This paper tackles the Knowledge Discovery in Databases (KDD) technique in order to qualify the information collected by the field crews during the restoration process of the distribution network. This improvement of the information enables the utilization of Artificial Intelligence (AI) techniques to support decision making in investments planning, operation and maintenance of distribution systems. This qualification allows the use of KDD results to train a Bayesian Network (BN). The aim of the proposed BN is to support the performance diagnostic of electrical networks, promoting an indirect identification of forced outage causes. The analysis of the collected data during a forced outage indicates that the main purpose of field crews is a quick restoration of the network and, several times, the causes surrounding the outage events have a high level of subjectivity and uncertainty, turning impossible its direct identification. To illustrate the methodology, is presented a case with 570.000 events, where KDD provides a new environment - with a significant amount of data - more suitable to train and validate a BN to identification of forced outage causes.
Knowledge discovery in databases; Data mining; Database; Forced outage cause identification; Bayesian Network
