Discovering Residential Electric Usage Patterns: A Data Mining Approach

Document Type

Thesis

Lead Author Type

CIS Masters Student

Advisors

Dr. Jamal Alsabbagh, alsabbaj@gvsu.edu

Committee Members

Dr. Yonglei Tao, taoy@gvsu.edu; Dr. Jerry Scripps, scrippsj@gvsu.edu

Embargo Period

12-9-2010

Abstract

The goal of this research was to discover patterns in hourly residential electricity usage. Understanding such patterns will allow management at the utility to design and offer appropriate incentives to its customers to influence their usage habits and, consequently, reduce the costs associated with providing electricity. Hourly electric consumption data was acquired for a subset of residential customers within the Holland Board of Public Works’ (Holland BPW) service territory in Holland, Michigan. SQL Server 2005 was used for data compilation and pre-processing. Machine learning algorithms were applied using R, a free software environment for statistical computing and graphics, with Kohonen Self Organizing Maps (SOMs) employed for discovering usage patterns within the time-series data. The resulting discoveries offered useful insight to management at the Holland BPW and form a solid foundation for further investigation

This document is currently not available here.

Share

COinS