Analysis of Residential Household Energy Consumption Using Smart Meter Data

Abstract

The need to change the source of electricity generation is apparent in the effect of climate change on the environment. Asides from the source change to renewable energy, the necessity for residents to understand their consumption rates and patterns is paramount to help reduce CO2 emission and thereby reduce climate change.This thesis discusses and implements the machine learning algorithm; K-Means clustering method, on a dataset derived from a town in Norway. The dataset is split into various features, to reveal the cluster and consumption patterns, their peak, off-peak, as well as mid-peak periods, in order to identify times where energy wastage can be minimized.It also experiments and compares two other algorithms; Hierarchical clustering and DBSCAN method, against the K-Means method, showing their differences and similarities,thereby deciding which algorithm is best suited for clustering the provided dataset