Pipeline optimization software focuses on finding a single optimal operating scenario for a given pipeline configuration. How the system behaves over ranges of particular parameters, however, is typically left unstudied. In this paper, I utilize an explorer tool to extract various subsets of the solution space for a liquid pipeline model currently optimized through commercial pipeline optimization software. Analysis of these subsets reveals unexpected system responses to changes in control variables, the knowledge of which can be exploited in operational planning. Additionally, knowledge gained through the parametric studies may be utilized to calibrate the optimization to achieve further reduction in operational costs.
The goal of most liquid pipeline optimization software is to find the pump operations that will minimize the operational cost of the pipeline system, given a large set of data that describes the setup and constraints of the system. Commonly, the pump operations along with possibly a few other variables make up the solution set that is sought after, and all other parameters of the configuration are held constant through the simulation. If one wants to vary a parameter outside of the solution set, this has to be done manually by simulating each value individually. In addition, the solution given by said software is a single minimal solution produced by the setup and constraints. This has the clear disadvantage in its lack of knowledge of how the objective function behaves near the minimum or in other regions of the solution space. Furthermore, the solution space produced by the pump combinations is many-dimensional. The addition of other variables increases the dimensionality even further, causing the solution space to become even more unwieldy and impossible to view graphically.
