Outlier Detection for Hydropower Generation Plant

Abstract

A hydropower generation plant is a complex system and composed of numerous physical components. To monitor the health of different components it is necessary to detect anomalous behavior in time. Establishing a performance guideline along with identification of the critical variables causing anomalous behavior can help the maintenance personnel to detect any potential shift in the process timely. To establish any guideline for future control, at first a mechanism is needed to differentiate anomalous observations from the normal ones. In our work we have employed three different approaches to detect the anomalous observations and compared their performances using a historical data set received from a hydropower plant. The outliers detected are verified by the domain experts. Making use of a decision tree and feature selection process, we have identified some critical variables which are potentially linked to the presence of the outliers. We further developed a one-class classifier using the outlier cleaned dataset, which defines the normal working condition, and therefore, violation of the normal conditions could identify anomalous observations in future operations.