Load clustering method and system considering load characteristics and adjustable capability

Abstract

The invention relates to the technical field of power system load clustering, and provides a load clustering method and system considering load characteristics and adjustable capability, and the method comprises the steps: constructing a comprehensive analysis framework of a user electrical load, and obtaining a user electrical load curve; and performing cleaning and normalization processing on a user power utilization load curve, performing primary clustering of user load characteristics on a standardized data curve, performing secondary clustering of user adjustable capability on a plurality of user groups with the same load characteristics, and obtaining a user load comprehensive clustering result according to a contour coefficient threshold. According to the method, a comprehensive analysis framework of the user electrical load is constructed, user private data security is guaranteed, the communication and control capability of the system is improved, primary clustering is performed on user load data through a clustering algorithm based on the Euclidean form distance, a secondary clustering result considering the user adjustable capability is obtained by utilizing SOM, the clustering efficiency is improved, and the user experience is improved. And the practicability is higher.

Description

Technical Field

[0001] The present invention relates to the technical field of load clustering in power systems, and in particular to a load clustering method and system that considers load characteristics and adjustable capabilities. Background Art

[0002] Electricity users are participating in various businesses of power grid companies, and the interactive regulation of loads, power sources and networks is increasing. However, the participation of large-scale users has brought challenges to the grid companies in carrying out demand-side dispatching. The load response of users varies, and the user behavior characteristics are highly differentiated. It is necessary to accurately classify electricity users to support higher-quality grid-friendly interaction and regulation.

[0003] The large-scale popularization of smart meters has made it easier to collect user electricity consumption data, and fine-grained electricity load data has provided a guarantee for the stability of supply and demand interaction. Based on advanced communication, metering and data management technologies, power grid companies have widely collected data from a large number of power users, reflecting the diversity of user electricity load characteristics.

[0004] However, the influx of massive data also means redundancy and confusion of power data information. Therefore, how to use data mining technology to extract effective information from power consumption data, analyze user power consumption patterns, and provide reference for power decision-making has become an important research direction in the field of power grids. Power load curve clustering is a common technology in power data mining. Its main purpose is to extract the distribution characteristics of user power load curves, identify similar user behaviors, and conduct classification analysis. At present, various clustering analysis methods have been applied to power load clustering, mainly including partition-based clustering methods, hierarchical clustering methods, and graph-based clustering methods.

[0005] The existing technology fails to effectively construct a comprehensive analysis framework for user power load that takes into account user load characteristics and adjustable capabilities, and when constructing the relevant framework, data transmission security requirements are rarely considered, resulting in low practicality.

[0006] Traditional methods often only consider clustering once when classifying and aggregating user electricity consumption data, and often cannot fully explore the user load characteristics and user load adjustable capacity information, which is not conducive to the upper-level power grid to carry out efficient regulation and control, and is not very practical. Summary of the invention

[0007] The present invention aims to solve at least one of the technical problems existing in the related art. To this end, the present invention provides a load clustering method and system that considers load characteristics and adjustable capacity. The present invention constructs a comprehensive analysis framework for user power loads. When constructing the relevant framework, the data transmission security requirements are fully considered, the user's private data security is guaranteed, and the communication and control capabilities of the system are improved. The user load data is clustered once by a clustering algorithm based on Euclidean morphological distance. Secondary clustering is performed for user groups with the same type of load characteristics, and the secondary clustering results that consider the user's adjustable capacity are obtained by using the pattern classification performance of SOM, which improves the clustering efficiency and has higher practicality.

[0008] The present invention provides a load clustering method considering load characteristics and adjustable capacity, comprising: S1: A comprehensive analysis framework for user power load is constructed, and user power load information data is collected to obtain user power load curves; S2: Clean and normalize the user's power load curve to obtain a standardized data curve; S3: clustering the user load characteristics of the standardized data curve to obtain multiple user groups with the same load characteristics; S4: performing secondary clustering of user adjustable capacity for multiple user groups with the same load characteristics to obtain user adjustable capacity clustering data; S5: Identify the user adjustable capacity clustering data according to the silhouette coefficient threshold to obtain a comprehensive clustering result of the user load.

[0009] Furthermore, the comprehensive analysis framework of user power load includes: Collect load data through smart meters, record the power consumption of each user in a fixed time interval in detail, and present the data in the form of a curve to obtain the user's power load characteristic information; Obtain information on users' power adjustment capabilities through questionnaire surveys; The user power load information data includes user power load characteristic information and user power adjustable capacity information.

[0010] Furthermore, step S2 includes: S21: Identify and delete abnormal data in the user's power load curve to obtain a first data curve; S22: by referring to the normal data on both sides of the abnormal data point, using Gaussian filtering to smooth the first data curve to obtain a second data curve; S23: performing normalization processing on the second data curve to obtain a standardized data curve.

[0011] Furthermore, step S3 includes: S31: Select any two standardized data curves of equal length; S32: calculating the Euclidean distance between two standardized data curves; S33: extracting the user power load characteristic sequence of the two standardized data curves to obtain two time series curves, and calculating the morphological distance between the two standardized data curves according to the time series curves; S34: calculating the Euclidean morphological distance between two standardized data curves according to the Euclidean distance and the morphological distance; S35: Calculating a correction index based on the Euclidean morphological distance according to the Euclidean morphological distance between the two standardized data curves, and obtaining a plurality of user groups having the same load characteristics according to the correction index based on the Euclidean morphological distance.

[0012] Furthermore, in step S33, the calculation expression of the morphological distance is: in, The standardized data curve and The shape distance between The standardized data curve No. values, The standardized data curve No. values, The standardized data curve The maximum amplitude of The standardized data curve The maximum amplitude of is the number of normalized data curves.

[0013] Furthermore, in step S34, the Euclidean morphological distance calculation expression is: in, The standardized data curve and The Euclidean distance between is the weight coefficient of Euclidean distance, The standardized data curve and The shape distance between is the weight coefficient of morphological distance, The standardized data curve and The Euclidean distance between The similarity of load curves is measured by Euclidean morphological distance.

[0014] Further, step S35 includes: S351: randomly selecting K standardized data curves as initial cluster centers, sorting the Euclidean distances between the data, clustering and forming a cluster set, and obtaining the number of clusters; S352: Calculate the correction index based on the Euclidean morphological distance according to the number of clusters. The calculation expression is: in, is a modified indicator based on Euclidean morphological distance, Cluster The average Euclidean distance between the sample points and the sample center, Cluster The average Euclidean distance between the sample points and the sample center; Cluster and Cluster The average Euclidean distance between sample centers, To find the maximum value function; S353: comparing the index threshold and the correction index based on the Euclidean morphological distance; If the correction index based on the Euclidean morphological distance is greater than the index threshold, execute S351; If the correction index based on the Euclidean morphological distance is less than or equal to the index threshold, execute S354; S354: Compare the number of clusters and the cluster number threshold; If the number of clusters is greater than the clustering threshold, the optimal number of clusters is obtained; If the number of clusters is less than or equal to the cluster threshold, execute S351; S355: Obtain multiple user groups with the same load characteristics according to the optimal number of clusters.

[0015] Furthermore, step S4 includes: S41: performing dimensionality reduction processing on a plurality of user group data having the same load characteristics by using a principal component analysis technology to obtain dimensionality reduction data; S42: performing secondary clustering of user adjustable capabilities on the dimension reduction data through a self-organizing competitive neural network to obtain user adjustable capability clustering data.

[0016] Furthermore, step S5 includes: The silhouette coefficient is selected as the evaluation index of the comprehensive clustering results of user load characteristics and adjustable capacity. If the silhouette coefficient is less than or equal to the silhouette coefficient threshold, the self-organizing competitive neural network is reconstructed; If the silhouette coefficient is greater than the silhouette coefficient threshold, the comprehensive clustering result of the user load is obtained.

[0017] The present invention also provides a load clustering system considering load characteristics and adjustable capacity, which is used to execute a load clustering method considering load characteristics and adjustable capacity, comprising: A construction and acquisition module, wherein the construction and acquisition module constructs a comprehensive analysis framework of the user's power load, collects user power load information data, and obtains the user's power load curve; A preprocessing module, which cleans and normalizes the user's power load curve to obtain a standardized data curve; A primary clustering module, wherein the primary clustering module performs primary clustering of user load characteristics on the standardized data curve to obtain multiple user groups with the same load characteristics; A secondary clustering module, wherein the secondary clustering module performs secondary clustering of user adjustable capabilities on a plurality of user groups having the same load characteristics to obtain user adjustable capability clustering data; A comprehensive clustering module identifies user adjustable capacity clustering data according to a silhouette coefficient threshold to obtain a comprehensive clustering result of user load.

[0018] The above one or more technical solutions in the embodiments of the present invention have at least one of the following technical effects: The present invention constructs a comprehensive analysis framework for user power loads. When constructing the relevant framework, data transmission security requirements are fully considered, user private data security is guaranteed, and the communication and control capabilities of the system are improved. User load data is clustered once using a clustering algorithm based on Euclidean morphological distance. Secondary clustering is performed for user groups with the same type of load characteristics, and the secondary clustering results that take into account the user's adjustable capabilities are obtained using the pattern classification performance of the SOM, thereby improving clustering efficiency and having higher practicality.

[0019] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the present invention or 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 some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0021] Figure 1 It is a flow chart of a load clustering method considering load characteristics and adjustable capacity provided by the present invention.

[0022] Figure 2It is a schematic diagram of the structure of a load clustering system provided by the present invention that takes load characteristics and adjustable capabilities into consideration.

[0023] Reference numerals: 101. Construction and acquisition module; 102. Preprocessing module; 103. Primary clustering module; 104. Secondary clustering module; 105. Comprehensive clustering module. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical scheme and advantages of the present invention clearer, the technical scheme in the present invention will be clearly and completely described below. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present invention. The following embodiments are used to illustrate the present invention, but cannot be used to limit the scope of the present invention.

[0025] In the description of the embodiments of the present invention, it should be noted that the terms "first", "second", and "third" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance. In the description of this specification, the description of reference terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the embodiments of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples without contradiction.

[0026] Combine the following Figure 1 to Figure 2 A load clustering method and system considering load characteristics and adjustable capacity of the present invention are described.

[0027] like Figure 1 As shown, a load clustering method considering load characteristics and adjustable capacity includes: S1: Build a comprehensive analysis framework for user power load, collect user power load information data, and obtain user power load curve; The comprehensive analysis framework of user power load includes: Collect load data through smart meters, record the power consumption of each user in a fixed time interval in detail, and present the data in the form of a curve to obtain the user's power load characteristic information; Obtain information on users' power adjustment capabilities through questionnaire surveys; The user power load information data includes user power load characteristic information and user power adjustable capacity information.

[0028] In some specific embodiments of the present invention, the user electricity load analysis based on power big data mainly focuses on two parts: mining and analysis of user basic electricity consumption information and mining and analysis of user adjustable electricity consumption information. By performing multi-dimensional analysis on user basic electricity consumption information, the user's electricity consumption pattern can be accurately identified. Based on the residential user electricity load, a multi-dimensional research framework for factors affecting user electricity load is proposed.

[0029] The factors affecting the user's electricity load can be divided into two categories: internal factors and natural factors. Self-factors: refers to the influence of users’ subjective factors on their electricity consumption behavior, mainly including users’ gender, age, behavioral preferences, etc.

[0030] Natural factors: refers to the impact of the external environment on the user's electricity load, mainly including temperature, humidity, weather conditions, etc.

[0031] The comprehensive analysis of user power load is based on the basic power information collected by the smart meter and the user power adjustable capacity information obtained through the questionnaire survey. The present invention fully considers the user's load characteristics and load adjustable capacity, and constructs a comprehensive analysis framework for user power load; In terms of mining and analysis of power load data, the load data collected by smart meters records in detail the power consumption of each user within a fixed time interval, and presents these data in the form of curves, intuitively showing the user's load characteristics. In the mining and analysis of power big data combined with user power adjustable capacity information, the adjustable capacity of user loads is classified according to different recognition patterns and influencing factors, thereby reflecting the different adjustment potentials of user loads.

[0032] The present invention provides a scientific basis for accurately evaluating user power load and its adjustable capacity, which helps to achieve more flexible and efficient power demand management.

[0033] In order to achieve remote transmission and analysis of user power consumption data, the present invention collects all data transmission and storage devices through an intelligent gateway, and uploads power generation and load data to the intranet cloud platform. The intranet cloud platform only allows intranet access, thereby ensuring the security of user data and effectively protecting user privacy.

[0034] The intelligent gateway used in the present invention supports multiple network standards, including GPRS, 5G / 4G / 3G, 1.8GHz power line wireless private network, 230MHz power wireless private network and fiber optic private network, and is compatible with multiple communication protocols such as TCP, DL / T645, DL / T698, CDT, Modbus, UDP, etc.

[0035] During the collection and storage of user power load information, data collection and storage equipment may be affected by electromagnetic interference, extreme natural conditions and other factors, resulting in abnormalities in power data, such as data loss, etc. Such abnormal data will not only affect the accuracy of data analysis, but may also interfere with subsequent data processing and analysis.

[0036] S2: Clean and normalize the user's power load curve to obtain a standardized data curve; S21: Identify and delete abnormal data in the user's power load curve to obtain a first data curve; Remove curves with zero load, constant load throughout the day, and load missing more than 10%.

[0037] S22: By referring to the normal data on both sides of the abnormal data point, the first data curve is smoothed by using Gaussian filtering to obtain a second data curve, thereby ensuring the integrity and reliability of the data.

[0038] S23: performing normalization processing on the second data curve to obtain a standardized data curve.

[0039] The present invention maps the data to The specific definition within the range is as follows: in, The amount of data is of The second data curve, ; After normalization dimensional normalized data curve, ; The maximum value of the normalized data curve for each time point, The data curves were normalized to the minimum value for each time point.

[0040] The data were normalized to ensure standardization.

[0041] S3: clustering the user load characteristics of the standardized data curve to obtain multiple user groups with the same load characteristics; S31: Select any two standardized data curves of equal length; Select the normalized data curve And the normalized data curve , , ; S32: calculating the Euclidean distance between two standardized data curves; and The Euclidean distance between them is: in, for and The Euclidean distance between for On Points, for On points; the Euclidean distance is used to measure the overall distribution characteristics of the user load curve.

[0042] S33: extracting the user power load characteristic sequence of the two standardized data curves to obtain two time series curves, and calculating the morphological distance between the two standardized data curves according to the time series curves; The power load curve reflects the power consumption of users at different times. In essence, it is the superposition of a series of power consumption behaviors of users. The present invention converts the power load curve of users into a set of feature sequences, aiming to describe the power consumption behavior characteristics of users at different time points. In order to measure the differences in power load characteristics of different users, the present invention ignores the differences in power consumption levels between different users during analysis.

[0043] The calculation expression of morphological distance is: in, The standardized data curve and The shape distance between The standardized data curve No. values, The standardized data curve No. values, The standardized data curve The maximum amplitude of The standardized data curve The maximum amplitude of is the number of normalized data curves.

[0044] The smaller the value, the longer the common feature subsequence of the standardized data curve is and the more similar their morphological change characteristics are.

[0045] S34: Calculate the Euclidean morphological distance between the two standardized data curves according to the Euclidean distance and the morphological distance; the Euclidean morphological distance calculation expression is: in, The standardized data curve and The Euclidean distance between is the weight coefficient of Euclidean distance, The standardized data curve and The shape distance between is the weight coefficient of morphological distance, The standardized data curve and The Euclidean distance between The present invention comprehensively considers the overall distribution characteristics of the curve morphology and the fluctuation characteristics of the curve, constructs a dual-scale similarity measurement of load curves, namely, the Euclidean morphological distance, and measures the similarity of the load curves by the Euclidean morphological distance.

[0046] S35: Calculating a correction index based on the Euclidean morphological distance according to the Euclidean morphological distance between the two standardized data curves, and obtaining a plurality of user groups having the same load characteristics according to the correction index based on the Euclidean morphological distance.

[0047] S351: randomly selecting K standardized data curves as initial cluster centers, sorting the Euclidean distances between the data, clustering and forming a cluster set, and obtaining the number of clusters; S352: Calculate the correction index based on the Euclidean morphological distance according to the number of clusters, The cluster internal evaluation index is generally used to evaluate the quality of unlabeled data clustering, requiring the data to have a small intra-class aggregation degree and a large inter-class difference degree. The present invention adopts the DB (Davies-Bouldin) index which has a good clustering evaluation effect on the power data set. The DB index can comprehensively consider the intra-cluster aggregation degree and the inter-cluster separation degree. The calculation expression is: in, is the separation between two clusters, Cluster The average Euclidean distance between the sample points and the sample center, and Cluster The average Euclidean distance between the sample points and the sample center, as well as Indicates the degree of aggregation between samples within a cluster; Cluster and Cluster The average Euclidean distance between sample centers in ; Considering the traditional Euclidean distance It is no longer possible to accurately evaluate the method of the present invention, and a correction index based on the Euclidean morphological distance is proposed; The calculation expression of the modified indicator based on Euclidean morphological distance is: in, is a modified indicator based on Euclidean morphological distance, Cluster The average Euclidean distance between the sample points and the sample center, Cluster The average Euclidean distance between the sample points and the sample center; Cluster and Cluster The average Euclidean distance between sample centers, To find the maximum value function; and The ratio of the intra-class aggregation degree to the inter-class separation degree between two clusters is used, so the lower the value, the higher the clustering quality; S353: comparing the index threshold and the correction index based on the Euclidean morphological distance; If the correction index based on the Euclidean morphological distance is greater than the index threshold, execute S351; If the correction index based on the Euclidean morphological distance is less than or equal to the index threshold, execute S354; When selecting the optimal number of clusters, we generally look for its minimum point. The cluster number corresponding to the minimum value is taken as the optimal cluster number. The present invention sets the initial cluster number to 2 and the cluster number threshold ,in, is the number of normalized data curves, The threshold is provided by the system.

[0048] S354: Compare the number of clusters and the cluster number threshold; If the number of clusters is greater than the clustering threshold, the optimal number of clusters is obtained; If the number of clusters is less than or equal to the cluster threshold, execute S351.

[0049] S355: Obtain multiple user groups with the same load characteristics according to the optimal number of clusters.

[0050] The purpose of primary clustering is to classify user load data into several user groups with the same load characteristics and aggregate these user groups with the same load characteristics.

[0051] S4: performing secondary clustering of user adjustable capacity for multiple user groups with the same load characteristics to obtain user adjustable capacity clustering data; S41: performing dimensionality reduction processing on multiple user group data with the same load characteristics through principal component analysis (PCA) technology to obtain dimensionality reduction data; Based on the primary clustering, the present invention conducts secondary classification and aggregation in the user group set with the same load characteristics. The data set of the secondary clustering is the power big data considering the factors affecting the user's power load. Since the dimension of the data set is high, it takes a long time to analyze it using the clustering algorithm, so it is necessary to perform dimensionality reduction on the data set. The present invention selects PCA to perform feature dimensionality reduction on the multi-dimensional influencing factor data of the user's power load.

[0052] There are multiple power users in a user group with the same load characteristics. The electricity consumption behavior of each user is affected by multiple indicators. The PCA technology is used to extract the principal components of the influencing factors of the user's electricity load in the user group with the same load characteristics, and the principal component vector is obtained. The principal component vector is the dimensionality reduction data.

[0053] S42: performing secondary clustering of user adjustable capabilities on the dimension reduction data through a self-organizing competitive neural network to obtain user adjustable capability clustering data.

[0054] Based on the dimensionality reduction results of PCA, the present invention uses a self-organizing competitive neural network (SOM) to perform secondary clustering on the user adjustable capabilities. This algorithm is a typical unsupervised learning algorithm, which automatically classifies the submitted input patterns through a training link. The algorithm steps can be divided into three links: S421: Vector normalization; the principal component vector As the input vector of the SOM network, The inner star weight vector corresponding to the neuron Perform normalization to obtain the normalized principal component vector and ,in, , is the number of inner star weight vectors.

[0055] S422: Find the winning neuron. When the network gets a When all the competition layers Both Compare similarities and The most similar Determined as the winning neuron in the competition, denoted as , where the similarity measure is calculated by and The calculation expression for finding the winning neuron is: in, is transposed.

[0056] S423: Network input and weight adjustment. According to the SOM "winner takes it all" competition rule, only the winning neuron has its weight vector adjusted. , and only the winning neuron outputs 1, and the outputs of the remaining neurons are 0. The adjusted weight vector is: in, For the The winning neuron of the competition is For the The winning neuron of the competition is is the learning rate, and its value gradually decreases as the learning progresses. , when , since the "winner" inhibits them and does not allow them to be excited, the weights of the corresponding neurons are not adjusted accordingly.

[0057] After the weight adjustment is completed, return to step S421 to continue training. When the decay reaches 0 or a specified value, the above operation is stopped and the training result is output.

[0058] S5: Identify the user adjustable capacity clustering data according to the silhouette coefficient threshold to obtain a comprehensive clustering result of the user load.

[0059] The silhouette coefficient is selected as the evaluation index of the comprehensive clustering results of user load characteristics and adjustable capacity. If the silhouette coefficient is less than or equal to the silhouette coefficient threshold, the self-organizing competitive neural network is reconstructed; If the silhouette coefficient is greater than the silhouette coefficient threshold, the comprehensive clustering result of the user load is obtained.

[0060] The present invention selects the silhouette coefficient as an evaluation index for the comprehensive clustering results of user load characteristics and adjustable capacity. The two factors of intra-class aggregation and inter-class separation are fully considered. The closer the value is to 1, the higher the clustering effectiveness is. The calculation expression is: in, is the distance from a certain feature vector to all other points in the cluster to which it belongs; is the average distance from a certain eigenvector to all points in the nearest cluster. is the maximum value function.

[0061] The present invention sets a threshold ,when When the SOM classification model is rebuilt, Provided by the system.

[0062] like Figure 2 As shown, a load clustering system considering load characteristics and adjustable capacity is used to execute the above-mentioned load clustering method considering load characteristics and adjustable capacity, including: The construction and acquisition module 101 constructs a comprehensive analysis framework of the user's power load, collects user's power load information data, and obtains the user's power load curve; The preprocessing module 102 cleans and normalizes the user's power load curve to obtain a standardized data curve; The primary clustering module 103 performs primary clustering of user load characteristics on the standardized data curve to obtain multiple user groups with the same load characteristics; The secondary clustering module 104 performs secondary clustering of user adjustable capacity on multiple user groups with the same load characteristics to obtain user adjustable capacity clustering data; The comprehensive clustering module 105 identifies the user adjustable capacity clustering data according to the silhouette coefficient threshold to obtain the user load comprehensive clustering result.

[0063] Through the collaborative work of the above modules, a comprehensive analysis framework for user power load is constructed. When constructing the relevant framework, the data transmission security requirements are fully considered to ensure the security of user private data and improve the communication and control capabilities of the system. The user load data is clustered once using a clustering algorithm based on Euclidean morphological distance. Secondary clustering is performed for user groups with the same type of load characteristics, and the secondary clustering results that take into account the user's adjustable capacity are obtained using the pattern classification performance of SOM, which improves the clustering efficiency and has higher practicality.

[0064] 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 aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A load clustering method considering load characteristics and adjustable capacity, characterized in that: include: S1: A comprehensive analysis framework for user power load is constructed, and user power load information data is collected to obtain user power load curves; S2: Clean and normalize the user's power load curve to obtain a standardized data curve; S3: clustering the user load characteristics of the standardized data curve to obtain multiple user groups with the same load characteristics; S4: performing secondary clustering of user adjustable capacity for multiple user groups with the same load characteristics to obtain user adjustable capacity clustering data; S5: Identify the user adjustable capacity clustering data according to the silhouette coefficient threshold to obtain a comprehensive clustering result of the user load.

2. A load clustering method considering load characteristics and adjustable capacity according to claim 1, characterized in that: The comprehensive analysis framework of user power load includes: Collect load data through smart meters, record the power consumption of each user in a fixed time interval in detail, and present the data in the form of a curve to obtain the user's power load characteristic information; Obtain information on users' power adjustment capabilities through questionnaire surveys; The user power load information data includes user power load characteristic information and user power adjustable capacity information.

3. A load clustering method considering load characteristics and adjustable capacity according to claim 1, characterized in that: The S2 step includes: S21: Identify and delete abnormal data in the user's power load curve to obtain a first data curve; S22: by referring to the normal data on both sides of the abnormal data point, using Gaussian filtering to smooth the first data curve to obtain a second data curve; S23: performing normalization processing on the second data curve to obtain a standardized data curve.

4. A load clustering method considering load characteristics and adjustable capacity according to claim 1, characterized in that: The S3 steps include: S31: Select any two standardized data curves of equal length; S32: calculating the Euclidean distance between two standardized data curves; S33: extracting the user power load characteristic sequence of the two standardized data curves to obtain two time series curves, and calculating the morphological distance between the two standardized data curves according to the time series curves; S34: calculating the Euclidean morphological distance between two standardized data curves according to the Euclidean distance and the morphological distance; S35: Calculating a correction index based on the Euclidean morphological distance according to the Euclidean morphological distance between the two standardized data curves, and obtaining a plurality of user groups having the same load characteristics according to the correction index based on the Euclidean morphological distance.

5. A load clustering method considering load characteristics and adjustable capacity according to claim 4, characterized in that: In step S33, the calculation expression of the morphological distance is: in, The standardized data curve and The shape distance between The standardized data curve No. values, The standardized data curve No. values, The standardized data curve The maximum amplitude of The standardized data curve The maximum amplitude of is the number of normalized data curves.

6. A load clustering method considering load characteristics and adjustable capacity according to claim 4, characterized in that: In step S34, the Euclidean morphological distance calculation expression is: in, The standardized data curve and The Euclidean distance between is the weight coefficient of Euclidean distance, The standardized data curve and The shape distance between is the weight coefficient of morphological distance, The standardized data curve and The Euclidean distance between The similarity of load curves is measured by Euclidean morphological distance.

7. A load clustering method considering load characteristics and adjustable capacity according to claim 4, characterized in that: Step S35 includes: S351: randomly selecting K standardized data curves as initial cluster centers, sorting the Euclidean distances between the data, clustering and forming a cluster set, and obtaining the number of clusters; S352: Calculate the correction index based on the Euclidean morphological distance according to the number of clusters. The calculation expression is: in, is a modified indicator based on Euclidean morphological distance, Cluster The average Euclidean distance between the sample points and the sample center, Cluster The average Euclidean distance between the sample points and the sample center; Cluster and Cluster The average Euclidean distance between sample centers, To find the maximum value function; S353: comparing the index threshold and the correction index based on the Euclidean morphological distance; If the correction index based on the Euclidean morphological distance is greater than the index threshold, execute S351; If the correction index based on the Euclidean morphological distance is less than or equal to the index threshold, execute S354; S354: Compare the number of clusters and the cluster number threshold; If the number of clusters is greater than the clustering threshold, the optimal number of clusters is obtained; If the number of clusters is less than or equal to the cluster threshold, execute S351; S355: Obtain multiple user groups with the same load characteristics according to the optimal number of clusters.

8. A load clustering method considering load characteristics and adjustable capacity according to claim 4, characterized in that: The S4 step includes: S41: performing dimensionality reduction processing on a plurality of user group data having the same load characteristics by using a principal component analysis technology to obtain dimensionality reduction data; S42: performing secondary clustering of user adjustable capabilities on the dimension reduction data through a self-organizing competitive neural network to obtain user adjustable capability clustering data.

9. A load clustering method considering load characteristics and adjustable capacity according to claim 4, characterized in that: Step S5 includes: The silhouette coefficient is selected as the evaluation index of the comprehensive clustering results of user load characteristics and adjustable capacity. If the silhouette coefficient is less than or equal to the silhouette coefficient threshold, the self-organizing competitive neural network is reconstructed; If the silhouette coefficient is greater than the silhouette coefficient threshold, the comprehensive clustering result of the user load is obtained.

10. A load clustering system considering load characteristics and adjustable capacity, characterized in that: A method for performing a load clustering method considering load characteristics and adjustable capacity as claimed in any one of claims 1 to 9, comprising: A construction and acquisition module, wherein the construction and acquisition module constructs a comprehensive analysis framework of the user's power load, collects user power load information data, and obtains the user's power load curve; A preprocessing module, which cleans and normalizes the user's power load curve to obtain a standardized data curve; A primary clustering module, wherein the primary clustering module performs primary clustering of user load characteristics on the standardized data curve to obtain multiple user groups with the same load characteristics; A secondary clustering module, wherein the secondary clustering module performs secondary clustering of user adjustable capabilities on a plurality of user groups having the same load characteristics to obtain user adjustable capability clustering data; A comprehensive clustering module identifies user adjustable capacity clustering data according to a silhouette coefficient threshold to obtain a comprehensive clustering result of user load.

Images