Page 219 - Special Topic Session (STS) - Volume 1
P. 219
STS425 Zaitul Marlizawati Z. et al.
Two-stage stochastic programming approach for
oil refinery production planning
Zaitul Marlizawati Zainuddin, Arifah Bahar, Norshela Mohd Noh
Universiti Teknologi Malaysia
Abstract
Two-stage stochastic models are the common model in oil refinery stochastic
optimization problem. We propose this model as a framework to maximize
the expected profit of production planning for oil refinery industry. Geometric
Brownian motion (GBM) is used to describe the uncertainty for price and
demand of petroleum products. This model generates the future realization of
the price and demand in scenario tree that provides input to the stochastic
programming. The prices and demand data was obtained from Malaysia
Energy Information Hub (MEIH), Suruhanjaya Tenaga Statistics and was tested
for the oil refinery production planning. The result indicates that stochastic
model provided a better prediction of oil refinery profit margin.
Keywords
Two-stage stochastic programming; geometric Brownian motion; scenario
tree; oil refinery optimization
1. Introduction
Oil refinery is one of the most complex industries due to many different
processes with different chemical reactions involved to produce multi finished
products. In order to optimize oil refinery profitability, important decisions
such as to determine the right amount of crude oil to purchase, amount of
products to produce, and amount of raw materials and finished products
inventory with the best use of the existing resources are needed. This is more
crucial now especially recently in facing uncertainties such as fluctuations of
crude oil prices, unpredictable demand of finish products and unstable oil
production that cause difficulty to know the company’s direction in next year
ahead.
Nowadays stochastic programming has become one of the main modeling
technique in dealing with refinery planning optimization problem under
uncertainty because the deterministic model has a limited capability in
handling uncertainties. Stochastic programming can be divided into
programming with recourse and probabilistic programming as described in
Sahinidis [1] and Khor et al [2]. Two-stage stochastic programming with
recourse is the most popular model in the review study of oil refineries
planning under uncertainty [3]. The scenario construction for stochastic
208 | I S I W S C 2 0 1 9