Desenvolvimento e aplicação de modelos de regionalização bayesianos para a análise dos índices de eficiência operacional e dos indicadores de duração equivalente de interrupção por unidade consumidora das empresas brasileiras de distribuição de energia elétrica

Abstract

In the economy segments that operate under a natural monopoly condition, the absence of competition among suppliers can generate unwanted results for consumers, such as low-quality services at a high cost. To avoid this scenario, the regulatory agency’s presence, which must determine the process parameters and evaluate these companies’ results, is usual. In the context of electricity distribution in Brazil, this role is exercised by the National Electric Energy Agency (ANEEL). Companies’ operating parameters and performance indices are affected by manageable and non-manageable variables related to their operating environment. These environmental effects are important in Brazil, given the breadth and heterogeneity of its territory. The first part of this work explores the effect of environmental conditions’ heterogeneity and its impact on electric energy distribution companies’ efficiency scores. To circumvent the problem, the Bayesian method of regionalization is implemented for spatial analysis of efficiency indexes, allowing the creation of contiguous regions that present more homogeneous environmental conditions. This methodology was proposed for the epidemiology problem and was adapted to the proposed problem. The second part of the work extends this methodology to include a spatial regression model where, in addition to the number and positions of the clusters, each covariate’s impact on each one can be estimated. The updated methodology is used to analyze the Duração Equivalente de Interrupção por Unidade Consumidora (DEC) index, an important performance indicator in the distribution sector. The results demonstrate the possibility of estimating the number of clusters, their positions, and the regression coefficients associated with the variables that impact the indicator.