Power analytics of power data

Abstract

The world has been working hard on technology to achieve the goal in order to reduce energy consumption. The widely used method of achieving that is making use of sensors to analyze and monitor the power consumption of electrical devices in a unit or building. However, although sensors are cheap, the maintenance and configuration tends to be tough and complicated. This project is to study on how electrical appliance can be identified through power consumption signal, by decomposing the signal into different phases.

To do this, Empirical Mode Decomposition has been utilized to enable closer look into power consumption signal. Some unique patterns can be observed through the decomposed signals. The observation is taken further by turning the time series graph of decomposed signals into frequency domain, via Fourier Transform technique. The unique features are then collected as knowledge base and a classification algorithm is used to predict the identity of electrical appliances used in an unknown dataset.