CN-121981765-A - Commercial virtual cluster recognition method for fuzzy clustering of electricity consumption data

CN121981765ACN 121981765 ACN121981765 ACN 121981765ACN-121981765-A

Abstract

The invention discloses a commercial virtual cluster identification method for fuzzy clustering of electricity consumption data, and belongs to the technical field of intelligent regulation and control of electric power systems. The method comprises the steps of extracting load vectors from processed electricity data based on the size of a rolling window, carrying out fuzzy clustering on the load vectors of all business users to obtain fuzzy clustering results, adjusting virtual clusters corresponding to all business users based on fuzzy membership matrixes of all business users, determining virtual cluster center load vectors corresponding to the adjusted virtual clusters, drawing a load curve based on the virtual cluster center load vectors, taking the load curve as a load mode of the adjusted virtual clusters, evaluating all the adjusted virtual clusters, outputting the load mode corresponding to the adjusted virtual clusters when the evaluation is passed, and carrying out power scheduling on all the adjusted virtual clusters based on the corresponding load modes. The method is suitable for intelligent aggregation and dynamic classification of large-scale commercial load data.

Inventors

WANG XIAOMING
BAI YUNLONG
ZHAO WENGUANG
WANG YUHANG
XU BIN
NI JINGYI

Assignees

国网安徽省电力有限公司电力科学研究院

Dates

Publication Date: 20260505
Application Date: 20251215

Claims (10)

1. A commercial virtual cluster identification method for carrying out fuzzy clustering on electricity consumption data is characterized by comprising the following steps: Step S1, acquiring electricity consumption data of commercial users, and cleaning and standardizing the electricity consumption data to obtain processed electricity consumption data; step S2, determining the size of a rolling window based on a preset sampling time interval, and extracting a load vector from the processed electricity data based on the size of the rolling window; S3, constructing a fuzzy clustering model, and performing fuzzy clustering on the load vectors of all commercial users to obtain a fuzzy clustering result; Step S4, constructing a fuzzy membership matrix of each commercial user based on a fuzzy clustering result, adjusting a virtual cluster corresponding to each commercial user based on the fuzzy membership matrix of each commercial user, determining a virtual cluster center load vector corresponding to the adjusted virtual cluster, and drawing a load curve based on the virtual cluster center load vector, wherein the load curve is used as the load mode of the adjusted virtual cluster; And step S5, evaluating each adjusted virtual cluster, outputting a load mode corresponding to the adjusted virtual cluster when the evaluation passes, and carrying out power scheduling on each adjusted virtual cluster based on the corresponding load mode.
2. The method of claim 1, wherein the step S2 is to determine a rolling window size based on a preset sampling time interval, and extract a load vector from the processed electricity data based on the rolling window size, wherein: Based on a preset sampling time interval, calculating the number of sampling points corresponding to 1 natural day, and taking the number of sampling points as the size of a rolling window; Determining a first sampling time interval based on the time span of the processed electricity data and the rolling window size, the first sampling time interval being equal to the time span of the processed electricity data divided by the rolling window size; and extracting data from the processed electricity data according to the first sampling time interval, wherein the extracted data form a load vector.
3. The method according to any one of claims 1-2, wherein in step S3, a fuzzy clustering model is constructed, and fuzzy clustering is performed on load vectors of all commercial users to obtain a fuzzy clustering result, wherein: the fuzzy clustering model is an objective function, and the objective function is as follows: Wherein, the Membership degree of the ith commercial user to the virtual cluster center of the jth virtual cluster in the fuzzy clustering result, m is a fuzzy coefficient, The virtual cluster center, which is the jth virtual cluster, N is the number of all commercial users, Is the weighted sum of squares of errors for all commercial users to all virtual cluster centers, For the number of virtual clusters, Load vector for the ith business user; And under the condition that the target is the minimum value of the objective function, fuzzy clustering is carried out on the load vectors of all commercial users by using the FCM algorithm, so as to obtain a fuzzy clustering result.
4. A method according to claim 3, wherein m is 1.5 to 2.5.
5. The method according to claim 3, wherein in the step S4, a fuzzy membership matrix of each commercial user is constructed based on the result of fuzzy clustering, wherein: the fuzzy membership matrix for commercial users is as follows: Matrix array Each element of (3) Representing the membership of user i to the jth virtual cluster.
6. The method as claimed in claim 5, wherein in the step S4, the virtual clusters corresponding to each commercial user are adjusted based on the fuzzy membership matrix of each commercial user, wherein: adjusting each business user to a virtual cluster with the largest membership value; Or each business user belongs to K virtual clusters at the same time, and for each business user, the degree of the business user belonging to each virtual cluster is determined according to the weight corresponding to the membership degree of the business user corresponding to each virtual cluster; Or each business user belongs to a plurality of virtual clusters at the same time, and for each business user, matrix elements with values exceeding a preset threshold value in the fuzzy membership matrix corresponding to the business user are determined to be selected matrix elements, and virtual clusters corresponding to the selected matrix elements are determined to be the virtual clusters to which the business user belongs at the same time.
7. The method according to claim 6, wherein in the step S4, a virtual cluster center load vector corresponding to the adjusted virtual cluster is determined, a load curve is drawn based on the virtual cluster center load vector, and the load curve is used as the adjusted load pattern of the virtual cluster, wherein: The calculation formula of the virtual cluster core load vector is as follows: Wherein, the Is the virtual cluster center of the kth adjusted virtual cluster, n is the total number of time points, Respectively at membership degree vectors The load value of the kth virtual cluster center at the 1 st to nth time points under action, For the ith business user pair Is used for the degree of membership of the group (a), Is the load vector of the ith business user at time t. The load curve comprises a load vector based on the virtual cluster center The daily load change curve, the peak-valley change characteristic curve and the time sequence load fluctuation curve are drawn and are used for representing typical power consumption mode characteristics of each virtual cluster in the whole time dimension.
8. A commercial virtual cluster recognition device for carrying out fuzzy clustering on electricity consumption data is characterized by comprising: The data acquisition module is configured to acquire electricity consumption data of commercial users, and clean and standardize the electricity consumption data to obtain processed electricity consumption data; The characteristic extraction module is configured to determine the size of a rolling window based on a preset sampling time interval, and extract a load vector from the processed electricity data based on the size of the rolling window; The clustering module is configured to construct a fuzzy clustering model, and perform fuzzy clustering on the load vectors of all commercial users to obtain a fuzzy clustering result; The load calculation module is configured to construct a fuzzy membership matrix of each commercial user based on a fuzzy clustering result, adjust virtual clusters corresponding to each commercial user based on the fuzzy membership matrix of each commercial user, determine virtual cluster center load vectors corresponding to the adjusted virtual clusters, and draw a load curve based on the virtual cluster center load vectors, wherein the load curve is used as a load mode of the adjusted virtual clusters; The scheduling module is configured to evaluate each adjusted virtual cluster, output a load mode corresponding to the adjusted virtual cluster when the evaluation passes, and perform power scheduling on each adjusted virtual cluster based on the corresponding load mode.
9. An electronic device, the device comprising: at least one processor, and A memory communicatively coupled to the at least one processor, wherein, The memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-7.
10. A non-transitory computer readable storage medium storing computer instructions for causing the computer to perform the method of any one of claims 1-7.

Description

Commercial virtual cluster recognition method for fuzzy clustering of electricity consumption data Technical Field The invention belongs to the technical field of intelligent regulation of power systems, and particularly relates to a commercial virtual cluster identification method for fuzzy clustering of power consumption data. Background Along with the development of modern power systems to intelligent, digital and multi-main-body cooperation, how to realize efficient load aggregation and optimal regulation at the demand side becomes an important subject for supporting the safe and stable operation of the power systems. In particular, in commercial user groups, the traditional static division and experience rule type management method cannot meet the requirements of fine regulation and real-time optimization due to various building types, complex operation rules and remarkable load fluctuation. Aiming at the problem, a virtual cluster model which can dynamically describe business load characteristics and reflect similarity and difference among users is constructed, so that the virtual cluster model becomes an important technical direction for realizing intelligent load management. In the existing research, load clustering is often used for mining power consumption modes of different types of users so as to support applications such as demand response, price adjustment, energy efficiency optimization and the like. However, the traditional hard clustering method (such as K-Means) has significant shortcomings when facing diversity and time sequence of commercial loads, namely firstly, hard partitioning assumes that each user only belongs to a certain fixed category and is difficult to reflect ambiguity and intersection of load behaviors in reality, secondly, load data have obvious random fluctuation and periodic variation characteristics, a clustering index purely based on Euclidean distance or a correlation coefficient cannot fully describe dynamic characteristics of the load data, thirdly, commercial users have differences in geographical positions, amateur scale, energy utilization modes and the like, and a traditional clustering result is difficult to be directly used for establishing a subsequent virtual cluster construction and regulation strategy. Disclosure of Invention The invention provides a commercial virtual cluster identification method for fuzzy clustering of electricity consumption data, which is used for solving the technical problems of insufficient user characteristic depiction, over-strong cluster partition rigidity and low suitability of a regulation strategy in the conventional load clustering technology. The first aspect of the invention provides a commercial virtual cluster identification method for carrying out fuzzy clustering on electricity utilization data, which comprises the following steps of. Step S1, acquiring electricity consumption data of commercial users, and cleaning and standardizing the electricity consumption data to obtain processed electricity consumption data; step S2, determining the size of a rolling window based on a preset sampling time interval, and extracting a load vector from the processed electricity data based on the size of the rolling window; S3, constructing a fuzzy clustering model, and performing fuzzy clustering on the load vectors of all commercial users to obtain a fuzzy clustering result; Step S4, constructing a fuzzy membership matrix of each commercial user based on a fuzzy clustering result, adjusting a virtual cluster corresponding to each commercial user based on the fuzzy membership matrix of each commercial user, determining a virtual cluster center load vector corresponding to the adjusted virtual cluster, and drawing a load curve based on the virtual cluster center load vector, wherein the load curve is used as the load mode of the adjusted virtual cluster; And step S5, evaluating each adjusted virtual cluster, outputting a load mode corresponding to the adjusted virtual cluster when the evaluation passes, and carrying out power scheduling on each adjusted virtual cluster based on the corresponding load mode. Preferably, the step S2 is to determine a rolling window size based on a preset sampling time interval, and extract a load vector from the processed electricity data based on the rolling window size, wherein: Based on a preset sampling time interval, calculating the number of sampling points corresponding to 1 natural day, and taking the number of sampling points as the size of a rolling window; Determining a first sampling time interval based on the time span of the processed electricity data and the rolling window size, the first sampling time interval being equal to the time span of the processed electricity data divided by the rolling window size; and extracting data from the processed electricity data according to the first sampling time interval, wherein the extracted data form a load vector. Preferably, in the step S3, a fuzz