Prediction of Power Output for Combined Cycle Power Plant Using Random Decision Tree Algorithms and ANFIS

Abstract

This paper presents methods for prediction of the power output of the combined cycle power plant (CCPP) with a full load. A dataset comprising 9568 samples include measurements of ambient temperature (AT), atmospheric pressure (AP), relative humidity (RH), exhaust steam pressure, i.e. vacuum (V) and power output of the CCPP (EP). The research was done two folded: Using all features and the reduced set of features. Random Forest, Random Tree, and Adaptive Neuro Fuzzy Inference System (ANFIS) were used for regression. The performance of the methods studied in both folds showed that the best obtained results are gained using Random Forest. Results obtained on all features showed (Root Means Square Error) RMSE of 3.0271 MW, while feature selection leads to the RMSE of 3.0527 MW and Correlation coefficient (CC) of 0.9843, both obtained on 90% Percentage split.