Machine Learning Approaches for Estimating the Efficiency of Combined Cycle Power Plant

Power plants are the main source of generating electricity. Maintaining the operation at its best possible thermal efficiency under the uncertainties of constantly changing environmental variables at various generating output loads is certainly a challenging task. Efforts and resources have been made to monitor and analyze the performance throughout the year along with periodic maintenance. Conventional analysis and monitoring methods require experts and engineers to closely evaluate rather complex sub-components with an enormous amount of data under the nonlinear nature of the complex systems from multiple measuring perspectives. In this paper, we investigate various machine learning techniques as the estimation tools for the thermal efficiencies of a combined cycle power plant in an effort to aid the monitoring process. The results depict that Random Forest performed best in regression techniques whereas Gradient boosting performed well in ensemble methods.