Power load characteristic data mean value clustering method and system based on cloud computing

Abstract

The invention relates to a power load characteristic data mean clustering method and system based on cloud computing. The method comprises the following steps: firstly, decomposing a load sample intoK classes; calculating a Zth-class mean value initial vector, and then randomly assigning a Zth-class clustering center; calculating the spatial distance between a jth load vector and an hth load vector in a Zth-class load; calculating an average distance between the two power load characteristic data points; setting a threshold, and for the Zth-class load, calculating the distance between each load vector in the class and the randomly specified clustering center of the class; and if the distance between each load vector and the randomly set clustering center is smaller than the threshold, indicating that the current clustering center is properly selected, and using the clustering center to represent the power consumption level and hierarchy of the class, otherwise, randomly setting one clustering center again, and repeating the calculation. The method is good in clustering effect and high in clustering efficiency.

Description

technical field [0001] The invention relates to the technical fields of electric power system and cloud computing, in particular to a method and system for clustering the mean value of electric load characteristic data based on cloud computing. Background technique [0002] The clustering of effective power load characteristics can timely understand the changing law of power load, and can provide accurate data support for power load forecasting, smart power price, peak shifting, and overall management for smart grids. [0003] For the power load clustering problem, the traditional method is to select a fixed number of clusters and use a more classical algorithm for clustering. For example, in the hierarchical clustering algorithm, it is necessary to perform classification tasks on the power load feature data set at different levels, treat all power load samples as a single class, and classify each sample into a class, and then classify the nearest The two categories are mer...

AI Technical Summary

Problems solved by technology

[0004] At present, the traditional clustering algorithm has the following deficiencies for the characteristic data of electric power load: (1) the selection of the threshold often cannot truly reflect the characteristics of the characteristic data of electric load; (2) the clustering of the characteristic data of electric power load The Euclidean distance between the sample and the sample center does not fit well; (3) The traditional clustering algorithm needs the initial cluster center to carry out the classification task. If the selection of the initial value is not suitable, a good clustering effect will not be obtained; (4) The real power load characteristic data is a huge amount of basic electrical parameter data, which cannot be analyzed and processed by traditional clustering algorithms

Images