Investigation on Machine learning based fault detection and estimation in hydro turbines of industrial hydro power plant

Abstract

An industrial plant safety depends on identification, detection and estimation of faults. Appropriate fault diagnosis can result in higher productivity in process industries. Real-time fault detection is achieved using data-driven modelling of an industrial hydro power plant. It is still very difficult to simulate complicated and nonlinear systems, like hydro turbines, despite efforts to detect faults. The industrial power plant at Mettur was modelled in order to gather the data set. This article focusses on mathematical modelling of nonlinear hydro-turbines. Also, investigates the application of machine learning algorithms to detect and estimate faults in hydro turbines, which are critical components in industrial hydro power plants. The goal is to improve operational efficiency and prevent costly downtimes by identifying faults early. In order to classify healthy or faulty turbines both qualitatively and quantitatively, machine learning algorithms such as support vector machines and linear regression are used. The experimental results show that the performance measures are in good agreement with the results generated by the Support Vector Machine for the nonlinear models. The R squared value, which falls between 0 and 1, shows a better fit, while the Mean Absolute Percentage Error, which falls between 20 and 50%, indicates fair accuracy.