Um estudo sobre o procedimento de despacho e sua influência na degradação dos equipamentos das centrais hidrelétricas

Abstract

Hydroelectric power plants (HPP) are frequently used in peak load hours to stabilize power grid operations. For this reason, peak-load units can experience between 10 and 12 start-stop cycles per day. During the starts, the turbines of the HPP operate in transient states, which may increase the wear and tear of the components. Therefore, more start-stop cycles increase maintenance costs. In this scenario, traditional maintenance techniques may be re-evaluated, and new proactive options should be more appropriate and efficient. For this, it is necessary to process and analyze data from the equipment to extract useful information that can help in equipment maintenance. This study aimed to fit time series models to a waveform data from the rolling bearing of an HPP. In order to do so, the Autoregressive Moving Average model (ARMA), the Autoregressive Fractionally Integrated Moving Average model (ARFIMA) and Autoregressive Conditional Heteroskedasticity model (ARCH) were used. This technique was applied to the original waveform data and to the post Wavelet Transformation data. The results obtained showed that the original data were better adjusted than the filtered, however, the statistical test results were proven to be unsatisfactory, due to the quasi brownian behavior of the data.