Research on performance prediction of gas turbine air filtration system based on transformation-gated LSTM method

Abstract

The air filtration system is the only line of defense for outside air to enter the gas turbine. It is of great significance to study and predict the variation trend of air filtration system performance to ensure the safety, economy and reliability of gas turbine. In this work, the gas turbine air filtration system test is carried out, and the test data of fine filter differential pressure, ambient temperature and relative humidity are measured. Then the LSTM and transformation-gated LSTM (GT-LSTM) methods are utilized to predict the variation trend of air filtration system performance. Finally, the root mean square error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE), and symmetric mean absolute percentage error (SMAPE) are employed to evaluate the prediction effect quantitatively. The research results show that the RMSE, MAE, MAPE and SMAPE are only 0.0110, 0.0076, 3.3778 and 3.4655 by TG-LSTM method. The prediction error based on TG-LSTM method is smaller than LSTM method. Thus the prediction effect of TG-LSTM method is better than that of the LSTM method, and the TG-LSTM method is suitable for the performance prediction of gas turbine air filtration system.