Optimizing Batched Operations over Time

Batched operations on a liquid pipeline pose distinct challenges to optimization analysis. Considerations of the position and movement of individual batches often dictate which operations are allowed, and the management of batch-specific controls can complicate the underlying simulation strategies employed by the analysis. In addition, time-of-day power rates and power cost ratchets require a comprehensive view of the operation rather than a simple series of individually optimized operating states. As one moves away from simple "snap-shot" optimization (optimizing the operation for a single moment in time) to the more complex "longterm" optimization (allowing short-term inefficiencies in order to garner better overall long-term performance), the analytical strategy must change.

Direct Simulation and Iterative Steady States

Batched operations are primarily addressed by two methodologies: direct transient simulation of the operation or an approximated simulation using iterative steady states. Under both of these methods the supporting simulator executes a chronological series of simulated pipeline states. The result of each step in time capitalizes on the previous step and provides information to the next. The connected series of simulation steps collectively captures the subject batching operation. This chronological stepping through time means that each step only sees a "snapshot" of the pipeline - not the overall operation. Techniques may be used to mitigate this essential isolation of the several individual steps, and optimization is usually applied to the overarching operations plan. Nevertheless, evaluation of the plan (or changes to the plan) require re-computation of some (or all) of the individual simulation steps. Since almost all of the optimization techniques require iterative processes in their own right, the calculation of each optimization iteration is compounded by a number of supporting state iterations.

Challenges of Optimizing Events Displaced in Time

Some operational objectives require anticipated actions to realize a later improvement. For example, the injection of DRA (Drag Reducing Agents) into a line may alleviate the need for a particular pumping station, allowing it to be shut-down and avoiding a potentially expensive power usage spike. However, for the DRA to be in the appropriate position in the pipeline at the appropriate time, it will need to be injected well before the actual power demand becomes hydraulically evident. Balancing and navigating such optimization opportunities can prove to be tricky. The overall optimization process can become complicated and consume considerable computing resources and time.

A Small Innovation

Perhaps it would be possible to restructure the formulation of the basic strategy to alleviate some of these issues? Suppose that the analytical model was expressed in such a manner as to permit the evaluation of all of the various time steps in a single computational step. Events separated in time but now no longer separated computationally could be mathematically expressed so that their simulated states naturally interrelate. Anticipatory operations would thus fall naturally from the optimization step computation instead of requiring tedious extraction from a complicated (and possibly problematic) set of sub-iterations.

This content is only available via PDF.
You can access this article if you purchase or spend a download.