Exhaust Temperature Prediction for Gas Turbine Performance Estimation by Using Deep Learning

Gas turbines are used to generate electricity in thermal power plants and are also used as a backup for renewable energy. Recently, following various environmental regulations, interest in technology for predicting the state of gas turbines for power generation and controlling the emission of air pollutants such as nitrogen oxides is increasing. Indices for predicting the state of a gas turbine include the turbine inlet temperature and exhaust temperature. Predicting the exhaust temperature according to the fluctuating operation method makes it possible to know the efficiency and condition of the gas turbine and enables more efficient operation. In this study, deep learning is used to predict the exhaust temperature of a gas turbine. The gas turbine data set consists of various variables, and by learning the data, it learns the trend of irregular changes in the exhaust temperature. A deep learning model is constructed by combining CNN and RNN algorithms, which are accessible for time series prediction. In particular, LSTM and GRU algorithms among RNN algorithms are applied respectively to obtain a more robust and accurate model. Accurate prediction of exhaust temperature can contribute to diagnosing the condition of a gas turbine and operating it efficiently.