TY - JOUR
T1 - Incorporating uncertainty in data driven regression models of fluidized bed gasification
T2 - A Bayesian approach
AU - Pan, Indranil
AU - Pandey, Daya Shankar
PY - 2016/2/1
Y1 - 2016/2/1
N2 - In recent years, different non-linear regression techniques using neural networks and genetic programming have been applied for data-driven modelling of fluidized bed gasification processes. However, none of these methods explicitly take into account the uncertainty of the measurements and predictions. In this paper, a Bayesian approach based on Gaussian processes is used to address this issue. This method is used to predict the syngas yield production and the lower heating value (LHV) for municipal solid waste (MSW) gasification in a fluidized bed gasifier. The model parameters are calculated using the maximum a-posteriori (MAP) estimate and compared with the Markov Chain Monte Carlo (MCMC) method. The simulations demonstrate that the Bayesian methodology is a powerful technique for handling the uncertainties in the model and making probabilistic predictions based on experimental data. The method is generic in nature and can be extended to other types of fuels as well.
AB - In recent years, different non-linear regression techniques using neural networks and genetic programming have been applied for data-driven modelling of fluidized bed gasification processes. However, none of these methods explicitly take into account the uncertainty of the measurements and predictions. In this paper, a Bayesian approach based on Gaussian processes is used to address this issue. This method is used to predict the syngas yield production and the lower heating value (LHV) for municipal solid waste (MSW) gasification in a fluidized bed gasifier. The model parameters are calculated using the maximum a-posteriori (MAP) estimate and compared with the Markov Chain Monte Carlo (MCMC) method. The simulations demonstrate that the Bayesian methodology is a powerful technique for handling the uncertainties in the model and making probabilistic predictions based on experimental data. The method is generic in nature and can be extended to other types of fuels as well.
KW - Bayesian statistics
KW - Fluidized bed gasifier
KW - Gasification
KW - Gaussian processes
KW - Municipal solid waste
UR - http://www.scopus.com/inward/record.url?scp=84946218743&partnerID=8YFLogxK
U2 - 10.1016/j.fuproc.2015.10.027
DO - 10.1016/j.fuproc.2015.10.027
M3 - Article
AN - SCOPUS:84946218743
VL - 142
SP - 305
EP - 314
JO - Fuel Processing Technology
JF - Fuel Processing Technology
SN - 0378-3820
ER -