Advancing AI Modeling for Prediction of Safety Parameters for Combustion of Hydrogen/Syngas/Natural Gas Mixtures

Abstract

Abstract One promising way supporting efforts to decarbonize energy production is to hybridize gas turbine fuels, and in particular to make increasing use of hydrogen. The use of alternative decarbonated or renewable fuels will significantly modify safety parameters. New simulation tools need to be developed that can rapidly assess and forecast the safety characteristics of different reactive fluids. This paper focuses on the development of a learning model to estimate Auto Ignition Temperature and Auto Ignition Delay Times (AIT and AID) during the combustion of several mixtures of dihydrogen, natural gas, and syngas. A numerical procedure has been developed, which generates automatically a dataset of AIT covering a wide range of operating conditions. The dataset contains 50,000 points and has been simulated using the comprehensive combustion model from the literature developed at National University of Ireland at Galway (NUIG). In order to fit optimally all the hyperparameters of a neural network model, several methods of optimization have been tested and are discussed. The model developed is capable of predicting AIT for pressures between 1 and 50 atm, temperatures between 200°C and 700°C and equivalence ratio between 0.2 and 5. For both the training and testing datasets, the average value of the correlation coefficient was above 99.95%, the Mean Absolute Error (MAE) and the Mean Square Error (MSE) around 0.003 and lower than 1e−4, respectively. The final machine learning model shows a high degree of robustness in reproducing the results of the detailed model over all the parameter domain studied. It was used eventually for the analysis of a typical case study of the risk of auto-ignition in pipes when gases are fed to a gas turbine.