Multivariate data analysis in hydroelectric system maintenance: A decision evaluation case study

Abstract

Hydroelectric generation is comprised of complex systems specifically designed to meet the dynamic load. Forced outages and unscheduled maintenance activities severely limit the generation output and oftentimes create undesired environmental effects within the immediate Dam/Reservoir area as well as the downstream surroundings. The introduction of multivariate descriptive data analysis and multi-criteria decision making in the maintenance sphere of hydroelectric generation are designed to eliminate reactive preservation methodologies while economically dispatching the unit. Moreover, the implementation of a correlation matrix for powertrain and auxiliary electrical systems produce a general method to localize the outage cause to a component level. Statistical regression techniques were used to evaluate the differences of inconsistencies between maintenance practices and theoretical systematic preservation methods experienced in the hydroelectric generation industry. The regression forecasting model minimizes the risks typically encountered in systems maintenance and prioritizes capital-intense projects within the power production envelope. Additionally, the application will assist the decision-makers with systematic and orderly ranking of projects competing for scarce resources (labor, material, and funding) over a multi-year period in a constrained power production environment. The identification of inadequate performing assets and its cost effectiveness throughout the electrical footprint is an important tenet of the program.