Numerical simulation of bioliquid fuels combustion

The world relies heavily upon fossil fuels as its major sources of energy for transportation, heating and electricity generation. Fossil fuel resources are, however, and its usage is accompanied by growing concerns over environmental implications such as impact on local air quality and the global problem of greenhouse gas emissions. Utilization of biomass, more specifically pyrolysis oil, was shown to have high potential to be used in a gas turbine to produce heat and electricity.

Pyrolysis oils, however, come with vastly different properties compare to that of normal liquid fossil fuels, and is shown to require a novel combustor design. CFD model to predict the behaviour of pyrolysis oil combustion is critical to deliver a proper combustor design. Modelling combustion of such a complex fuel, however, is not that easy. For that reasons, intermediate models which use simpler fuel such as ethanol and nheptane, are investigated. Important factors which affect the combustions process itself, such as droplet diameter distribution and droplet evaporation behaviour, are studied as well. Finally, the development of the actual pyrolysis model is done. All the simulations are performed using the OPRA's gas turbine combustor.

The main objective of this project is to develop a working model for pyrolysis oil combustion, taking into account only the evaporation process. Other process such as components polymerization, cracking, and char combustion are not considered Simulations have been done successfully for several operating parameters. The most important conclusions are: a) The model for pyrolysis oil combustion works well and it is able to mimic the physical behaviour of the fuel; b) It was found that ethanol and nheptane has similar evaporation constant, which approximately is ± 15 x 10-7 m2/s. Pyrolysis oil, however, evaporate three to four times slower compare to this value. The combustor design may have difficulty to accommodate such a low evaporation process; c) it was found that the models are sensitive to one of the input parameters which is estimated. The model, thus, can be improved by providing this input through experimental measurements.