Academic Journals Database
Disseminating quality controlled scientific knowledge

Stochastic Programming with Tractable Meanb-risk Objectives for Refinery Planning under Uncertainty

ADD TO MY LIST
 
Author(s): C.S. Khor

Journal: Journal of Applied Sciences
ISSN 1812-5654

Volume: 10;
Issue: 21;
Start page: 2618;
Date: 2010;
VIEW PDF   PDF DOWNLOAD PDF   Download PDF Original page

Keywords: Mean-Absolute Deviation (MAD) | refinery planning | Two-Stage stochastic programming | Conditional Value-at-Risk (CVaR)

ABSTRACT
The application of Information Technology (IT) and Information Systems (IS) have been crucial in enhancing the operating flexibility and resiliency of refineries. In particular, the Process Systems Engineering (PSE) community has been instrumental in carrying out a key role in extending the systems engineering boundaries from mere chemical process systems to the incorporation of business process systems with consideration for risk. Thus, this study considers a robust framework for the economic and operational risk management of a refinery under uncertainty by extending an existing two-stage stochastic program with fixed recourse via scenario analysis. The problem is mathematically formulated as a two-stage stochastic nonlinear program with a tractable meanbrisk structure in the objective function. Two measures of risk are considered, namely the metrics of Mean-Absolute Deviation (MAD) and Conditional Value-at-Risk (CVaR). The scenario analysis approach is adopted to represent uncertainties in three types of stochastic parameters, namely prices of crude oil and commercial products, market demands and production yields. However, a large number of scenarios are required to capture the stochasticity of the problem. Therefore, to circumvent the problem of the resulting large-scale model, we implement a Monte Carlo simulation approach based on the Sample Average Approximation (SAA) technique to generate the scenarios. A statistical-based scenario reduction strategy is applied to determine the minimum number of scenarios required yet still able to compute the true optimal solution for a desired level of accuracy within the specified confidence intervals. The proposed model is illustrated through a representative numerical example, with computational results demonstrating how risk-averse-and risk-inclined solutions in the face of uncertainty can be attained in a risk-conscious model.

Tango Rapperswil
Tango Rapperswil

     Save time & money - Smart Internet Solutions