Detecting defaulters for consumed electric energy with neural network clustering

Abstract

There are two main fields in the business activity of electric power industry: electric energy distribution and its collection. The first field includes the building and maintenance of electric power facilities according to the needs of consumers with the appropriate strategy of electric energy distribution in the situations of extremely poor and brisk demand. The second field, electric energy collection exerts influence on liquidity and business performance. This second field, electric energy collection, has priority in electric power industry, and it is quite understandable. The work deals with the application of neural network clustering analysis for detecting defaulters for consumed electric energy in the territory of the City of Subotica in the period of 2001 to 2006. The results of applying this Data Mining technique are precious for the Subotica Electric Power Distribution Enterprise because in this way a campaign can timely begin for regular payment of electric energy for those consumers being late to pay in the last five years. I.INTRODUCTION. THE SPECIFICITIES OF ELECTRIC ENERGY ACTIVITY All electric energy distributors have, besides the problems of electric energy distribution, common and ever lasting problem of control, collecting money for consumed electric energy and distribution management. However, electric power distribution enterprises are occupied by the biggest problems because they have many consumers and the longest and most involved network system. It is also known fact that requirements for electric energy change every year very dynamically ( increase per year is between 10% and 15 %). At the same time, electric energy is very specific because it cannot be accumulated and stored. Electric energy that is consumed must be permanently reproduced in the network by the generators. Aggregate power of all generators in one national network is limited disregarding how powerful it is. There are temporal intervals when the requirements for electric energy are so expressive that generators are not available to meet the needs of all consumers and than some restrictions in delivering electric energy are done. It would be desirable if electric power industry could independently manage electric energy delivery in the territory it covers, especially in critical cases of shortage of electricity, and in cases of its surplus. Therefore, it is necessary to know all dimensions of consumers' behavior and needs, to forecast their requirements and anticipate their future activities in order to satisfy their requirements and increase income of the enterprise.

Sistema Inteligente Híbrido Para Classificação Do Perfil De Pagamento Dos Consumidores Não-Residenciais De Energia Elétrica

“…Nessas campanhas também se busca um resultado positivo para a imagem da empresa para a população atendida em sua área de concessão. Depende, naturalmente, de um aporte de gastos com mídias de grande alcance (televisão, rádio, jornais, etc. Nos últimos anos, estudos têm sido desenvolvidos propondo sistemas que detectem e minimizem perdas não-técnicas (Araujo, 2007;Bastos, 2011;Cabral et al, 2006;Calili, 2005;Carvalho et al, 2014;Costa et al, 2013;Dias, 2006;Eller, 2003;Faria, 2012;Huang et al, 2013;Jiang et al 2002;León et al, 2011;Medeiros, 2013;Monedero et al, 2012;Nagi et al, 2008;Nizar et al, 2006;Nizar et al, 2008;Ortega, 2008;Penin, 2008;Petkovic & Balaban, 2008;Queiroga, 2005;Ramos, 2014;Ramos et al, 2011;Reis Filho, 2006;Ribeiro et al, 2012;Souza et al, 2013;Trevisan, 2014 Bayesiano Simples (Naive Bayes) como técnicas soft computing que podem ser aplicadas na segmentação do mercado consumidor.…”