APPLICATION OF SELF-ORGANIZING MAPS FOR CLASSIFICATION AND FILTERING OF ELECTRICAL CUSTOMER LOAD PATTERNS

S.V. Verdu, M.O. Garc´a, C.S. Blanes, F.J.G. Franco, and A.G. Mar´n ´ ı ı

Keywords

Demand management, selforganizing maps, electrical customersegmentation, load patterns

Abstract

The objective of this research is to show the capability of the self-organizing maps (SOMs) to organize, filter, classify and extract patterns from distributor, commercializer, aggregator or customer electrical demand databases (the objective known as data mining). This approach basically uses – to reach the above-mentioned objectives – the historic load demand curves of each user. To get a better classification, some anomalous data – holidays, wrong measurements due to recorder failures – should be filtered before starting the map training. This preliminary step has been performed through an SOM map too. To show the proposed method in the paper only two typical medium users are studied on the filtering stage: an industry and a university both located in Spain. Subsequently, the filtering process is applied to a larger group of customers to finally prove the customer clustering capacity of SOM. The results clearly show the suitability of SOM approach to improve data management and to find easily coherent clusters between electrical users.

Important Links:



Go Back