Predicción de parada de máquinas generadoras en una central hidroeléctrica por medio de minería de datos

Abstract

A methodology is developed to predict shutdowns of generating machines in a hydroelectric power station by using the platform Weka and the algorithm J48. A file was built with 300 real data registrations and 11 variables consisting of a single dependent variable (machine trip) and ten independent variables: Temperature, water flow, regulatory pressure, pipeline pressure, flow rate of the reservoir level, generated load, frequency, oil temperature, and climate. Using XRealStats tools, a Pearson correlation analysis was performed between each of the independent variables and the dependent variable. The results showed that it is possible to predict machine trips with a success rate of 94%. It is concluded that the classification tree generated in this research predicts future machine trips with over 94% accuracy.