MorphingProjections: Interactive Visualization of Electric Power Demand Time Series

Abstract

MorphingProjections is an interactive tool to explore time series of electric power demand in a public building. The core idea of the interface is a smooth continuous transition morphing between several 2D views, each providing a different insight on the data. For power demand analysis three types of encodings were considered: a) manifold learning projections, that result from mapping 24-dim day patterns on 2D points organized according to similarity of demand profile; b) "clock-type encodings" that give insight into daily, weekly and yearly periodicities and c) specific encodings, providing insight into calendar-dependent demand patterns. Morphing allows the user 1) to track interesting points or selections between conceptually different views, and 2) to produce new views by blending basic ones. The combination of morphing and other interaction elements such as zoom, pan and multiple selection, results in a highly dynamic interface that allows the user to link and combine the demand information in ways unseen in previous approaches.