(373v) A Mixed Integer Non-Linear Programming Model for Optimizing the Operation of a Complex Gas-Oil Separation Network

Gathering and processing significant volumes of crude oil involves connecting to wells in different fields that are usually spread across large geographical areas. This crude can then be separated by Gas Oil Separation Plants (GOSPs) which are also located remotely from each other. These facilities are often connected laterally via swing and transfer lines which would allow shifting part or all of the production from one GOSP to another. The purpose of those pipelines is to provide an added flexibility to the operation of the overall network. For example, when a bottleneck exists in one GOSP relating to water processing capacity, the production from wells with higher water-cut can be diverted to a GOSP which is not similarly bottlenecked. This allows processing higher total crude volumes and leads to improving the utilization of these assets. Also, this often allows distributing crude oil more optimally to minimize energy consumption.

This work addresses optimizing the operation of a complex network of GOSPs in one of Saudi Arabia’s largest oilfields. These GOSPs are connected by pipelines which are constructed only between nearby plants where wells can flow naturally based on excess reservoir pressure and without the need to use any artificial surface boosting or subsurface lifting. The goal is to operate this network such that oil production targets are met at minimum energy consumption. A Mixed Integer Non-Linear Programming (MINLP) model is formulated to optimize swing line flows and processing equipment utilization. This allows the systematic identification of optimal operating points, as opposed to current common practices, which are largely based on heuristics.

This paper proposes a methodology to formulate and solve this problem. It describes the level of fidelity used to represent physical process units. This varies between use of detailed first-principles models for certain equipment to a more simplified representation elsewhere. This is done systematically based on the overall impact on the solution’s accuracy and robustness. This results in minimizing energy consumption while meeting oil production targets without added investment.