Aplicação de técnicas de previsão de séries temporais para determinação da vida útil de para-raios de óxido de zinco.

Abstract

Metal Oxide Surge Arresters (MOSAs) have been employed in order to increase the reliability, economy and continuity of operation of Electric Power Systems (EPSs). Considering the importance of the MOSAs for EPSs, it is necessary to study, develop and optimize techniques for forecating the MOSA lifetime. This paper presents an evaluative study of techniques that employ pure and hybrid models to forecast the behavior of Time Series (TSs), in order to apply those techniques in the estimation of MOSA lifetime. Although there are several techniques in the literature that use pure and hybrid models for forecasting, there is lack of designed techniques to forecast the MOSA lifetime. Among the sensitive indicators to the MOSA degradation level, the ones that can be used to form the TS are the harmonic content of the total leakage current and its resistive component, considering that these indicators are the most used by the EPSs. To perform this work, a database was initially built with TSs with information on the third harmonic component of the MOSA total leakage current. Then, the TS forecasting techniques were implemented on a computational platform. The selected TS forecasting techniques for implementation employ models based on artificial intelligence, with the focus of analysis being the Neuro-Fuzzy system, the Long Short-Term Memory network and the Support Vector Machine. The combinational model-based technique uses these three types artificial intelligence to perform TS behavior forecasting. Finally, the performance evaluation of each of the implemented techniques was accomplished using the adopted performance indicators, such as: mean square error, mean absolute error and mean absolute percentage error. Tests performed so far indicate that the implemented techniques present satisfactory results in the forecasting, especially in relation to the technique that employs a combinational model. Thus, an overview of the forecasting techniques of the MOSA lifetime was obtained, mainly contributing to the asset management of the EPS.