Autonomous Bayesian neural models for fault diagnosis in transmission lines

Abstract

Fault diagnosis constitutes a problem in electric power systems with relevant economic impact for operators and stakeholders. Artificial neural networks have been proposed in the literature to deal with this problem in a significant number of applications. However, most proposals are based in ad-hoc structure specification and model regularization, which compromises the direct application of the algorithms to other transmission systems. Given the above, this paper will present the application of Autonomous Bayesian Neural models to fault diagnosis in electric power system transmission lines. This modeling strategy includes automatic and analytical procedures for input selection, model specification and regularization without the need of a validation set. The system performance will be analyzed considering data from a real transmission line of the Brazilian interconnected system.