A hybrid intelligent system for alarm processing in power distribution substations

Article type: Research Article

Authors: de Souza, J.C. Stacchini^{a; *} | Do Coutto Filho, M.B.^a | Freund, R.S.^b

Affiliations: [a] Fluminense Federal University, Electrical Engineering Department, Institute of Computing, Rua Passo da Pátria 156 – Bloco E – 3° andar São Domingos Niterói – RJ – 24210-240, Brazil | [b] LIGHT Services of Electricity, Automation Department, Av. Marechal Floriano 168, Centro, Rio de Janeiro – RJ – 20080-002, Brazil

Correspondence: [*] Corresponding author. Tel.: +55 21 26295472; Fax: +55 21 26295627; E-mail: [email protected]

Abstract: This application paper presents an intelligent system for alarm processing and fault location in power substations. A hybrid model is constructed using rule-based systems and an artificial neural network. Incoming alarms are initially handled by an input expert module, responsible for presenting selected input information to an artificial neural network. This network processes the alarm information and provides classifications that will be further processed by an output expert module, which is responsible for obtaining final diagnoses that will be presented to the substation operator. The concept of typical substation is presented and discussed. The alarm processing automation obtained with the proposed model enhances the reliability and quickness of power system operators' decision making. Simulation results using data from a real substation show that the proposed model presents excellent performance under many different situations.

Keywords: Expert systems, neural networks, fault location

DOI: 10.3233/HIS-2010-0109

Journal: International Journal of Hybrid Intelligent Systems, vol. 7, no. 2, pp. 125-136, 2010