Metodologia multicritério aplicada na priorização de serviços de manutenção em equipamentos do sistema de distribuição de uma concessionária de energia elétrica

Abstract

The uninterrupted and high-quality supply of electric power has been one of the main challenges faced by utility companies in the electricity sector, with the prioritization of system maintenance services being a key issue for mitigating failures and interruptions. In this context, this work proposes a methodology for assigning weights in multicriteria analysis problems, applied to a real case of maintenance prioritization for equipment in a utility’s distribution system. The methodology developed performs the optimal allocation of weights assigned to the criteria, defined based on the assessments provided by experts involved in the problem. Grounded in Decision Theory, the study describes the expert evaluation process using the Analytic Hierarchy Process (AHP), as well as the procedure used for ranking the alternatives—the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS), and the method employed to evaluate the rankings, Grey Relational Analysis (GRA). Specifically in the GRA stage, the dispersion of the Grey Relational Grade (GRG) of the first 150 samples from each ranking was analyzed using several statistical metrics. Among these metrics, the one that yielded results most aligned with the problem under study was correlation. Additionally, in the search for the best rankings, eight optimization algorithms were used together with statistical techniques aimed at selecting the most suitable model. The results obtained show that the aggregation method, the arrangement, and the proportion of the weights generated in the evaluation stage significantly influence the prioritization rankings. Furthermore, the rankings derived from the proposed technique far outperformed those obtained from a single decision-maker, from all decision-makers, and from the best-performing decision-makers involved in the study, increasing from a correlation classified as insignificant, with a value of –0.0819 in the first case, to a correlation classified as very high negative, with a value of –0.9429 after optimization. Given these results, the research offers a relevant contribution to the generation of more robust rankings by reducing the subjectivity inherent in individual evaluations and in traditional decision aggregation methods.