A Two Stages Pattern Recognition for Time-of-use Customers based on Behavior Analytic by Using Gaussian Mixture Models and K-mean Clustering: a Case Study of PEA, Thailand

Abstract

Data and information become valuable possession in digital era where we are surrounded with big data. Data mining is supposed to be major and first process to tackle with big data. This study investigates featured features of Time-of-Use (TOU) based electricity customers using Gaussian mixture process. K-means clustering clusters TOU based electricity customer into various groups i.e., majority and minority consumption profile. Then, confidential interval (CI) corresponding with forecasted α-level confidential is formulated for each customer’s major load profile. The input data is collected from 1,000 PEA’s TOU customers during January to December 2016. Then, all individual consumption patterns of both working and nonworking day are grouping into 12 groups to be represented overall pattern of the sample of 1,000 TOU’s PEA customers. The outcome of this study shows that feature extraction with data clustering processes using could help to extract intrinsic features and formulate consumption patterns of metadata of TOU customers.