Power load mode mining method and device, equipment and storage medium

Abstract

The invention discloses a power load mode mining method and device, equipment and a storage medium. In the scheme, unsupervised clustering is performed on a target sample under a plurality of different cluster numbers, and an unsupervised clustering result under the plurality of different cluster numbers is obtained. And evaluating the unsupervised clustering results under the plurality of different clustering cluster numbers based on a plurality of clustering effectiveness evaluation indexes to obtain a clustering effectiveness evaluation result which can indicate the clustering effects under the different clustering cluster numbers. And determining an optimal clustering cluster number L according to the clustering validity evaluation result. And performing clustering analysis on the power load curve of the user to be mined in the target scene based on the optimal cluster number L, and determining the power load mode of the user to be mined. The mining accuracy is remarkably improved while the mining efficiency of the power load mode is guaranteed, and efficient and accurate mining of the user power load mode is achieved.

Description

Technical Field

[0001] The present application relates to the technical field of power systems, and in particular to a method, device, equipment and storage medium for mining power load patterns. Background Art

[0002] Power load pattern mining mainly utilizes the power consumption data information of various users accumulated by the power system on the user side to identify, classify and detect the user's power consumption, load distribution and other power consumption behavior characteristics.

[0003] With the popularization of smart grid technology, the Advanced Metering Infrastructure (AMI) can collect high-frequency and large-scale user electricity consumption information on the user side, which provides solid data support for in-depth analysis and understanding of user electricity consumption behavior. However, while the high-frequency collection of user electricity consumption feature information brings new data insight opportunities, it also seriously affects the efficiency of power load pattern mining due to the large amount of data and low effective information density.

[0004] In the existing technology, when faced with high-dimensional power data, dimensionality reduction is usually used to improve the efficiency of power load pattern mining. However, this processing method will sacrifice the information integrity of some user load characteristics and it is difficult to fully retain the original information of the data, thus making it difficult to meet the needs of accurate mining of power load patterns. Summary of the invention

[0005] Based on the above problems, the present application provides a method, device, equipment and storage medium for mining power load patterns, the purpose of which is to achieve efficient and accurate mining of power load patterns.

[0006] The embodiments of the present application disclose the following technical solutions:

[0007] The first aspect of the present application provides a method for mining a power load pattern, the method comprising:

[0008] Performing unsupervised clustering on the target samples under multiple different numbers of clusters to obtain unsupervised clustering results under the multiple different numbers of clusters; the target samples are power load records of users in the target scenario after being cleaned and standardized;

[0009] Based on a plurality of clustering effectiveness evaluation indicators, the unsupervised clustering results under the plurality of different numbers of clustering clusters are evaluated to obtain a clustering effectiveness evaluation result; the clustering effectiveness evaluation result is used to indicate the clustering effect under the different numbers of clustering clusters;

[0010] Determine the optimal number of clusters L according to the clustering effectiveness evaluation result;

[0011] Based on the optimal number of clusters L, cluster analysis is performed on the power load curve of the user to be mined in the target scenario to determine the power load pattern of the user to be mined.

[0012] In an optional implementation, before performing cluster analysis on the power load curve of the user to be mined in the target scenario based on the optimal number of clusters L to determine the power load pattern of the user to be mined, the method further includes:

[0013] Clustering the target sample according to the optimal number of clusters L to obtain L clusters; the L clusters indicate L typical power load modes;

[0014] The average power consumption pattern of each cluster and the variance of the hourly power load of the cluster are calculated; the average power consumption pattern is the cluster center of the corresponding cluster.

[0015] In an optional implementation, the clustering analysis of the power load curve of the user to be mined in the target scenario based on the optimal number of clusters L to determine the power load pattern of the user to be mined includes:

[0016] Based on the matching degree between the power load curve of the user to be mined and the L clusters, obtaining the probability distribution of the power load curve of the user to be mined under the L clusters;

[0017] The power load pattern of the user to be mined is determined based on the probability distribution.

[0018] In an optional implementation, determining the power load pattern of the user to be mined based on the probability distribution includes:

[0019] When the probability of the power load curve of the user to be mined being in a certain cluster is the highest, the average power consumption pattern of the cluster is used as the power load pattern of the user to be mined.

[0020] In an optional implementation, the power load pattern mining method further includes:

[0021] Based on the probability distribution of the power load curve of the user to be mined under the L clusters, the power load volatility of the user to be mined is evaluated; the power load volatility indicates the stability of the power consumption behavior of the user to be mined.

[0022] In an optional implementation, the evaluating the power load volatility of the to-be-mined user based on the probability distribution of the power load curve of the to-be-mined user under the L clusters includes:

[0023] Calculating the power load volatility index of the user to be mined based on the probability distribution of the power load curve of the user to be mined under the L clusters;

[0024] The larger the value of the power load characteristic volatility index is, the greater the power load volatility of the to-be-mined user is; the smaller the value of the power load characteristic volatility index is, the smaller the power load volatility of the to-be-mined user is.

[0025] In an optional implementation, the multiple clustering effectiveness evaluation indicators include: Davies-Bouldin index, S_Dbw index, Akaike information criterion and Bayesian information criterion.

[0026] A second aspect of the present application provides a power load pattern mining device, the device comprising:

[0027] An unsupervised clustering module, used to perform unsupervised clustering on target samples under multiple different numbers of clusters to obtain unsupervised clustering results under the multiple different numbers of clusters; the target samples are power load records of users in target scenarios after cleaning and standardization;

[0028] An effectiveness evaluation module, used to evaluate the unsupervised clustering results under the plurality of different numbers of cluster clusters based on a plurality of clustering effectiveness evaluation indicators to obtain clustering effectiveness evaluation results; the clustering effectiveness evaluation results are used to indicate the clustering effects under the different numbers of cluster clusters;

[0029] A clustering cluster number determination module, used to determine the optimal clustering cluster number L according to the clustering effectiveness evaluation result;

[0030] The power load mining module is used to perform cluster analysis on the power load curve of the user to be mined in the target scenario based on the optimal number of clusters L, and determine the power load pattern of the user to be mined.

[0031] The third aspect of the present application provides a power load pattern mining device, the device comprising: a processor and a memory:

[0032] The memory is used to store program code and transmit the program code to the processor;

[0033] The processor is used to execute the steps of the power load pattern mining method introduced in any implementation manner of the first aspect according to the instructions in the program code.

[0034] A fourth aspect of the present application provides a computer-readable storage medium for storing program code, wherein the program code is used to execute the steps of the power load pattern mining method introduced in any implementation manner of the first aspect.

[0035] Compared with the prior art, this application has the following beneficial effects:

[0036] In the technical solution of the present application, firstly, unsupervised clustering is performed on the target samples under multiple different numbers of clusters to obtain the unsupervised clustering results under the multiple different numbers of clusters, wherein the target samples are the power load records of the users in the target scenario after cleaning and standardization; then, the unsupervised clustering results under the multiple different numbers of clusters are evaluated based on multiple clustering effectiveness evaluation indicators to obtain clustering effectiveness evaluation results, which can indicate the clustering effect under different numbers of clusters; then, the optimal number of clusters L is determined according to the clustering effectiveness evaluation results; finally, based on the optimal number of clusters L, the power load curves of the users to be mined in the target scenario are clustered and analyzed to determine the power load pattern of the users to be mined. In the technical solution of the present application, by adopting unsupervised learning technology, in the face of large-scale high-dimensional power data and low information density, the original information can be effectively retained and redundancy can be reduced, thereby significantly improving the accuracy of mining while ensuring the efficiency of power load pattern mining, and realizing efficient and accurate mining of user power load patterns. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0038] Figure 1 A flow chart of a method for mining power load patterns provided in an embodiment of the present application;

[0039] Figure 2 A schematic diagram of an IDC clustering effectiveness evaluation result provided in an embodiment of the present application;

[0040] Figure 3 A schematic diagram of the effectiveness evaluation results of clustering of electric vehicle charging stations provided in an embodiment of the present application;

[0041] Figure 4 A schematic diagram of the effectiveness evaluation results of clustering of ordinary large industrial and commercial enterprises provided in an embodiment of the present application;

[0042] Figure 5 A flow chart of another method for mining electric load patterns provided in an embodiment of the present application;

[0043] Figure 6A schematic diagram of a typical IDC power load mode curve provided in an embodiment of the present application;

[0044] Figure 7 A schematic diagram of a typical power load mode curve of an electric vehicle charging station provided in an embodiment of the present application;

[0045] Figure 8 A schematic diagram of a typical power load pattern curve for a common large industrial and commercial enterprise provided in an embodiment of the present application;

[0046] Fig. 9 A schematic diagram of power load characteristic volatility under three target scenarios provided in an embodiment of the present application;

[0047] Fig.10 A schematic diagram of the structure of a power load pattern mining device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0048] As described above, the current power load pattern mining work mainly relies on the power consumption data information accumulated by the power system on the user side. With the popularization of smart grid technology and the widespread application of advanced power measurement systems, the frequency and quantity of user power consumption feature information collection have increased significantly. However, the amount of data collected at high frequency is huge and the effective information density is low, which seriously affects the efficiency of power load pattern mining. Although the dimensionality reduction methods in the existing technology can improve efficiency to a certain extent, they often sacrifice the original information and integrity of the data, making it difficult to meet the needs of accurate mining.

[0049] After research, the inventor proposed a method, device, equipment and storage medium for mining power load patterns.

[0050] First, unsupervised clustering is performed on the target samples under multiple different numbers of clusters to obtain unsupervised clustering results under the multiple different numbers of clusters, wherein the target samples are the power load records of users in the target scenario after cleaning and standardization; then, the unsupervised clustering results under the multiple different numbers of clusters are evaluated based on multiple clustering effectiveness evaluation indicators to obtain clustering effectiveness evaluation results, which can indicate the clustering effects under different numbers of clusters; then, the optimal number of clusters L is determined according to the clustering effectiveness evaluation results; finally, based on the optimal number of clusters L, cluster analysis is performed on the power load curves of the users to be mined in the target scenario to determine the power load patterns of the users to be mined.

[0051] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0052] See also Figure 1 , which is a flow chart of a method for mining power load patterns provided by an embodiment of the present application. Figure 1 As shown, the method comprises the following steps:

[0053] S101. Perform unsupervised clustering on target samples under multiple different numbers of clusters to obtain unsupervised clustering results under multiple different numbers of clusters.

[0054] The target samples are the electricity load records of users in the target scenario after cleaning and standardization.

[0055] The target scenario refers to the scenario where the power load pattern mining is to be performed. The target scenario may be a large-scale industrial and commercial enterprise in a certain area, an Internet data center (IDC), or an electric vehicle charging station. In the specific implementation, the target scenario can be selected according to the actual need for power load pattern mining. Accordingly, the target sample is selected according to the selection of the target scenario, which is not specifically limited here.

[0056] Unsupervised clustering refers to a method of classification based on the characteristics of the data itself without pre-labeled category information.

[0057] In the embodiment of the present application, the absolute values ​​of electricity consumption between different users vary greatly. In order to avoid clustering electricity load curves with similar shapes but large differences in electricity consumption into different categories, the electricity load records of users in the target scenario that have been cleaned and standardized are used as target samples. The cleaning and standardization methods are not specifically limited here.

[0058] In an exemplary implementation, the target sample includes the electricity load records of n users, where the i-th user has M i The power load record of the segment, the power load curve sample of the i-th user is

[0059]

[0060] The target sample is

[0061] Take the logarithm z of the target sample x as the input vector for unsupervised clustering

[0062]

[0063] In an exemplary implementation, the target sample is a mixed Gaussian distribution composed of K Gaussian distributions. Based on the probability density function of the mixed Gaussian distribution, the probability density function of z is:

[0064]

[0065] Where p(z) is the probability density function of z, which represents the weighted sum of each component in the mixed Gaussian distribution, where each component corresponds to a Gaussian distribution. d (z|μ k ,∑ k ) is the kth Gaussian distribution, μ k is the mean of the kth Gaussian distribution, Σ k is the covariance matrix of the kth Gaussian distribution, d is the sampling frequency of the power load curve in the target sample, π k is the weight of the kth Gaussian distribution, which represents the relative contribution of the kth Gaussian distribution in the overall probability distribution, and π k Satisfy π k ≥0 and

[0066] In an exemplary implementation, the above-mentioned mixed Gaussian distribution model is solved by the expectation maximization (EM) algorithm, and the user power load curve clustering category is determined by the Gaussian distribution with the highest probability in the mixed Gaussian distribution model, that is, the clustering cluster to which the power load curve of the i-th user belongs is expressed as:

[0067]

[0068] Where K is the number of different Gaussian distribution components in the mixed Gaussian distribution, that is, the total number of clustering categories, k is the kth Gaussian distribution, and k is a positive integer not greater than K, π i，m，k is the probability of the kth Gaussian distribution in the mth section of the power load curve of the ith user, is the cluster to which the power load curve of the mth segment of the ith user belongs when the number of cluster categories is K.

[0069] In the embodiment of the present application, when performing unsupervised clustering on the target sample, it is necessary to preset the number of clusters K, which determines the number of categories of the final power load pattern. By presetting different numbers of clusters K, unsupervised clustering results under multiple different numbers of clusters are obtained. Multiple different numbers of clusters can be selected according to actual conditions and are not specifically limited here.

[0070] In an example implementation, when K is 2-15 clusters, unsupervised clustering can be performed on the target sample to obtain an unsupervised clustering result when K is 2-15 clusters.

[0071] S102. Evaluate unsupervised clustering results under multiple different numbers of clusters based on multiple clustering effectiveness evaluation indicators to obtain clustering effectiveness evaluation results.

[0072] The clustering effectiveness evaluation result is used to indicate the clustering effect under different numbers of clusters.

[0073] In the embodiments of the present application, for unsupervised clustering in which data do not have definite labels, the evaluation of the selection of clustering categories can be based on the compactness of the data within the cluster and the dispersion between clusters after clustering as evaluation criteria, such as: Davies-Bouldin index (DBI), Silhouette Coefficient Index (SI), etc.; it can also be based on measuring the information conveyed by the clustering model as a judgment of effectiveness, such as: Akaike's Information Criterion (AIC), Minimum Description Length Principle (MDL), etc. to evaluate the effectiveness of category division, which is not specifically limited here.

[0074] In an embodiment of the present application, unsupervised clustering results with multiple different numbers of cluster clusters are evaluated based on multiple clustering effectiveness evaluation indicators.

[0075] Optionally, multiple clustering effectiveness evaluation indicators include: Davies-Bouldin Index (DBI), S_Dbw Index, Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC).

[0076] In an example implementation, for the target samples, which are the standardized weekly electricity load curves of Internet Data Centers (IDCs), electric vehicle charging stations, and ordinary large-scale industrial and commercial enterprises, unsupervised clustering of Gaussian mixtures is repeated multiple times under 2-15 Gaussian distribution numbers, and AIC, BIC, DBI index and S_Dbw index are calculated to comprehensively evaluate the clustering effect of Gaussian mixture models with different mixture components on the samples.

[0077] Figure 2 This is a schematic diagram of the IDC clustering effectiveness evaluation results. Figure 3 This is a schematic diagram of the effectiveness evaluation results of the clustering of electric vehicle charging stations. Figure 4 This is a schematic diagram of the effectiveness evaluation results of clustering of ordinary large industrial and commercial enterprises.

[0078] like Figure 2 , Figure 3 and Figure 4 As shown, each figure has four sub-figures, which show the calculation results of DBI index, S_Dbw index, AIC and BIC in turn. The horizontal axis is the number of components K of the Gaussian mixture model, and the vertical axis is the value of each indicator.

[0079] S103: Determine the optimal number of clusters L according to the clustering effectiveness evaluation result.

[0080] In the embodiment of the present application, according to the clustering effectiveness evaluation result obtained in the previous step, a comprehensive analysis is performed to determine the optimal number of clusters L. In the case of the optimal number of clusters L, the data within each cluster is compact, the distribution between clusters is dispersed, and the data can be effectively fitted without excessively increasing the complexity of the model.

[0081] In one example implementation, Figure 2 In the schematic diagram of IDC clustering effectiveness evaluation results shown in the figure, the overall trends of AIC and BIC are basically the same, that is, with the increase in the number of Gaussian distributions, AIC and BIC are both declining, but the rate of decline is slowing down. When K>3, the rate of decline of both almost stops increasing and remains at a relatively small decline level. The DBI index shows a large fluctuation, while the S_Dbw index is maintained at a relatively low level as a whole and tends to be stable after K=4.

[0082] Based on the calculation results of the four clustering effectiveness evaluation indicators, the optimal number of clusters L = 4, that is, the optimal number of components of the Gaussian mixture model is 4. Since the number of components of the Gaussian mixture model determines the number of clustering categories, the number of typical power load patterns is 4. In this clustering case, the data within each cluster is compact, the distribution between clusters is dispersed, and the data can be effectively fitted without excessively increasing the complexity of the model.

[0083] Accordingly, based on Figure 3 and Figure 4 The clustering effectiveness evaluation results shown in Figure 1 show that the optimal number of clusters L for both electric vehicle charging stations and ordinary large-scale industrial and commercial enterprises is 4.

[0084] S104: Perform cluster analysis on the power load curve of the user to be mined in the target scenario based on the optimal number of clusters L, and determine the power load pattern of the user to be mined.

[0085] In the embodiment of the present application, according to the optimal number of clusters L determined in the previous step, a cluster analysis can be performed on the power load area lines of the users to be mined in the target scenario to determine the power load patterns of the users to be mined.

[0086] In an embodiment of the present application, by using the user's electricity load record in the target scenario as the input of unsupervised clustering, the finally determined pattern not only contains the information of the user's load curve, that is, the high-frequency characteristics of the user's demand for electricity and electricity consumption behavior, but also retains the dynamic change characteristics of the user's electricity load pattern over time. By determining the optimal clustering cluster based on a variety of clustering effectiveness evaluation indicators, the inaccurate evaluation of a single clustering effectiveness evaluation indicator is avoided. Through unsupervised learning technology, in the face of large-scale high-dimensional power data and low information density, the original information can be effectively retained and redundancy can be reduced, thereby significantly improving the accuracy of mining while ensuring the efficiency of power load pattern mining, and realizing efficient and accurate mining of user power load patterns.

[0087] Figure 5 Another flow chart of a method for mining power load patterns provided in an embodiment of the present application. In the embodiment introduced by this figure, a more detailed description is provided for the implementation of the method for mining power load patterns.

[0088] like Figure 5 The power load pattern mining method shown includes the following steps:

[0089] S501, performing unsupervised clustering on the target sample under multiple different numbers of clusters to obtain unsupervised clustering results under multiple different numbers of clusters.

[0090] S502: Evaluate unsupervised clustering results with multiple different numbers of clusters based on multiple clustering effectiveness evaluation indicators to obtain clustering effectiveness evaluation results.

[0091] S503: Determine the optimal number of clusters L according to the clustering effectiveness evaluation result.

[0092] S501-S503 are implemented in substantially the same manner as S101-S103 in the method embodiment described above, and will not be elaborated here. For related technical implementations, reference may be made to the description of S101-S103 described above.

[0093] S504: Cluster the target samples according to the optimal number of clusters L to obtain L clusters.

[0094] In the embodiment of the present application, L clusters indicate L typical power load modes.

[0095] In the embodiment of the present application, the target samples are clustered according to the optimal number of clusters L determined in the previous step to obtain L clusters.

[0096] In one example implementation,

[0097] right Perform cluster analysis, and the corresponding target sample x can be divided into L clusters: C * ={c1 * , c2 * , …, c L *}, where, for any 1≤k, l≤L, There is c k * ∩c l * =φ and

[0098] S505: Calculate the average power consumption pattern of each cluster and the variance of the power load per hour in the cluster.

[0099] In the embodiment of the present application, the average power consumption pattern is the cluster center of the corresponding cluster.

[0100] In the embodiment of the present application, the average power consumption pattern of the L clusters divided in the previous step, that is, the cluster center, and the variance of the power load per hour under each cluster are calculated.

[0101] The kth cluster c k * The average power consumption pattern and the variance of the power load per hour in this cluster are calculated as follows:

[0102]

[0103] Among them, c k * ∈C * , x i，m ∈c k * , z i,m ∈c k * .x i，m is the vector corresponding to the mth segment power load record of the ith user in the target sample x, z i，m is the vector corresponding to the m-th segment electricity load record of the ith user in the sample space z after taking the logarithm of the target sample x, ECMc(c k * ) is the cluster c k * The average power consumption pattern, ECMvar(ck * ) is the cluster c k * The variance of the hourly electricity load.

[0104] In the embodiment of the present application, ECMc and ECMvar together give a typical power load pattern. The typical power load pattern is a curve with bandwidth after clustering, ECMc is the center of the curve (a line), and ECMc±ECMvar are the upper and lower limits of the bandwidth.

[0105] For different target scenarios, users' typical power load pattern curves also vary significantly.

[0106] In an exemplary implementation, when the optimal number of clusters L is 4, typical power load pattern curves of IDC, electric vehicle charging station and ordinary large industrial and commercial enterprises can be obtained, as shown in FIG. Figure 6 , Figure 7 and Figure 8 shown.

[0107] like Figure 6 , Figure 7 and Figure 8 As shown, each figure contains four different typical power load patterns. Because the original power load curves of users have been standardized, the vertical axes in the three figures do not correspond to any concept of actual power consumption, but only represent the "high and low" in the shape of the typical power load patterns obtained by clustering the standardized user power load curves. The horizontal axes represent the sampling points of a week.

[0108] S506: Based on the matching degree between the power load curve of the user to be mined and the L clusters, the probability distribution of the power load curve of the user to be mined under the L clusters is obtained.

[0109] In the embodiment of the present application, for user i, the i Daily electricity load curve The matching degree with the L clusters determined in the previous step obtains the probability distribution of the user's power load curve under the L clusters, that is:

[0110]

[0111] Among them, x i，m ∈c k * Determine whether the power load curve of user i belongs to cluster c k * , Calculated in M i The power load curve of Tianzhong belongs to cluster c k *The number of days, The power load curve of user i belongs to cluster c k * The probability of RM i is the cluster probability distribution of the power load curve of user i under L-type Gaussian distribution clustering.

[0112] S507: Determine the power load pattern of the user to be mined based on the probability distribution.

[0113] In the embodiment of the present application, when the probability of the power load curve of the user to be mined being in a certain cluster is the highest, the average power consumption pattern of the cluster is used as the power load pattern of the user to be mined.

[0114] In one example implementation, RM is calculated i The maximum probability of obtaining the power load pattern RMM of user i i , the specific calculation formula is as follows:

[0115]

[0116] Optionally, the power load volatility of the user to be mined may be evaluated based on the probability distribution of the power load curve of the user to be mined under L clusters.

[0117] In the embodiment of the present application, the power load volatility indicates the stability of the power consumption behavior of the user to be mined.

[0118] In the embodiment of the present application, the power load volatility index of the user to be mined can be calculated based on the probability distribution of the power load curve of the user to be mined under L types of clustering clusters.

[0119] The power load volatility index is constructed based on the information entropy theory and is used to evaluate the load characteristic fluctuation. Each user i The power load pattern set It can be regarded as an independent system. According to the probability distribution of the user's power load curve RM i , calculate the user's power load volatility index, the calculation formula is as follows:

[0120]

[0121] in, is the power load volatility index of user i, L is the number of categories of typical power load patterns, is the set of power load patterns at user i c appears in k The probability of a typical power load pattern, ε is a minimum term and satisfies ε>0, avoiding the problem that logarithmic calculation cannot be performed when the probability is 0.

[0122] In the embodiment of the present application, the power load characteristic volatility index obtained is The larger the value of is, the greater the volatility of the power load of the user to be mined is, which means that the power load pattern of the user is more dispersed and the power consumption is more unstable.

[0123] Power load characteristic volatility index The smaller the value of is, the smaller the volatility of the power load of the user to be mined is, which means that the power load pattern of the user is more concentrated and the power consumption behavior is more stable.

[0124] In the embodiment of the present application, according to Figure 6 , Figure 7 and Figure 8 The schematic diagram of typical power load pattern curves under the three target scenarios shown in FIG. 1 can be used to obtain a corresponding table of typical power load pattern distribution under the three target scenarios, as shown in Table 1.

[0125] Table 1 Correspondence between three target scenarios and typical power load pattern distribution

[0126] Target scenario Mode 1 Mode 2 Mode 3 Mode 4 IDC 34.42％ 15.19％ 34.23％ 16.15％ Electric vehicle charging stations 31.23％ 49.16％ 18.39％ 1.23％ Ordinary large industrial and commercial enterprises 16.95％ 15.86％ 24.21％ 42.98％

[0127] According to the typical power load pattern distribution under the three target scenarios shown in Table 1, the power load characteristic volatility index under the three target scenarios can be obtained. Fig. 9 As shown, Fig. 9 The following is a schematic diagram of the power load characteristic volatility under three target scenarios. Charging station is an electric vehicle charging station scenario, General firm is a general large-scale industrial and commercial enterprise scenario, the vertical axis is the power load volatility index, and the horizontal axis is the three target scenarios.

[0128] Optionally, user power load volatility indicators under three target scenarios in different seasons are shown in Table 2.

[0129] Table 2 Correspondence table of power load volatility indicators for three target scenarios in different seasons

[0130] season IDC Electric vehicle charging stations Ordinary large industrial and commercial enterprises spring 0.5363 0.4656 0.5454 summer 0.5609 0.4654 0.5676 autumn 0.5887 0.4707 0.5713 winter 0.5682 0.4699 0.5651

[0131] In an embodiment of the present application, by using the user's power load record in the target scenario as the input of unsupervised clustering, the finally determined pattern can not only contain the information of the user's load curve, that is, the high-frequency characteristics of the user's demand for electricity and power consumption behavior, but also retain the dynamic change characteristics of the user's power load pattern over time. By determining the optimal clustering cluster based on a variety of clustering effectiveness evaluation indicators, the inaccurate evaluation of a single clustering effectiveness evaluation indicator is avoided. Based on the typical power load pattern extracted from the load curve, the group commonalities and individual differences of user power loads at a microscale can be more accurately excavated. Through unsupervised learning technology, in the face of large-scale high-dimensional power data and low information density, the original information can be effectively retained and redundancy can be reduced. At the same time, the embodiment of the present application achieves a more accurate evaluation of power load volatility by constructing a more comprehensive and intuitive power load volatility evaluation indicator.

[0132] Based on the power load pattern mining method provided in the aforementioned embodiment, the present application also provides a power load pattern mining device accordingly. Fig.10 This is a schematic diagram of the structure of a power load mode mining device provided in an embodiment of the present application. Fig.10 As shown, the power load pattern mining device includes:

[0133] The unsupervised clustering module 1001 is used to perform unsupervised clustering on the target samples under multiple different numbers of clusters to obtain unsupervised clustering results under the multiple different numbers of clusters; the target samples are the power load records of users in the target scenario after cleaning and standardization;

[0134] The effectiveness evaluation module 1002 is used to evaluate the unsupervised clustering results under the plurality of different numbers of cluster clusters based on a plurality of clustering effectiveness evaluation indicators to obtain clustering effectiveness evaluation results; the clustering effectiveness evaluation results are used to indicate the clustering effects under the different numbers of cluster clusters;

[0135] A clustering cluster number determination module 1003 is used to determine the optimal clustering cluster number L according to the clustering effectiveness evaluation result;

[0136] The power load mining module 1004 is used to perform cluster analysis on the power load curve of the user to be mined in the target scenario based on the optimal number of clusters L, and determine the power load pattern of the user to be mined.

[0137] In the embodiment of the present application, the optimal clustering cluster is determined based on multiple clustering effectiveness evaluation indicators, thereby avoiding the situation where a single clustering effectiveness evaluation indicator is inaccurately evaluated. Through unsupervised learning technology, in the face of large-scale high-dimensional power data and low information density, the original information can be effectively retained and redundancy can be reduced, thereby significantly improving the accuracy of mining while ensuring the efficiency of power load pattern mining, and realizing efficient and accurate mining of user power load patterns.

[0138] Optionally, the power load pattern mining device also includes a clustering module, which is used to cluster the target samples according to the optimal number of clusters L to obtain L clusters; the L clusters indicate L typical power load patterns.

[0139] The average power consumption pattern of each cluster and the variance of the hourly power load of the cluster are calculated; the average power consumption pattern is the cluster center of the corresponding cluster.

[0140] Optionally, the power load mining module 1004 is specifically used for:

[0141] Based on the matching degree between the power load curve of the user to be mined and the L clusters, obtaining the probability distribution of the power load curve of the user to be mined under the L clusters;

[0142] The power load pattern of the user to be mined is determined based on the probability distribution.

[0143] When the probability of the power load curve of the user to be mined being in a certain cluster is the highest, the average power consumption pattern of the cluster is used as the power load pattern of the user to be mined.

[0144] Optionally, the power load pattern mining device also includes a power load volatility assessment module, which is used to evaluate the power load volatility of the user to be mined based on the probability distribution of the power load curve of the user to be mined under the L clusters; the power load volatility indicates the stability of the power consumption behavior of the user to be mined.

[0145] Optionally, the power load volatility assessment module is specifically used for:

[0146] Based on the probability distribution of the power load curve of the user to be mined under the L clusters, the power load volatility index of the user to be mined is calculated.

[0147] The larger the value of the power load characteristic volatility index is, the greater the power load volatility of the to-be-mined user is; the smaller the value of the power load characteristic volatility index is, the smaller the power load volatility of the to-be-mined user is.

[0148] Optionally, the multiple clustering effectiveness evaluation indicators in the effectiveness evaluation module 1002 include: Davies-Bouldin index, S_Dbw index, Akaike information criterion and Bayesian information criterion.

[0149] In addition, an embodiment of the present application also provides a power load pattern mining device, which includes a processor and a memory.

[0150] The memory is used to store computer program code and transmit the program code to the processor;

[0151] The processor is used to execute the steps of the power load pattern mining method introduced in any of the above method embodiments according to the instructions in the program code.

[0152] In addition, an embodiment of the present application also provides a computer-readable storage medium, in which a computer program code is stored. When the program code is executed by a processor, the power load pattern mining method introduced in any manner of the method embodiment is implemented.

[0153] It should be noted that each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device and equipment embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments. The device and equipment embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components indicated as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. A person of ordinary skill in the art can understand and implement it without paying creative labor.

[0154] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for mining power load patterns, characterized in that: The method comprises: Performing unsupervised clustering on the target samples under multiple different numbers of clusters to obtain unsupervised clustering results under the multiple different numbers of clusters; the target samples are power load records of users in the target scenario after being cleaned and standardized; Based on a plurality of clustering effectiveness evaluation indicators, the unsupervised clustering results under the plurality of different numbers of clustering clusters are evaluated to obtain a clustering effectiveness evaluation result; the clustering effectiveness evaluation result is used to indicate the clustering effect under the different numbers of clustering clusters; Determine the optimal number of clusters L according to the clustering effectiveness evaluation result; Based on the optimal number of clusters L, cluster analysis is performed on the power load curve of the user to be mined in the target scenario to determine the power load pattern of the user to be mined.

2. The method according to claim 1, characterized in that Before performing cluster analysis on the power load curve of the user to be mined in the target scenario based on the optimal number of clusters L to determine the power load pattern of the user to be mined, the method further includes: Clustering the target sample according to the optimal number of clusters L to obtain L clusters; the L clusters indicate L typical power load modes; The average power consumption pattern of each cluster and the variance of the hourly power load of the cluster are calculated; the average power consumption pattern is the cluster center of the corresponding cluster.

3. The method according to claim 2, characterized in that The clustering analysis is performed on the power load curve of the user to be mined in the target scenario based on the optimal number of clusters L to determine the power load pattern of the user to be mined, including: Based on the matching degree between the power load curve of the user to be mined and the L clusters, obtaining the probability distribution of the power load curve of the user to be mined under the L clusters; The power load pattern of the user to be mined is determined based on the probability distribution.

4. The method according to claim 3, characterized in that The determining the power load pattern of the user to be mined based on the probability distribution includes: When the probability of the power load curve of the user to be mined being in a certain cluster is the highest, the average power consumption pattern of the cluster is used as the power load pattern of the user to be mined.

5. The method according to claim 3, characterized in that: The method further comprises: Based on the probability distribution of the power load curve of the user to be mined under the L clusters, the power load volatility of the user to be mined is evaluated; the power load volatility indicates the stability of the power consumption behavior of the user to be mined.

6. The method according to claim 5, characterized in that The evaluating the power load volatility of the user to be mined based on the probability distribution of the power load curve of the user to be mined under the L clusters includes: Calculating the power load volatility index of the user to be mined based on the probability distribution of the power load curve of the user to be mined under the L clusters; The larger the value of the power load characteristic volatility index is, the greater the power load volatility of the to-be-mined user is; the smaller the value of the power load characteristic volatility index is, the smaller the power load volatility of the to-be-mined user is.

7. The method according to claim 1, characterized in that The multiple clustering effectiveness evaluation indicators include: Davies-Bouldin index, S_Dbw index, Akaike information criterion and Bayesian information criterion.

8. A power load pattern mining device, characterized in that: The device comprises: An unsupervised clustering module, used to perform unsupervised clustering on target samples under multiple different numbers of clusters to obtain unsupervised clustering results under the multiple different numbers of clusters; the target samples are power load records of users in target scenarios after cleaning and standardization; An effectiveness evaluation module, used to evaluate the unsupervised clustering results under the plurality of different numbers of cluster clusters based on a plurality of clustering effectiveness evaluation indicators to obtain clustering effectiveness evaluation results; the clustering effectiveness evaluation results are used to indicate the clustering effects under the different numbers of cluster clusters; A clustering cluster number determination module, used to determine the optimal clustering cluster number L according to the clustering effectiveness evaluation result; The power load mining module is used to perform cluster analysis on the power load curve of the user to be mined in the target scenario based on the optimal number of clusters L, and determine the power load pattern of the user to be mined.

9. A power load pattern mining device, characterized in that: include: Processor and memory: The memory is used to store program code and transmit the program code to the processor; The processor is used to execute the steps of the power load pattern mining method according to any one of claims 1 to 7 according to the instructions in the program code.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium is used to store program codes, and the program codes are used to execute the steps of the power load pattern mining method according to any one of claims 1 to 7.

Images