Page 261 - Biofuels for a More Sustainable Future

Life-cycle costing: Analysis of biofuel production systems 235

1.2 Uncertainty in life-cycle costs
These costs are based on predicted and actual assumptions. Many LCC stud-
ies have assumed that all input parameters of the LCC model are determin-
istic (Tang et al., 2015). However, LCC calculation could involve many
uncertainties, even at the lowest level of fLCC. The methods to address
them have yet to be developed systematically (Ilg et al., 2017) The problem
is even more pronounced for prospective systems like proposed biofuel sys-
tems using novel feedstocks which strongly affect the operating cost (Myint
and El-Halwagi, 2009) or new process equipment that strongly affect the
investment cost (Brownbridge et al., 2014). For novel processes under
development, effects of technology maturity and scale-up are also difficult
to gauge (Chan et al., 2018). Data is usually difficult to obtain because of
corporate confidentiality concerns and they are also time sensitive.
A good example of a time-sensitive parameter that strongly influences the
LCC is interest rate. Since the randomness and the uncertainties behind
these parameters are uncontrollable, there is a need to systematically assess
the impacts of these parameters on the cost of the product.
There are many methods to assess the uncertainties in LCA or LCC
including interval analysis (Chevalier and Le T eno, 1996), probabilistic
(Kennedy et al., 1996), and fuzzy numbers methods (Tan, 2008). The uncer-
tainty of single parameters and their effects on the final results can be assessed
through sensitivity and matrix perturbation analysis (Heijungs, 2010). In
Heijungs (1996), the key issues analysis method was used to guide improve-
ments in data collection.
The simultaneous analysis of the effects of multiple parameters can be
done via a method that has been traditionally used guiding experiments
in the physical world: design of experiments (DOE). Long used for the anal-
ysis of complex systems like those found in agriculture and other living sys-
tems, DOE has also been used for designing computational experiments
(Giunta et al., 2003). That is, DOE is used to guide how parameter values
may be systematically varied such that the effects of individual parameters
may be assessed simultaneously, the overall uncertainty may also be gauged.
This is also known as global sensitivity analysis (GSA). Among the more
popular alternative or competing methods is Monte Carlo simulation which
has been used to determine lumped uncertainty in LCA (Ciroth et al., 2004).
The Monte Carlo method, however, requires a large amount of calculation
(Lloyd and Ries, 2008) although there have been algorithms proposed to
reduce computational times (Peters, 2007). Other types of DOE are