Application of clustering algorithms and self organising maps to classify electricity customers

Load Time-Series Classification Based on Pattern Recognition Methods

“…Similarity matrix indicator (SMI) (Chicco et al, 2003b), which is defined as the maximum off-diagonal element of the symmetrical similarity matrix, whose terms are calculated by using a logarithmic function of the Euclidean distance between any kind of class representative load curves:…”

“…• the "modified follow the leader" (Chicco et al, 2003a;-,2003b;-, 2006), • the self-organizing map (Beccali et al, 2004;Chicco et al, 2004;-, 2006 ;Figueiredo et al, 2003;Lamedica et al 1996;Verdu et al, 2003), • the k-means Figueiredo et al, 2003), • the average and Ward hierarchical methods (Chicco et al, 2004;-, 2006;Gerber et al, 2003) and • the fuzzy k-means (Chicco et al, 2004;-, 2006;Gerber et al, 2003;). All the above methods generally belong to pattern recognition techniques (Theodoridis & Koutroumbas, 1999).…”

“…• the mean index adequacy (Chicco et al, 2003a;-, 2003b;, • the clustering dispersion indicator (Chicco et al, 2003a;-, 2003b;-, 2006), • the similarity matrix indicator (Chicco et al, 2004), • the Davies-Bouldin indicator (Beccali et al, 2004;-, 2006;Gerbec et al, 2004;, • the modified Dunn index , • the scatter index ) and • the mean square error (Gerbec et al, 2003;.…”

“…In the second stage, classification of customers is performed by the same methods and measures, together with the representative load patterns of customers being obtained from the first stage. The basic contribution of this methodology is that its first stage enables the modification of the representative day, such as the most populated day, and avoids the a priori definition of a single day or the "mean" day of a specific time period (as it is suggested by previously published methodologies (Chicco et al, 2002 ;-,2003a;-,2003b;-,2006;Figueiredo et al, 2003;Gerbec et al, 2003 ;). The results of the second stage provide valuable information for electricity suppliers in competitive energy markets.…”

Robust Smart Meter Data Analytics Using Smoothed ALS and Dynamic Time Warping

“…Alternatively, the customer classification problem can be solved by using data mining, frequency-domain data [11,12], etc. The most commonly used respective adequacy measures are the mean index adequacy, the clustering dispersion indicator, the similarity matrix indicator, the Davies-Bouldin indicator, the modified Dunn index, the scatter index, and the mean square error [13].…”