Prediction of operating characteristics for industrial gas turbine combustor using an optimized artificial neural network. (15th December 2020)

• Record Type: Journal Article
• Authors: Park, Yeseul; Choi, Minsung; Kim, Kibeom; Li, Xinzhuo; Jung, Chanho; Na, Sangkyung; Choi, Gyungmin
• Is Part Of: Energy. Volume 213(2020)
• Publication Date: 2020-12-15
• Subjects: Gas turbine combustor -- Operating prediction -- Neural network -- Predict combustion mode -- Artificial intelligence -- Intelligence digital power plant; Power resources -- Periodicals; Power (Mechanics) -- Periodicals; Energy consumption -- Periodicals; 333.7905
• DOI: 10.1016/j.energy.2020.118769
• Languages: English
• ISSNs: 0360-5442

Abstract

Abstract: In this study, the operating characteristics of a gas turbine combustor are predicted using real-time data from industrial gas turbines. The turbine exhaust temperature (TET) and major gas turbine design parameters are used as input parameters to predict the combustor operation characteristics such as fuel mass flow, turbine inlet temperature, fuel distribution of each nozzle, NOx, operating pressure of combustor, and inlet air temperature of combustor. The sensitivity analysis of input parameters is conducted to optimize predictive neural network structure. The average predicted root mean square error (RMSE) is below 0.02296. Also, for the expandability of the predictive model, 27 turbine exhaust temperature data changes to the mean/median/mean, standard deviation (std)/mid, and std are used as input temperature data. The case that uses the mean TET (TET_mean) shows the highest accuracy. The biggest influence on the prediction error is that when there is a sudden change in operation in a short time, the prediction error becomes large. The peak error occurs at start-up and shutdown process and the nitrogen oxides emission (NOx ) has the largest peak RMSE is 0.489. The RMSE from 0.2 to 0.5 occurs just for 20 s at startup and shutdown procedure. Highlights: The predictive model was developed based on real-time operating data of gas turbine. The operation characteristics of gas turbine combustor are predicted using ANN. The model was optimized through sensitivity analysis, and the mean RMSE is 0.02296. The peak error of NOx is largest and occurs at start-up and shutdown for 20 s. The model can be used for monitoring and diagnosing combustor operation.