Petroleum Refinery Planning Under Uncertainty: A Multiobjective Optimization Approach with Economic and Operational Risk Management

In the current modernized globalization era, crude oil prices have reached a record high of USD 147 per barrel according to the NYMEX exchange on June 2008. It is forecast to spiral upwards (with the current graph trend) to a much higher price level. The current situation of fluctuating high petroleum crude oil prices is affecting the markets and industries worldwide by the uncertainty and volatility of the petroleum industry. As oil refining is the downstream of the petroleum industry, it is increasingly important for refineries to operate at an optimal level in the presence of volatility of crude oil prices. Downstream refineries must assess the potential impact that may affect its optimal profit margin by considering the costs of purchasing the raw material of crude oils and prices of saleable intermediates and products as well as production yields. With optimization, refinery will be able to operate at optimal condition.
In this work, we have attempted to solve model formulation concerning the petroleum refinery planning under uncertainty. We use stochastic programming optimization incorporating the weighted sum method as well as the epsilon constraint method to solve the model formulation of the petroleum refinery planning under uncertainty.
The objective of this research project is to formulate a deterministic model followed by a two stage stochastic programming model with recourse problem for a petroleum refinery planning. The two stage stochastic risk model is then reformulated using Mean Absolute Deviation as the risk measure. After formulating the stochastic model using Mean Absolute Deviation, the problem is then investigated using the Pareto front solution of efficient frontier of the resulting multiobjective optimization problem by using the Weighted Sum Method as well as the ε-constraint method in order to obtain the Pareto Optimal Curve which generates a wide selection of optimization solutions for our problem. The implementation of the multiobjective optimization problem is then automated to report the model solution by capturing the solution values using the GAMS looping system. Note that some of the major parameters used throughout the formulated stochastic programming model include prices of the raw material crude oil and saleable products, market demands for products, and production yields.
The main contribution on this work in the first part is to conduct a further study/research on the implementation of the model formulation in Khor et al. (2008) where the model formulated by Khor et al. (2008) uses variance as the risk measure. The results obtain in the previous paper will be compared with the method in this paper that incorporates Mean Absolute Deviation as the risk measure. To further study the model formulated, the solution obtain is further enhanced using the Weighted Sum Method as well as the Epsilon constraint method to obtain the Pareto Optimal Curve generation. Hence, most of the exposition on the model formulation and solution algorithms are taken directly from the original paper so as to provide the readers with the most accurate information possible.