A CHANCE-CONSTRAINED APPROACH FOR OPTIMIZATION OF GAS PROCESSING PLANT OPERATION UNDER UNCERTAIN CONDITIONS

Natural gas plant operations contribute hugely to the economies of many developed nations that depend on hydrocarbon resources. The plant operation is usually subjected to continuous variations in upstream conditions, such as flow rate, composition, temperature and pressure, which propagate through the plant and affect its stable operations. As a result, decision making for optimal operating conditions of an in-operation plant is a complex problem and it is exacerbated with the changing product specifications and variations in energy supplies. This work presents a new solution method to the problem, which is based on chance constrained optimization method. A deterministic model is initially developed from process simulation using Aspen HYSYS and later converted to a chance constrained model. The probabilistic model is then relaxed to its equivalent deterministic form and solved for optimum solution using GAMS. The optimum solution is determined probabilistically using chance constraints that are held at a user-defined confidence level. Optimal solution is represented graphically as a trade-off between reliability of holding the process constraints and profitability of the plant. Three case studies are presented to demonstrate the new method. Optimization results show that uncertainty of plant parameters significantly affect the economic performance of the plant operation. The solution approach developed in this work is able to increase the reliability of maintaining the profit by more than 95% confidence level. As a result, the risk of constraints violation is reduced from more than 50% using the typical deterministic optimization to less than 5% with the chance constrained optimization approach. In addition, the results from this study indicate that the variation of material flow from the plant inlet has greater impact by more than 85.5% on profit compared to variation from the plant outlet, which is less than 2%. The variations of energy flow affect on profit is mainly changes with confidence level measurement higher than 95%, although material flow uncertainty is more sensitive to profit changes than uncertainty in energy flow. Final computational results also highlight the advantage of the developed chance constrained approach, which combines both the profit and the
vi
reliability of the process constraints, over “worst case” and two-stage programming approaches. Decisions from the “worst case” approach may reach to more than 99% confidence level which can drastically decrease the profit while the optimal decision from the two-stage programming does not clearly show to how much extent that the profit has been held. The developed solution approach in this work can aid as guidelines to flexible plant operation decision making for the in-operating plant by satisfying all the process constraints at certain confidence level.