Research on load data classification method based on the characteristics of electricity consumption behavior by extreme learning machine
Puhui Wang, Wei Zhang, Min Ma, Wanggang Jin, Xueyao Wang · 2023
In the background of the era of grid information intelligence, with the comprehensive development of power sales enterprises, the data mining work of electric power customers has been emphasized and grown rapidly. Therefore, this paper proposes an Extreme Learning Machine (ELM)-based algorithm for electricity customer behavior analysis. Initially, we conduct an analysis of typical load curves using multiple indicators. Firstly, we employ a feature selection strategy to extract the most relevant features from load curves, which enables us to classify and analyze customers' electricity consumption data effectively. Subsequently, we utilize the feature preference set as input and evaluate the accuracy of the training and test sets using various activation functions and different numbers of nodes in the hidden layer. This optimization process aims to determine the best input parameters for the ELM algorithm in customer behavior analysis. Finally, the arithmetic simulation experiments are carried out to compare and analyze with BP neural network. The results show that the load data classification method based on the electricity consumption behavior features of the extreme learning machine can meet the goal of flexible and interactive intelligent electricity consumption, and improve the efficiency of abnormal electricity consumption detection and identify the patterns of users' electricity consumption behavior through data mining and electricity consumption behavior analysis of electricity users' daily electricity consumption.