Ballester J., Hernandez R., Sanz A., Smolarz A.

There is a growing interest in using new fuels (e.g., gasification products, H2-enriched gases) for different combustion applications, including industrial gas turbines as one of the most promising options. However, this poses some technological challenges that need to be addressed. In particular, the difficulties to achieve an optimal and stable combustion increase when burning those alternative fuels, usually characterised by low calorific values and fluctuating properties. The present work is aimed at characterising the combustion performance of different fuels and at exploring novel methods for the permanent monitoring and optimisation of the process. Natural gas and its blends with CO2 and H2 were burnt in an atmospheric, lean premixed combustor. In the first place, their combustion characteristics are compared in terms of CO and NOx emissions and stability ranges; particular attention is given to the magnitudes and frequency spectra of the fluctuations in pressure and emitted radiation as a function of equivalence ratio. In the second place, different methods for real-time monitoring of the combustor are explored. These methods are based on the analysis of average and fluctuating signals from fast pressure transducers and radiation sensors. The capabilities to identify the actual combustion conditions are evaluated using artificial neural networks. Finally, some optimisation trials are presented, which are designed to minimise a cost function combining pollutant emissions and pressure fluctuations. The tests showed that a direct application of minimisation methods entails significant risks of reaching unstable conditions. However, the combination of those algorithms with a continuous monitoring of fluctuating magnitudes has demonstrated to enable a safe optimisation process, by preventing the onset of combustion oscillations during the search for minimal emissions.