in combination with the set effectiveness index, and then using the fuzzy C-means FCM clustering algorithm to optimize the membership matrix and cluster center of the current load curve clustering, and finally obtaining the load curve clustering results of each industrial user.

[0012] Optionally, the step of preprocessing the load data includes:

[0013] Using KNN nearest neighbor algorithm to supplement missing values ​​and replace abnormal values ​​of the load curve data;

[0014] Perform mean smoothing on each load curve data to obtain the overall shape trend of each curve;

[0015] The smoothed load curve data was normalized using z-score.

[0016] Optionally, the expression of the effectiveness indicator is as follows:

[0017]

[0018] Where: VSC(c) represents the effectiveness index under the current load curve; U = {u ij} is the membership matrix of the current load curve aggregation result, c means that the load curve data set X is divided into c non-intersecting subsets S 1 ,S2 ,...,S c , n means there are n industrial load curves in the load curve data set X, u ij represents the jth load curve x j For the i-th subset S i The degree of membership; A = {a 1 ,a 2 ,…,a c} represents the current cluster center matrix; Var(U,V,c), Sep(U,c) and Cop(U,c) respectively represent the compactness evaluation index, separation evaluation index and overlap comprehensive evaluation index under the current load curve clustering result; Var(U,A) N 、Sep(U,c) N and Cop(U,c) N They represent the measurement range of [0,1] obtained by dividing the compactness evaluation index, separation evaluation index and overlap evaluation index by the maximum index value in their corresponding cluster numbers.

[0019] Optionally, the calculation formula of the compactness evaluation index Var(U,V,c) is as follows:

[0020]

[0021] u ij ≥0(1≤i≤c,1≤j≤n)

[0022] In the formula, U={u ij} is the membership matrix of cxn, c means that the load curve data set X is divided into c non-intersecting subsets S 1 ,S 2 ,…,S c , n means there are n industrial load curves in the load curve data set X, u ij represents the jth load curve sample x j For the i-th subset S i The degree of membership; A = {a 1 ,a 2 ,…,a c} represents the matrix of c cluster centers obtained after clustering the current load; a i is the i-th cluster center in the cluster center matrix A; represents the cluster centroid of n load curves; β represents the compactness of the cluster; exp() represents the exponential function with the mathematical constant e as the base.

[0023] Optionally, the calculation formula of the separation evaluation index Sep(U,c) is as follows:

[0024]

[0025] In the formula, log means taking the logarithm.

[0026] Optionally, the calculation formula of the overlap evaluation index Cop(U,c) is as follows:

[0027]

[0028] In the formula, O hg represents the overlap between the distances of n industrial load curves and the hth and gth cluster centers; u hk represents the kth load curve sample x k The degree of membership to the h-th cluster center; u gk represents the kth load curve sample x k The degree of membership to the g-th cluster center of the division.

[0029] Optionally, the step of clustering the load curve data set using the improved WOA-FCM algorithm includes:

[0030] Step 1: Initialize the number of clusters c = 2 and set the maximum value of the number of clusters c max ;

[0031] Step 2: According to the currently updated cluster number c, randomly select c cluster centers a from the load curve data set X 1 (t),…,a c (t);

[0032] Step 3: Calculate the distance between each load curve sample and the currently updated cluster center a 1 (t),…,a c (t), and select the cluster center with the shortest average distance as the target prey X*;

[0033] Step 4: According to the current number of iterations, calculate the coefficient A* of simulating the whale's behavior of surrounding its prey; if |A*|≤1, the current cluster center a is updated by the bubble net hunting position update formula based on the target prey X*. 1 (t),…,a c (t) performing location updates respectively;

[0034] If |A*|>1, the current cluster center a is updated by randomly selecting a search object from the population to replace the target prey X*. 1 (t),…,a c (t) performing location updates respectively;

[0035] Step 5: Determine whether the current number of iterations reaches the set first number of iterations tmax ; If not, set t = t + 1 and return to step 3; If so, record the positions of the cluster centers after the final update under the current number of clusters and execute step 6;

[0036] Step 6: Based on the load curve clustering results of each cluster center finally updated in step 5, calculate the effectiveness index VSC(c) based on the current number of clusters;

[0037] Step 7: Determine whether the current number of clusters satisfies c≤c max ; If yes, set c = c + 1 and return to step 2, otherwise go to step 8;

[0038] Step 8: Determine the value of the effectiveness index VSC(c) corresponding to each cluster number, and take the cluster number with the smallest effectiveness index VSC(c) value as the optimal cluster number C. best ;

[0039] Step 9: According to the optimal number of clusters C recorded in step 5 best The positions of the cluster centers after the final update are obtained to obtain the current cluster center matrix And calculate the membership matrix U of the current load curve clustering;

[0040] Step 10: Calculate the FCM objective function of the current clustering result of the load curve data set X according to the current cluster center matrix A and the membership matrix U;

[0041] Step 11: Determine whether the current FCM objective function converges or whether the current number of iterations reaches the set second number of iterations T max ; If not, then update the cluster center matrix A and the membership matrix U through the membership and cluster center update function, set t = t + 1, and return to execute step 10; If so, output the industrial load curve clustering results corresponding to the final cluster center matrix A and the membership matrix U.

[0042] Optionally, the calculation formula of the coefficient A* for simulating the behavior of a whale surrounding its prey is as follows:

[0043] A*=2(a*)·ra*

[0044] Where a* decreases linearly from 2 to 0 with the number of iterations, and r is a random vector in [0,1].

[0045] Optionally, the bubble net hunting position update formula based on the target prey X* is as follows:

[0046]

[0047] Where a i(t+1) represents the i-th cluster center a at the current t-th iteration i (t) updated position; i = 1, 2…c; X * (t) represents the target prey at the current iteration t; b is the constant coefficient of the logarithmic spiral shape, l represents a random number in the interval [0,1], and e is a mathematical constant; D(t) = |C*·X * (t)-a i (t)| is the target prey X * (t) and the cluster center a as the search object i (t), C*=2r, r is a random vector in [0,1]; p is a random number in the interval [0,1].

[0048] Optionally, the prey search position update formula based on randomly selecting a search body in the population to replace the target prey X* is as follows:

[0049] a i (t+1)=X rand -A*·D'(t)

[0050] In the formula, a i (t+1) represents the i-th cluster center a at the current t-th iteration i (t) updated position; i = 1, 2…c; X rand Indicates that a cluster center randomly selected from the cluster center set is used as the current target prey; D'(t) = |C*·X rand -a i (t)|Target prey X rand and the cluster center a as the search object i (t); C*=2r, r is a random vector in [0,1].

[0051] Optionally, the expression of the FCM objective function is as follows:

[0052]

[0053] In the formula, represents the objective function value of the load curve data set X under the current cluster center matrix A and membership matrix U; u ij represents the jth element in the i-th category of data in the membership matrix; n represents that there are n industrial load curves in the load curve data set X; Indicates the current optimal number of clusters C best The matrix of cluster centers under ; λ is the Lagrangian operator; m represents softness; Indicates that the jth load curve sample in the load curve data set X is x jand the i-th cluster center a in the cluster center matrix A i The Euclidean distance.

[0054] Optionally, the expressions of the update functions of the membership degree and cluster center are as follows:

[0055]

[0056] In the formula, represents the jth load curve sample x after the t+1th iteration j The degree of membership to the i-th cluster center of the partition; a i (t) and a i (t+1) represents the cluster center a after the tth and t+1th iterations respectively. i location; k (t) represents the cluster center a after the tth iteration k The position of k = 1, 2…C best ; m indicates softness.

[0057] In a second aspect, the present invention provides a device for analyzing different power consumption modes of industrial loads based on curve clustering, and executes the steps of any method described in the first aspect of the present invention, the device comprising:

[0058] A data acquisition module is used to acquire load curve data of electricity consumption of each industrial user;

[0059] A preprocessing module, used for preprocessing the load curve data to obtain a complete and usable load curve data set;

[0060] A curve clustering module, used for clustering the load curve data set using an improved WOA-FCM algorithm;

[0061] The power consumption mode differentiation module is used to differentiate the power consumption modes of various industries according to the clustering results;

[0062] Among them, the improved WOA-FCM algorithm includes: optimizing the positions of each cluster center of the load curve through the WOA whale algorithm, and determining the optimal number of clusters of the load curve in combination with the set effectiveness index, and then using the fuzzy C-means FCM clustering algorithm to optimize the membership matrix and cluster center of the current load curve clustering, and finally obtaining the load curve clustering results of each industrial user.

[0063] In a third aspect, the present invention provides a terminal, including a processor and a storage medium;

[0064] The storage medium is used to store instructions;

[0065] The processor is used to operate according to the instructions to execute the steps of any one of the methods described in the first aspect of the present invention.

[0066] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any one of the methods described in the first aspect of the present invention.

[0067] The beneficial effects of the present invention are as follows:

[0068] The present invention ensures the quality of industrial load curve data through strict data preprocessing, and proposes an improved WOA-FCM algorithm to realize clustering division of different industrial user point patterns, which significantly enhances the ability of clustering analysis of industrial load curves.

[0069] In view of the shortcomings of traditional industrial load clustering algorithms, such as the need to manually set the number of clusters and being sensitive to cluster centers, the present invention proposes a new clustering effectiveness index VSC to automatically obtain the number of clusters. The selection of the number of clusters is more reasonable, and the complex characteristics of the industrial load curve can be quickly and efficiently revealed without adding additional computational burden. Performance tests are carried out on different data sets, and experiments show that the method of the present invention achieves good results in determining the number of clusters and selecting the center point.

[0070] In order to effectively solve the problem of falling into local optimality caused by random selection of cluster centers due to the complexity of industrial load, the present invention adopts the WOA whale algorithm with strong optimization ability and fast convergence speed to optimize the selection of cluster centers in the clustering process of the FCM algorithm. It can not only effectively deal with data noise and outlier problems, but also intelligently identify and divide unique pattern categories in industrial load data without presetting the number of clusters, which greatly improves the ability to process complex data sets; thereby achieving a significant improvement in the accuracy of industrial load monitoring and the sophistication of management without sacrificing computing efficiency.

[0071] After experimental verification, compared with the traditional FCM industrial load clustering method, the method proposed in the present invention has significant advantages in performance evaluation and iteration time, has better clustering ability and robustness, can accurately and effectively cluster power loads, and provides theoretical guidance and data support for the extraction of steady-state features and component identification of industrial loads. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 It is a schematic diagram of the steps of the method for analyzing different power consumption modes of industrial loads based on curve clustering in the present invention;

[0073] Figure 2 It is a flowchart of the improved WOA-FCM algorithm in the present invention;

[0074] Figure 3A schematic diagram of the relationship between the number of clusters and the VSC effectiveness index value in the present invention;

[0075] Figure 4 It is a schematic diagram of the clustering result of the load curve using the traditional FCM algorithm in the simulation experiment of the present invention;

[0076] Figure 5 It is a schematic diagram of the clustering result of the load curve using the improved WOA-FCM algorithm in the simulation experiment of the present invention;

[0077] Figure 6 This is a structural principle block diagram of the device for analyzing different power consumption modes of industrial loads based on curve clustering in the present invention. DETAILED DESCRIPTION

[0078] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. The embodiments described in the present invention are only part of the embodiments of the present invention, not all of the embodiments. Based on the spirit of the present invention, other embodiments obtained by ordinary technicians in this field without creative work are all within the scope of protection of the present invention.

[0079] Embodiment 1:

[0080] Reference Figure 1 The embodiment of the present invention provides a method for analyzing different power consumption modes of industrial loads based on curve clustering, which specifically includes the following steps:

[0081] Step S1, obtaining load curve data of electricity consumption of each industrial user;

[0082] Step S2: preprocessing the load curve data to obtain a complete and usable load curve data set;

[0083] Step S3, clustering the load curve data set using an improved WOA-FCM algorithm;

[0084] Step S4: Differentiate each industrial electricity consumption mode according to the clustering result;

[0085] As an embodiment of the present invention, the improved WOA-FCM algorithm in step S3 includes: optimizing the positions of each cluster center of the load curve through the WOA whale algorithm, and determining the optimal number of clusters of the load curve in combination with the set effectiveness index, and then using the fuzzy C-means FCM clustering algorithm to optimize the membership matrix and cluster center of the current load curve cluster, and finally obtaining the load curve clustering results of each industrial user.

[0086] In this embodiment, the preprocessing process of step S2 includes:

[0087] Step 2.1, using the KNN nearest neighbor algorithm to supplement missing values ​​and replace abnormal values ​​in the load curve data;

[0088] Among them, the main principle of the KNN nearest neighbor algorithm is that when there is a new unknown sample that needs to be predicted, KNN will find the K samples closest to it in the existing training data set, and then determine the prediction result of the new sample based on the majority category (classification task) or average value (regression task) of these neighbors. The KNN nearest neighbor algorithm is often used to supplement or replace missing values ​​or outliers in the data, which will not be elaborated here.

[0089] Step 2.2, perform mean smoothing processing on each load curve data to obtain the overall shape trend of each curve;

[0090] Since the data points collected every 15 minutes by the industrial load power data acquisition system are too small in granularity and the data dimension is too high, this leads to an increase in the complexity of the algorithm; in the actual operation process, due to various reasons such as communication interruption, hardware and software failure, and signal interference, noise may appear in the load data, thereby causing significant fluctuations in the load curve, further affecting the accuracy of clustering. In response to this situation, the traditional method is to use Gaussian filtering to smooth the original load curve and then perform clustering analysis, but this method will lead to shortcomings such as inaccurate curve fitting and large errors. The present invention adopts mean smoothing technology to make the power load curve smoother, thereby more effectively highlighting the shape and overall trend of the curve; the steps of mean smoothing specifically include:

[0091] 2.2.1: Input a set of sampling points, smoothing times and smoothing threshold tv;

[0092] 2.2.2: The absolute value of the difference between samples is expressed as the difference between two adjacent sample points, denoted by dv;

[0093] 2.2.3: Estimate the smoothing threshold tv of the interpolated value dv; if the value is lower than this value, add it to the average of the two variables, otherwise the value of the sampling point remains unchanged;

[0094] 2.2.4: Repeat step 2.2.3 according to the smooth frequency ο;

[0095] 2.2.5: After smoothing the data, the corresponding curve set is obtained and the curve is output.

[0096] Step 2.3: Perform z-score normalization on the smoothed load curve data.

[0097] Furthermore, the core idea of ​​z-score normalization is to normalize the original data set so that its mean is set to 0 and its variance is set to 1. This method is more in line with the characteristics of the original data distribution being close to the Gaussian distribution. The normalization formula of z-score is as follows:

[0098]

[0099] Among them, X represents the original data sequence, μ represents the average value of the original data, and σ is the standard deviation of the original data. novm Used to represent a standardized data sequence.

[0100] Linear function normalization is to perform a linear transformation on the original data so that it is within the range of [0,1]. Its expression is:

[0101]

[0102] Among them, X represents the original data sequence, X min Represents the minimum value in the original data, X max Represents the maximum value in the original data.

[0103] When using traditional methods to process different types of load data, the data types are diverse and often cannot meet the pre-defined conditions of Gaussian distribution; therefore, different methods are needed to perform standardization processing to obtain statistics that meet specific requirements.

[0104] In step S3 of this embodiment, the WOA-FCM algorithm is improved to achieve efficient clustering of industrial load curves, and the clustering speed is improved without increasing the amount of calculation. Among them, the flexible clustering algorithm in the traditional industrial load clustering problem is improved, and a cluster number selection method based on clustering effectiveness index is proposed to fully reveal the high-dimensional properties of the industrial load curve; the specific contents include:

[0105] In order to further ensure that the number of clusters is optimal, the present invention redefines the cluster validity index function, and well completes the determination of the number of clusters in cluster analysis.

[0106] In view of this deficiency of the fuzzy clustering effectiveness index, a new clustering effectiveness index VSC is redefined by defining the dispersion, compactness and overlap between data. Among them, the compactness is mainly represented by the intra-class distance, the separation is composed of the minimum membership, and the overlap is completed by the combination of membership and partition entropy. The smaller the compactness and overlap, the greater the separation, and the more significant the effect of data clustering. This fully demonstrates the hiding of various information in the data and improves the level of determination of the number of clusters in the clustered data; the VSC index provided by the present invention can effectively eliminate the influence of noise and overlap on the clustering results, and can more accurately determine the optimal number of clusters.

[0107] (1) Firmness

[0108] Var(U,V,c) is defined as the weighted square deviation within the cluster, which can reflect the compactness of the clustering results. The calculation formula of the compactness evaluation index Var(U,V,c) is as follows:

[0109]

[0110]

[0111] u ij ≥0(1≤i≤c,1≤j≤n)

[0112] In the formula, U={u ij} is the membership matrix of cxn, c means that the load curve data set X is divided into c non-intersecting subsets S 1 ,S 2 ,…,S c , n means there are n industrial load curves in the load curve data set X, u ij represents the jth load curve sample x j For the i-th subset S i The degree of membership; A = {a 1 ,a 2 ,…,a c} represents the matrix of c cluster centers obtained after clustering the current load; a i is the i-th cluster center in the cluster center matrix A; represents the cluster centroid of n load curves; β represents the compactness of the cluster; exp() represents the exponential function with the mathematical constant e as the base.

[0113] It should be noted that the main purpose of using compactness in this embodiment is to use an exponential function to measure the distance between each load curve and the cluster centroid; the exponential function can effectively reduce the interference of noise and outliers, suppress the numerical deviation caused by noise points and outliers, and greatly improve the robustness of noise and outlier data; the smaller the compactness, the better the effect of fuzzy partitioning of industrial loads.

[0114] (2) Separation

[0115] The calculation formula of the separation evaluation index Sep(U,c) is as follows:

[0116]

[0117] In the formula, log means taking the logarithm.

[0118] It should be noted that the introduction of the separation degree in this embodiment well describes the fuzziness and uncertainty of the industrial load curve division results. The larger the value, the smaller the uncertainty and the more reliable the classification effect.

[0119] (3) Overlap

[0120] The traditional effective function is mainly evaluated by the compactness within the class and the separation between the classes, while the distance-based evaluation standard cannot effectively solve the classification of overlapping classes and has the disadvantage of poor sensitivity to noise points. This embodiment evaluates by the following overlap, and the calculation formula of the overlap evaluation index Cop(U,c) is as follows:

[0121]

[0122] In the formula, O hg represents the overlap between the distances of n industrial load curves and the hth and gth cluster centers; u hk represents the kth load curve sample x k The degree of membership to the h-th cluster center; u gk represents the kth load curve sample x k The degree of membership to the g-th cluster center of the division.

[0123] It should be noted that in this embodiment, the measurement boundary of the degree of overlap between different classes is mainly represented by the degree of overlap. The clearer the division, the greater the degree of membership, which makes the class division and clustering results more specific and clear; therefore, it can be concluded that the smaller the degree of overlap, the clearer the division between the two classes, and the smaller the overlap, the more significant the clustering division result.

[0124] Furthermore, the dimensions of compactness, separation, and overlap are very different, so they need to be normalized. The index value of each cluster number is divided by the largest index value in the corresponding cluster number to obtain a measurement range of [0,1]. The formula is as follows:

[0125]

[0126] Then the expression of effectiveness index is obtained as follows:

[0127]

[0128] Where: VSC(c) represents the effectiveness index under the current load curve; U = {u ij} is the membership matrix of the current load curve aggregation result, c means that the load curve data set X is divided into c non-intersecting subsets S 1 ,S 2 ,...,S c , n means there are n industrial load curves in the load curve data set X, u ij represents the jth load curve x j For the i-th subset S i The degree of membership; A = {a 1 ,a 2 ,…,a c} represents the current cluster center matrix; Var(U,V,c), Sep(U,c) and Cop(U,c) respectively represent the compactness evaluation index, separation evaluation index and overlap comprehensive evaluation index under the current load curve clustering result; Var(U,A) N 、Sep(U,c) N and Cop(U,c) N Respectively represent the compactness evaluation index, separation evaluation index and overlap evaluation index divided by the maximum index value in each corresponding cluster number to obtain the measurement range of [0,1]. Among them, Var(U,A) obtained by compactness N The smaller it is, the closer the distance within the class is; Sep(U,c) obtained by separation N The larger the value, the more dispersed the distance between classes; Cop(U,c) obtained by the overlap angle N The smaller it is, the smaller the overlap is; therefore, the smaller VSC(c) is, the clearer the data samples are in the cluster and the better the clustering effect is.

[0129] Furthermore, in order to address the shortcomings of traditional industrial load clustering algorithms that are sensitive to cluster centers, the embodiments of the present invention propose a WOA whale optimization algorithm to effectively improve the optimization of the initial center. In order to address the shortcomings of the inability to reasonably divide data and the lack of interpretability of clustering results due to empirical settings, a new validity indicator VSC is incorporated to enable the algorithm to automatically obtain the number of clusters during operation.

[0130] Reference Figure 2 As shown, the step of clustering the load curve data set using the improved WOA-FCM algorithm in step S3 includes:

[0131] Step 1: Initialize the number of clusters c = 2 and set the maximum value of the number of clusters c max ;

[0132] Step 2: According to the currently updated cluster number c, randomly select c cluster centers a from the load curve data set X 1 (t),…,a c (t);

[0133] Step 3: Calculate the distance between each load curve sample and the currently updated cluster center a 1 (t),…,a c (t), and select the cluster center with the shortest average distance as the target prey X*; the calculation formula of the average distance is as follows:

[0134]

[0135] In the formula, a νg [a i (t)] represents the distance between each load curve sample and the currently updated cluster center a i (t), i = 1, 2…c; d(x j ,a i (t)) represents the load curve sample x j To cluster center a i (t) is the Euclidean distance.

[0136] Step 4: According to the current number of iterations, calculate the coefficient A* of simulating the whale's behavior of surrounding its prey; if |A*|≤1, the current cluster center a is updated by the bubble net hunting position update formula based on the target prey X*. 1 (t),…,a c (t) performing location updates respectively;

[0137] If |A*|>1, the current cluster center a is updated by randomly selecting a search object from the population to replace the target prey X*. 1 (t),…,ac (t) Perform location update respectively.

[0138] It is further explained that the whales of the WOA algorithm involved in step 4 need to determine the position of the prey in order to surround and capture the prey during the hunting process. However, the position of the prey in the search space is usually unknown. The WOA algorithm assumes that the best candidate solution in terms of fitness in the current population is the prey position or the position that is close to the target prey, and updates its own position based on the current best candidate solution. Based on the predation behavior of humpback whales hunting in bubble nets, the WOA algorithm designs two strategies: contraction and encirclement and spiral update of position; among them, contraction and encirclement is achieved by reducing the convergence factor a* value, and the value of A* in this process is a random vector in the interval [-a*, a*]. Specifically, the calculation formula of the coefficient A* that simulates the behavior of whales surrounding prey in an embodiment of the present invention is as follows:

[0139] A*=2(a*)·ra*

[0140] Where a* decreases linearly from 2 to 0 with the number of iterations, and r is a random vector in [0,1].

[0141] If |A*|≤1, the individual whale moves closer to the target prey position from the original position after the updated position to complete the contraction and encirclement of the prey. In the spiral update position, the spiral motion state of the whale is simulated to construct a spiral. Since the whale shrinks the encirclement while moving along the spiral path, in order to simulate this synchronization process, the WOA algorithm selects the same probability p for contraction and encirclement and spiral position update. Therefore, the bubble net hunting position update formula based on the target prey X* in this embodiment is obtained as follows:

[0142]

[0143] Where a i (t+1) represents the i-th cluster center a at the current t-th iteration i (t) updated position; i = 1, 2…c; X * (t) represents the target prey at the current iteration t; b is the constant coefficient of the logarithmic spiral shape, l represents a random number in the interval [0,1], and e is a mathematical constant; D(t) = |C*·X * (t)-a i (t)| is the target prey X * (t) and the cluster center a as the search object i (t), C*=2r, r is a random vector in [0,1]; p is a random number in the interval [0,1].

[0144] When |A*|>1, the WOA algorithm no longer selects the target prey to update its own position, but randomly selects a search body in the group to replace the target prey. In order to find prey, the individual whale is forced to deviate from the original target prey, so as to enhance the global exploration ability of the algorithm; therefore, the prey search position update formula based on randomly selecting a search body in the group to replace the target prey X* in this embodiment is as follows:

[0145] a i (t+1)=X rand -A*·D'(t)

[0146] In the formula, a i (t+1) represents the i-th cluster center a at the current t-th iteration i (t) updated position; i = 1, 2…c; X rand Indicates that a cluster center randomly selected from the cluster center set is used as the current target prey; D'(t) = |C*·X rand -a i (t)|Target prey X rand and the cluster center a as the search object i (t); C*=2r, r is a random vector in [0,1].

[0147] Step 5: Determine whether the current number of iterations reaches the set first number of iterations t max ; If not, set t = t + 1 and return to step 3; If so, record the positions of the cluster centers after the final update under the current number of clusters and execute step 6;

[0148] Step 6: Based on the load curve clustering results of each cluster center finally updated in step 5, calculate the effectiveness index VSC(c) based on the current number of clusters;

[0149] Step 7: Determine whether the current number of clusters satisfies c≤c max ; If yes, set c = c + 1 and return to step 2, otherwise go to step 8;

[0150] Step 8: Determine the value of the effectiveness index VSC(c) corresponding to each cluster number, and take the cluster number with the smallest effectiveness index VSC(c) value as the optimal cluster number C. best ; that is, 2≤c≤c max The minimum value in VSC min = min{VSC(c)}, get the optimal number of clusters C best =VSC min .

[0151] Step 9: According to the optimal number of clusters C recorded in step 5 bestThe positions of the cluster centers after the final update are obtained to obtain the current cluster center matrix And calculate the membership matrix U of the current load curve clustering;

[0152] Step 10: Calculate the FCM objective function of the current clustering result of the load curve data set X according to the current cluster center matrix A and the membership matrix U;

[0153] As an embodiment of the present invention, the expression of the FCM objective function is as follows:

[0154]

[0155] In the formula, represents the objective function value of the load curve data set X under the current cluster center matrix A and membership matrix U; u ij represents the jth element in the i-th category of data in the membership matrix; n represents that there are n industrial load curves in the load curve data set X; Indicates the current optimal number of clusters C best The matrix of cluster centers under ; λ is the Lagrangian operator; m represents softness, and the value of m determines the fuzziness of the clustering process; Indicates that the jth load curve sample in the load curve data set X is x j and the i-th cluster center a in the cluster center matrix A i The Euclidean distance.

[0156] Step 11: Determine whether the current FCM objective function converges or whether the current number of iterations reaches the set second number of iterations T max ; If not, then update the cluster center matrix A and the membership matrix U through the membership and cluster center update function, set t = t + 1, and return to execute step 10; If so, output the industrial load curve clustering results corresponding to the final cluster center matrix A and the membership matrix U.

[0157] As an embodiment of the present invention, the expressions of the update functions of the membership degree and the cluster center are respectively as follows:

[0158]

[0159] In the formula, represents the jth load curve sample x after the t+1th iteration j The degree of membership to the i-th cluster center of the partition; a i (t) and a i (t+1) represents the cluster center a after the tth and t+1th iterations respectively. i location; k (t) represents the cluster center a after the tth iterationk The position of k = 1, 2…C best ; m indicates softness.

[0160] In order to verify the influence of the cluster center determined by the WOA whale algorithm adopted in the present invention on the clustering results, this embodiment uses the public data set Iris containing 150 data and a dimension of 4 to conduct experiments. The effect of the selection of the center point on the clustering results is shown in Table 1.

[0161] Table 1 Iris cluster center comparison

[0162]

[0163] As can be seen from Table 1, the WOA whale algorithm used in the present invention to optimize the cluster center can improve the problem of low accuracy of clustering results caused by unreasonable selection of cluster centers, and effectively reduce the number of algorithm iterations in the clustering process.

[0164] In addition, this embodiment uses the improved WOA-FCM algorithm of the present invention to conduct simulation experiments and comparisons with the traditional FCM algorithm; the experimental data is analyzed by taking the industrial load data collected in a certain area in one day as an example, with a total of 148 user data, and the collection time is once per minute. The clustering effectiveness index VSC defined in the present invention is used to dynamically optimize the optimal number of clusters for load data, and the change trend of the index function is obtained as follows: Figure 3 As shown by Figure 3 It can be seen that the effectiveness index VSC reaches the minimum value when c=5, so the number of load data clusters is set to 5. The load curve data are clustered using the traditional FCM algorithm and the improved WOA-FCM algorithm. Figure 4 and Figure 5 They are the clustering results of load data in the traditional FCM algorithm and the adaptive WOA-FCM algorithm.

[0165] As shown in Table 2, this embodiment uses the DB index and the Dunn index to evaluate the clustering effects of the two methods. Among them, the DB index (Davies-Bouldin Index) is a quality evaluation index for measuring clustering effects. The index works by calculating the ratio of the compactness within each cluster to the separation between its nearest neighboring clusters. The smaller the value, the better the clustering effect. The Dunn index (Dunn Index, DI) is an indicator for evaluating clustering effects. The index measures clustering quality by calculating the ratio of the smallest internal distance among all clusters to the maximum distance between all cluster pairs. The larger the value, the better the clustering effect.

[0166] Table 2 Comparison of traditional FCM algorithm and improved WOA-FCM algorithm

[0167]

[0168] From the data analysis in Table 2, it can be seen that the clustering effect of the method of the present invention is relatively ideal, which is manifested in a lower DB value and a higher Dunn index. That is, compared with the traditional FCM algorithm, the distances of most samples from their respective cluster centers are reduced to varying degrees in the improved WOA-FCM algorithm of the present invention. At the same time, the load curve shows relatively regular local peaks and troughs in each part. The method of the present invention has better clustering performance and smaller similarity between classes. Experiments show that the improved WOA-FCM algorithm can obtain clustering results with a higher degree of intra-class aggregation, better inter-class division effect, and better clustering ability and robustness.

[0169] The beneficial effects of the present invention are as follows:

[0170] The present invention ensures the quality of industrial load curve data through strict data preprocessing, and proposes an improved WOA-FCM algorithm to realize clustering division of different industrial user point patterns, which significantly enhances the ability of clustering analysis of industrial load curves.

[0171] In view of the shortcomings of traditional industrial load clustering algorithms, such as the need to manually set the number of clusters and being sensitive to cluster centers, the present invention proposes a new clustering effectiveness index VSC to automatically obtain the number of clusters. The selection of the number of clusters is more reasonable, and the complex characteristics of the industrial load curve can be quickly and efficiently revealed without adding additional computational burden. Performance tests are carried out on different data sets, and experiments show that the method of the present invention achieves good results in determining the number of clusters and selecting the center point.

[0172] In order to effectively solve the problem of falling into local optimality caused by random selection of cluster centers due to the complexity of industrial load, the present invention adopts the WOA whale algorithm with strong optimization ability and fast convergence speed to optimize the selection of cluster centers in the clustering process of the FCM algorithm. It can not only effectively deal with data noise and outlier problems, but also intelligently identify and divide unique pattern categories in industrial load data without presetting the number of clusters, which greatly improves the ability to process complex data sets; thereby achieving a significant improvement in the accuracy of industrial load monitoring and the sophistication of management without sacrificing computing efficiency.

[0173] After experimental verification, compared with the traditional FCM industrial load clustering method, the method proposed in the present invention has significant advantages in performance evaluation and iteration time, has better clustering ability and robustness, can accurately and effectively cluster power loads, and provides theoretical guidance and data support for the extraction of steady-state features and component identification of industrial loads.

[0174] Embodiment 2:

[0175] like Figure 6 As shown, the present invention provides a device for analyzing different power consumption modes of industrial loads based on curve clustering, the device is used to implement the steps of the method in the above embodiment 1, and the device specifically includes:

[0176] A data acquisition module is used to acquire load curve data of electricity consumption of each industrial user;

[0177] A preprocessing module, used for preprocessing the load curve data to obtain a complete and usable load curve data set;

[0178] A curve clustering module, used for clustering the load curve data set using an improved WOA-FCM algorithm;

[0179] The power consumption mode differentiation module is used to differentiate the power consumption modes of various industries according to the clustering results;

[0180] As an embodiment of the present invention, the improved WOA-FCM algorithm in the curve clustering module includes: optimizing the positions of each cluster center of the load curve through the WOA whale algorithm, and determining the optimal number of clusters of the load curve in combination with the set effectiveness index, and then using the fuzzy C-means FCM clustering algorithm to optimize the membership matrix and cluster center of the current load curve clustering, and finally obtaining the load curve clustering results of each industrial user.

[0181] The device for analyzing different power consumption patterns of industrial loads based on curve clustering provided in an embodiment of the present invention and the method for analyzing different power consumption patterns of industrial loads based on curve clustering provided in Example 1 are based on the same technical concept, and can produce the beneficial effects described in Example 1. For the contents not described in detail in this embodiment, please refer to Example 1.

[0182] Embodiment three:

[0183] A terminal provided by an embodiment of the present invention includes a processor and a storage medium;

[0184] The storage medium is used to store instructions;

[0185] The processor is used to operate according to the instruction to execute the steps of the method according to any one of the embodiments.

[0186] Embodiment 4:

[0187] An embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the steps of any method described in Embodiment 1 are implemented.

[0188] The present disclosure may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0189] A computer-readable storage medium may be a tangible device that can hold and store instructions used by an instruction execution device. A computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples of computer-readable storage media (a non-exhaustive list) include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination of the foregoing. As used herein, a computer-readable storage medium is not to be interpreted as a transient signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through a wire.

[0190] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0191] The computer program instructions for performing the operation of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. Computer-readable program instructions may be executed completely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may be customized by utilizing the state information of the computer-readable program instructions, and the electronic circuit may execute the computer-readable program instructions, thereby realizing various aspects of the present disclosure.

[0192] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for analyzing different power consumption modes of industrial loads based on curve clustering, characterized in that: Methods include: Obtain the load curve data of electricity consumption of each industrial user; Preprocessing the load curve data to obtain a complete and available load curve data set; Using an improved WOA-FCM algorithm to cluster the load curve data set; Based on the clustering results, the industrial electricity consumption patterns are distinguished; Among them, the improved WOA-FCM algorithm includes: optimizing the positions of each cluster center of the load curve through the WOA whale algorithm, and determining the optimal number of clusters of the load curve in combination with the set effectiveness index, and then using the fuzzy C-means FCM clustering algorithm to optimize the membership matrix and cluster center of the current load curve clustering, and finally obtaining the load curve clustering results of each industrial user.

2. The method for analyzing different power consumption modes of industrial loads based on curve clustering according to claim 1 is characterized in that: The step of preprocessing the load data includes: Using KNN nearest neighbor algorithm to supplement missing values ​​and replace abnormal values ​​of the load curve data; Perform mean smoothing on each load curve data to obtain the overall shape trend of each curve; The smoothed load curve data was normalized using z-score.

3. The method for analyzing different power consumption modes of industrial loads based on curve clustering according to claim 1 is characterized in that: The expression of the effectiveness index is as follows: Where: VSC(c) represents the effectiveness index under the current load curve; U = {u ij } is the membership matrix of the current load curve aggregation result, c means that the load curve data set X is divided into c non-intersecting subsets S1, S2, …, S c , n means there are n industrial load curves in the load curve data set X, u ij represents the jth load curve x j For the i-th subset S i The degree of membership; A = {a1, a2, …, a c } represents the current cluster center matrix; Var(U,V,c), Sep(U,c) and Cop(U,c) respectively represent the compactness evaluation index, separation evaluation index and overlap comprehensive evaluation index under the current load curve clustering result; Var(U,A) N 、Sep(U,c) N and Cop(U,c) N They represent the measurement range of [0,1] obtained by dividing the compactness evaluation index, separation evaluation index and overlap evaluation index by the maximum index value in their corresponding cluster numbers.

4. The method for analyzing different power consumption modes of industrial loads based on curve clustering according to claim 3 is characterized in that: The calculation formula of the compactness evaluation index Var(U,V,c) is as follows: in ij ≥0(1≤i≤c,1≤j≤n) In the formula, U={u ij } is the membership matrix of cxn, c means that the load curve data set X is divided into c non-intersecting subsets S1, S2, …, S c , n means there are n industrial load curves in the load curve data set X, u ij represents the jth load curve sample x j For the i-th subset S i The degree of membership; A = {a1, a2, ..., a c } represents the matrix of c cluster centers obtained after clustering the current load; a i is the i-th cluster center in the cluster center matrix A; represents the cluster centroid of n load curves; β represents the compactness of the cluster; exp() represents the exponential function with the mathematical constant e as the base.

5. The method for analyzing different power consumption modes of industrial loads based on curve clustering according to claim 4 is characterized in that: The calculation formula of the separation evaluation index Sep(U,c) is as follows: In the formula, log means taking the logarithm.

6. The method for analyzing different power consumption modes of industrial loads based on curve clustering according to claim 4 is characterized in that: The calculation formula of the overlap evaluation index Cop(U,c) is as follows: In the formula, O hg represents the overlap between the distances of n industrial load curves and the hth and gth cluster centers; u hk represents the kth load curve sample x k The degree of membership to the h-th cluster center; u gk represents the kth load curve sample x k The degree of membership to the g-th cluster center of the division.

7. The method for analyzing different power consumption modes of industrial loads based on curve clustering according to claim 1 is characterized in that: The steps of clustering the load curve data set using the improved WOA-FCM algorithm include: Step 1: Initialize the number of clusters c = 2 and set the maximum value of the number of clusters c max ; Step 2: According to the currently updated cluster number c, randomly select c cluster centers a1(t),…,a1(t) from the load curve data set X. c (t); Step 3: Calculate the distance between each load curve sample and the currently updated cluster center a1(t),…,a c (t), and select the cluster center with the shortest average distance as the target prey X*; Step 4: According to the current number of iterations, calculate the coefficient A* of simulating the whale's behavior of surrounding its prey; if |A*|≤1, the current cluster centers a1(t), ..., a1(t) are updated using the bubble net hunting position update formula based on the target prey X*. c (t) performing location updates respectively; If |A*|>1, the prey search position update formula is updated based on the current cluster center a1(t), ..., a by randomly selecting a search body from the group to replace the target prey X*. c (t) performing location updates respectively; Step 5: Determine whether the current number of iterations reaches the set first number of iterations t max ; If not, set t = t + 1 and return to step 3; If so, record the positions of the cluster centers after the final update under the current number of clusters and execute step 6; Step 6: Based on the load curve clustering results of each cluster center finally updated in step 5, calculate the effectiveness index VSC(c) based on the current number of clusters; Step 7: Determine whether the current number of clusters satisfies c≤c max ; If yes, set c = c + 1 and return to step 2, otherwise go to step 8; Step 8: Determine the value of the effectiveness index VSC(c) corresponding to each cluster number, and take the cluster number with the smallest effectiveness index VSC(c) value as the optimal cluster number C. best ; Step 9: According to the optimal number of clusters C recorded in step 5 best The positions of the cluster centers after the final update are obtained to obtain the current cluster center matrix And calculate the membership matrix U of the current load curve clustering; Step 10: Calculate the FCM objective function of the current clustering result of the load curve data set X according to the current cluster center matrix A and the membership matrix U; Step 11: Determine whether the current FCM objective function converges or whether the current number of iterations reaches the set second number of iterations T max ; If not, then update the cluster center matrix A and the membership matrix U through the membership and cluster center update function, set t = t + 1, and return to execute step 10; If so, output the industrial load curve clustering results corresponding to the final cluster center matrix A and the membership matrix U.

8. The method for analyzing different power consumption modes of industrial loads based on curve clustering according to claim 7 is characterized in that: The calculation formula of the coefficient A* for simulating the whale's surrounding prey behavior is as follows: A*=2(a*)·ra* Where a* decreases linearly from 2 to 0 with the number of iterations, and r is a random vector in [0,1].

9. The method for analyzing different power consumption modes of industrial loads based on curve clustering according to claim 8 is characterized in that: The bubble net hunting position update formula based on the target prey X* is as follows: Where a i (t+1) represents the i-th cluster center a at the current t-th iteration i (t) updated position; i = 1, 2…c; X * (t) represents the target prey at the current iteration t; b is the constant coefficient of the logarithmic spiral shape, l represents a random number in the interval [0,1], and e is a mathematical constant; D(t) = |C*·X * (t)-a i (t)| is the target prey X * (t) and the cluster center a as the search object i (t), C*=2r, r is a random vector in [0,1]; p is a random number in the interval [0,1].

10. The method for analyzing different power consumption modes of industrial loads based on curve clustering according to claim 8, characterized in that: The prey search position update formula based on randomly selecting a search body from the population to replace the target prey X* is as follows: a i (t+1)=X rand -A*·D'(t) In the formula, a i (t+1) represents the i-th cluster center a at the current t-th iteration i (t) updated position; i = 1, 2…c; X rand Indicates that a cluster center randomly selected from the cluster center set is used as the current target prey; D'(t) = |C*·X rand -a i (t)|Target prey X rand and the cluster center a as the search object i (t); C*=2r, r is a random vector in [0,1].

11. The method for analyzing different power consumption modes of industrial loads based on curve clustering according to claim 7, characterized in that: The expression of the FCM objective function is as follows: In the formula, represents the objective function value of the load curve data set X under the current cluster center matrix A and membership matrix U; u ij represents the jth element in the i-th category of data in the membership matrix; n represents that there are n industrial load curves in the load curve data set X; Indicates the current optimal number of clusters C best The matrix of cluster centers under ; λ is the Lagrangian operator; m represents softness; Indicates that the jth load curve sample in the load curve data set X is x j and the i-th cluster center a in the cluster center matrix A i The Euclidean distance.

12. The method for analyzing different power consumption modes of industrial loads based on curve clustering according to claim 11, characterized in that: The expressions of the update functions of the membership degree and cluster center are as follows: In the formula, represents the jth load curve sample x after the t+1th iteration j The degree of membership to the i-th cluster center of the partition; a i (t) and a i (t+1) represents the cluster center a after the tth and t+1th iterations respectively. i location; k (t) represents the cluster center a after the tth iteration k The position of k = 1, 2…C best ; m indicates softness.

13. A device for analyzing different power consumption modes of industrial loads based on curve clustering, running the method for analyzing different power consumption modes of industrial loads based on curve clustering as claimed in any one of claims 1 to 11, characterized in that: The device includes: A data acquisition module is used to acquire load curve data of electricity consumption of each industrial user; A preprocessing module, used for preprocessing the load curve data to obtain a complete and usable load curve data set; A curve clustering module, used for clustering the load curve data set using an improved WOA-FCM algorithm; The power consumption mode differentiation module is used to differentiate the power consumption modes of various industries according to the clustering results; Among them, the improved WOA-FCM algorithm includes: optimizing the positions of each cluster center of the load curve through the WOA whale algorithm, and determining the optimal number of clusters of the load curve in combination with the set effectiveness index, and then using the fuzzy C-means FCM clustering algorithm to optimize the membership matrix and cluster center of the current load curve clustering, and finally obtaining the load curve clustering results of each industrial user.

14. A terminal comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the method according to any one of claims 1-12.

15. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 12 are implemented.

Images