From number to fire: supercomputing accelerates hydrogen gas turbine development

A KAIST research team conducted extreme high-fidelity simulations on a supercomputer to predict and suppress the onset of combustion instability in hydrogen gas turbines, or hydrogen fire.

Article | Spring 2022

Have you watched the movies ‘Matrix’ or ‘Ready Player One’? They both features virtual reality. Such technology is still far from reality, as accurate physics simulations at the scale of the living world will not be within reach for several generations. However, at smaller sizes, VR is possible. Have you seen how a smoker blows out a smoke and seen small wrinkles inside the smoke? That is about the size at which computer simulations can accurately reproduce all details. Figure 1(a) shows an experimental image of gas issuing from a turbulent jet (similar to smoke); Figure 2(b) shows a simulation image of a similar condition. Due to the different color schemes, these two images do not look exactly the same, but the simulations can actually reproduce all complex flow features.

 

These extreme high-fidelity simulations can help fight global warming. Around 60 % of our electricity is currently generated by burning fossil fuels. Hence, carbon-free power generation systems such as hydrogen gas turbines need to be developed quickly to replace conventional systems. However, changing fuels is not a simple task; as drivers know, one cannot use diesel fuel in a gasoline car.

In developing hydrogen gas turbines, preventing onset of combustion instability is a key technological challenge. Combustion instability is a resonance phenomenon caused by unstable flames, and it can lead to system failure. In this study, extreme high-fidelity simulations were employed to solve the challenge.

A KAIST research team used supercomputers to solve by brute force the Navier-Stokes equation and chemical kinetics so that no ad-hoc turbulence models would need to be used (see Figure 2(a)). The simulations revealed that hydrogen flames are more susceptible to topological changes of flame geometry than are methane flames, leading to different flame transfer functions.

 

The obtained hydrogen flame characteristics are fed into a global modeling tool to predict the onset of combustion instability. These predictions help hydrogen gas turbine manufacturers to design systems that can avoid the resonance frequencies and withstand the predicted oscillation amplitudes.