Design for dynamic performance: Application to an air separation unit Conferences

The significant effect that the design of a plant can have on its dynamic performance has led to methodologies for systematic analysis of the interaction between design and control, and for inclusion of dynamic performance considerations in plant design. In this thesis, an optimization-based framework is presented for improving the agility of a N₂ plant in response to the highly dynamic market, particularly demand and electricity price fluctuations. In this research, a decomposition optimization strategy is followed to identify limiting plant constraints and investigate selected design modifications using a rigorous dynamic air separation plant model. The plant model comprises of dynamic models for the distillation column with an integrated reboiler/condenser and a primary heat exchanger, and algebraic models for a compressor and turbine. The models presented follow first principles and/or empirical approaches. An index reduction procedure is utilized to reduce a high-index dynamic distillation model to an index-1 system. The proposed plant model is validated and reconciled using plant data through parameter estimation. Excellent prediction accuracy is obtained.

Steady state optimization is the first tier of the optimization problem. In this stage, different scenarios in terms of demand changes and electricity price fluctuations are explored. In all optimized cases, the liquid nitrogen (LN₂ ) production rate and rate of evaporation remain at their lower bounds due to the zero revenue for LN₂ product and the high cost of evaporation. It is demonstrated that operating close to the maximum allowable impurity level is optimal as it gives a high recovery rate. Two major findings in this stage are: (1) the operating window of the plant is defined by the flooding constraint of the distillation column and the surge constraint of the compressor when the feed flow rate is considered; and (2) there is a break-even point between revenue generated from and compression cost required by one mole of air feed into the system when electricity price fluctuates.

Dynamic optimization is performed to switch the system from the base case operating point to the new operating point determined from the steady state optimization without violating plant constraints . At rajectory tracking objective function with endpoint constraints to "pin-down" the final states to the pre-determined optimal values is solved in each case. With optimized control action, a fast transition without constraint violations can be achieved. Plants can complete the transit ion in less than 30 minutes with step-like responses of gas nitrogen (GN₂ ) production. Two plant modifications for aiding transitions are evaluated: (1) introducing external LN₂ during transit ions for cases of increasing demand, and (2) allowing a vent stream after the compressor for cases of decreasing demand. Even though for this particular plant setup (impurity requirements, tray design, etc.), introducing external LN₂ may not be cost-effective, this design modification is very attractive as it allows a smaller operating safety margin in the impurity requirement, which could result in more profitable steady state operation.