Electricity consumption multielement space clustering method based on Getis-Ord Gi*

Abstract

The present invention provides an electricity consumption multielement space clustering method based on the Getis-Ord Gi*, and relates to a clustering method. At present, the clustering method does not take the consideration of the generated gathering caused by the variable space dependence relation to cause an inaccurate clustering result and not objective reflection of electricity consumption hot spot and cold spot gathering conditions of multi-industry electricity consumption areas. The technical scheme of the electricity consumption multielement space clustering method comprises the stepsof: 1) constructing a space dependence relation for a located area, and determining a space weight; 2) performing spatial autocorrelation analysis for each variable to determine an existing space dependence relation; 3) converting attribute variables to Gi* statistical variables with space elements, and reconstructing a multi-variable space matrix; and 4) employing a contour line coefficient method to the matrix to determine an optimal clustering number, and employing the k-means clustering method to perform clustering. The electricity consumption multielement space clustering method solves the space clustering problem of multiple variables so as to effectively improve the variable clustering effect and objectively reflect the electricity consumption hot spot and cold spot gathering conditions of the multi-industry electricity consumption areas.

Description

technical field [0001] The invention relates to a spatial clustering method, in particular to a multivariate spatial clustering method for electricity consumption based on Getis-Ord Gi*. Background technique [0002] In the past, the analysis of electricity consumption was mostly concentrated in a single industry. However, the electricity consumption of different industries may be related to each other and reflect the local economic development in an overall manner. Analyzing the agglomeration and dispersion of power consumption in different industries in the power consumption area is of great help to power distribution companies in formulating power consumption policies for the area. Because electricity consumption in similar areas can learn from each other, a city's electricity consumption policy can be transplanted to other cities that are similar to it. Generally, unsupervised clustering methods are used for the analysis of multivariate electricity consumption industrie...

AI Technical Summary

Problems solved by technology

Generally, unsupervised clustering methods are used for the analysis of multivariate electricity consumption industries. The common methods are segmentation clustering, hierarchical clustering, density-based clustering and grid-based clustering methods, but these methods ignore different The spatial characteristics of industrial electricity consumption, because the spatial characteristics of electricity consumption will produce the phenomenon of the first law of geography "everything is related, but similar things are more closely related", so the electricity consumption will be due to the proximity between regions Aggregation

However, the general clustering method does not take into account the agglomeration due to the spatial dependence of variables. The clustering generated by ignoring this spatial relationship is incomplete, resulting in inaccurate clustering results and cannot objectively reflect the multi-industry electricity consumption area. Agglomeration of hot and cold spots of electricity consumption

Images