Page 220 - Special Topic Session (STS) - Volume 1
P. 220
STS425 Zaitul Marlizawati Z. et al.
parameters and dimensional problem are the main challenges in solving two-
stage stochastic programming. Therefore we propose this model as a
framework to maximize the expected profit of production planning for oil
refinery industry. Leiras et al.[3] conducted a review of oil refineries planning
under uncertainties for journal articles from the 90s to 2010 and none of the
articles have emphasised on an approach to apply the behaviour of
uncertainties and the method of representation for stochastic parameters in
oil refinery optimization. Moreover, their study are limited to articles before
2011 thus, we focus on papers on oil refinery from 2011 onwards with the
representation of stochastic parameters. Ribas et al. [4] constructed the
stochastic parameter scenario tree based on the Brazilian refineries expertise
of employees to evaluate between stochastic and robust approach in
considering uncertainty in oil refinery activity in 2012 and stochastic model
gain highest profit expected value compared to robust approach .Geometric
Brownian motion is used to model the end products price uncertainties in
developing stochastic linear programming to maximizing profit of biofuel
supply chain in North Dakota as studied by Awudu and Zhang [5] in the year
2013. However, scenarios generated by the solution of GBM increased the size
of the problem and they applied Benders decomposition to solve large scale
mathematical programming. In 2015, Ruoran Chen et al. [6] proposed an
approach to apply the behaviour of crude oil prices follow GBM and employed
approximate dynamic programming to solve multiperiod multiproduct oil
refinery optimization problem at Shandong, China and they are also facing
difficulty in a high dimensional problem. Meanwhile, Relvas et al [7] built a
scenarios to describe the future realization of uncertainties, oil price and
demand with ARIMA and SARIMA model that provides the input to the two-
stage stochastic programming in maximizing the expected profit of Portugues
oil network. In this study, we improve the formulation of two-stage stochastic
programming by Khor et al [2] with constructing the scenario tree based on
GBM to generates all possible future realization of the price and demand
instead of only considering element in the set of event sequences is highest
and lowest value. The prices and demand data from 1990 to 2015 was
obtained from Malaysia Energy Information Hub (MEIH), Suruhanjaya Tenaga
Statistics and was tested for the oil refinery production planning to maximize
oil refinery profitability.
2. Methodology
2.1 Deterministic model
In the deterministic model, crude oil price, finish products sales price and
operating cost are constant with mean values from historical data are used
and each cost or sales prices are in dollar/barrel. The objective function is to
209 | I S I W S C 2 0 1 9