i we /z_ om .c to ho kp oc St ©i can be observed, and any other interesting structure can be examined. We will present a solution to this problem by first converting the data from each smart meter into a series of probability distributions, which are then used to compute pairwise distances between load profiles. the households are embedded in 2-D space to enable simple but informative plots to be constructed. Irish Smart-Meter Data to illustrate, we will use data collected during a smart-metering trial conducted by the commission for energy regulation (cer) in ireland. for demonstration purposes, we will use measurements of half-hourly electricity consumption gathered from 500 residential consumers over 535 consecumay/june 2018 tive days. every meter provides the electricity consumption between 14 July 2009 and 31 December 2010. many days in the series have periods of missing data. the cer data set does not account for energy consumed by heating and cooling systems. either the households use a different source of energy for heating, such as oil and gas, or a separate meter is used to measure consumption due to heating. installed cooling systems are not reported in the study. Data from two smart meters are shown as time-series plots in figure 1. While it is obvious that these meters have very different demand patterns, it is not possible to say much more; the time-of-day and day-of-week patterns are hidden due to the volume of data, and even the median demand is not clear from such plots. ieee power & energy magazine 19http://www.iStockphoto.com/z_wei