Adaptive phase shift control of thermoacoustic combustion instabilities using model-free reinforcement learning

Abstract

Combustion instability is a significant risk in the development of new engines when using novel zero-carbon fuels such as ammonia and hydrogen. These instabilities can be difficult to predict and control, making them a major barrier to the adoption of carbon-free gas turbine technologies. In order to address this challenge, we propose the use of model-free reinforcement learning (RL) to adjust the parameters of a phase-shift controller in a time-varying combustion system. Our proposed algorithm was tested in a simulated time-varying combustion system, where it demonstrated excellent performance compared to other model-free and model-based methods, including extremum seeking controllers and self-tuning regulators. The ability of RL to effectively adjust the parameters of a phase-shift controller in a time-varying system, while also considering the safety implications of online system exploration, makes it a promising tool for mitigating combustion instabilities and enabling the development of safer, more efficient carbon-free gas turbine technologies.