(300f) Rolling Horizon Model for Gasoline Blend Planning Under Uncertainty in Demands Using an Updated Inventory Level Position Based on Chance Constraint Formulation

In this paper, a rolling horizon procedure is implemented to deal with gasoline blend planning problem under uncertainty in products demands. While most products from oil refineries are produced to meet contracted demand, there is an additional uncertain demands which refineries can satisfy to generate extra profit. Using a deterministic rolling horizon model leads to a suboptimal solution since such model fails to account for future production plan to meet additional uncertain demand. Inventory safety stocks approach can lead to an improved solution compared to the deterministic model, but it leaves open the question what is the right amount of safety stocks. In this paper, chance constrained programming is utilized to improve the inventory level position considering the uncertainty in future demands which is better than deterministic solution. A better prediction plan and products inventory levels can results in higher revenue by satisfying more demand and lower blending cost by using a more efficient blend recipes (lower cost per unit). Our stochastic model updates the production plan after every period based on the realized demand of the next period and new information obtained about future periods uncertain demands. The results shows that higher probability specification of the chance constraint for future planned periods will usually result in higher revenues but the blending cost per unit increase might have bigger impact on profit. Therefore different probability specification of the appropriate inventory level should be tested before making a final production plan.