(662f) Robust Scheduling of Crude Oil Operations under Demand and Ship Arrival Uncertainty

Scheduling of crude oil operations is an important and complex routine task in a refinery. It involves crude oil unloading, tank allocation, storage and blending of crudes, and CDU charging. Optimal crude oil scheduling can increase profits by exploiting cheaper but poor quality crudes, minimizing crude changeovers, avoiding ship demurrage, and managing crude inventory optimally. In our previous work, we have developed robust algorithms for obtaining optimal schedules for operations without any uncertainty, but in a practice, uncertainties are unavoidable. Some common and frequent uncertainties in refinery operations include ship arrival delays, demand fluctuations, equipment malfunction, etc. In the face of these uncertainties, an optimal schedule obtained using nominal parameter values may often be suboptimal or even become infeasible. Thus, it is critical to develop algorithms that can consider future uncertainty at the scheduling stage to improve schedule feasibility and robustness.

So far, scheduling and planning under uncertainty has been studied in specific fields such as capacity expansion, production planning, batch plant scheduling, etc. Demand and processing time uncertainties have been the focus of most existing work in the batch area. However, little work exists on refinery planning and scheduling under uncertainty. Arief et al. (2004) proposed a heuristic approach to reschedule operations of a given schedule to accommodate disruptions. Neiro and Pinto (2005) proposed a production-planning model incorporating product price and demand uncertainties in a refinery. Li et al. (2004) addressed the problem of refinery planning under demand or other economic parameters uncertainties with two-stage stochastic programming approach. Li et al. (2005) developed a planning model for refinery under correlated and truncated price and demand uncertainties. However, no work has so far addressed the development of robust schedules for crude oil scheduling in the face of uncertainties.

In this paper, we modify the deterministic MILP approach of Reddy et al. (2004) to address two important uncertainties in crude operations, namely product demand and ship arrival uncertainties. For crude oil scheduling, we define schedule robustness in terms of the schedule effectiveness, predictability and stability (Gan and Wirth, 2004). For demand uncertainty, we propose a scenario-based multi-period programming model in which all decision variables are treated as first-stage variables in a two-stage stochastic programming model. We then develop a reactvie scheduling strategy for different scenarios to evaluate the robustness of the schedule obtained by multiperiod programming model. It is found that this schedule is more robust than the original schedule. Moreover, this schedule is feasible over the entire expected range of uncertainty, whereas the original schedule is infeasible over some part of the uncertainty range. For VLCC arrival uncertainty, we treat all decision variables as second-stage variables in a two-stage stochastic programming model. We also develop a scenario-based multiperiod programming model in which we maximize the expected total profit and minimize connection changes (i.e. parcel to SBM or jetty connections and storage tank to CDU connections) among different scenarios. However, the large size of the model, increasing exponentially with the number of scenarios, makes it very difficult to solve. To overcome this difficulty, we propose an approximate decomposition strategy, in which we first solve each scenario independently except the nominal scenario and then solve the nominal scenario by maximizing the expected profit and minimizing connection changes from the individual scenarios. This decomposition strategy greatly reduces the model size and makes it possible to incorporate many scenarios and solve examples with up to two weeks of scheduling horizons. We also use the reactive scheduling strategy for different scenarios to evaluate the robustness of the schedule obtained with decomposition strategy. Our evaluation shows that the schedule obtained from our decomposition algorithms improves robustness and yields greater profits than the original schedule.

Key words: crude oil, scheduling, uncertainty, robustness, scenario-based approach, decomposition strategy, reactive scheduling

References

Adhitya, A., Srinivasan, R., Karimi, I. A., A Heuristic Reactive Scheduling Strategy for Recovering from Refinery Supply Chain Disruptions, AIChE Annual Meeting, Austin, TX, Nov 7-12, 2004.

Gan, H, S., Wirth, A., Generating Robust Schedules on Identical Parallel Machines: Heuristic Approaches, ICOTA 6, Ballarat, Austra, Dec 9-11, 2004

Li, W. K., Hui, C. W., Li, P., Li, A. X., Refinery Planning under Uncertainty, Industrial and Engineering Chemistry Research, 2004, 43, 6742-6755.

Li, W. K., Karimi, I. A., Srinivasan, R., Planning under correlated and truncated price and demand uncertainties, AIChE Annual Meeting, Cincinnati, OH, Oct 30- Nov 4, 2005

Neiro, S., Pinto, J. M., Multiperiod Optimization for Production Planning of Petroleum Refineries, Chemical Engineering Communications, 2005, 192 (1): 62-88.