(465b) Short Term Planning of Upstream Natural Gas Supply Chain Operations | AIChE

(465b) Short Term Planning of Upstream Natural Gas Supply Chain Operations

Authors 

Selot, A. - Presenter, Massachusetts Institute of Technology
Barton, P. I. - Presenter, Massachusetts Institute of Technology


Making routine operational decisions in natural gas production systems tends to be difficult due to highly volatile demand and uncertainty in the availability of the infrastructure required for gas transportation and storage. Extensive contractual frameworks have evolved between producers and consumers to mitigate the risks associated with these factors. This further complicates the decision making process.

Operational planning of the upstream supply chain in natural gas production systems can therefore aid in making production decisions efficiently and informatively to achieve various economic and operational objectives. This can help to ensure reliable supplies to fulfill highly volatile demands, minimize costs through efficient operation and meet customer specifications consistently. The upstream supply chain consists of the supply chain from reservoirs to large volume consumers.

We present a technical and economic model for optimizing the operations of the upstream natural gas supply chain using the Sarawak Gas Production System (SGPS), located in East Malaysia, as the case study. The SGPS comprises a network of offshore fields in the South China Sea that supplies a liquefied natural gas (LNG) plant complex consisting of three LNG plants in Bintulu, Malaysia. The system produces around 4 billion standard cubic feet per day of dry gas. It has an annual revenue of around $4 billion that forms approximately 4% of Malaysia's GDP.

The complexity of the system presents a two fold challenge: developing a reasonable mathematical representation of the system and also solving the resulting mathematical programming problem. The overall model can be viewed as a contractual framework model superimposed on top of a model of the infrastructure.

The infrastructure model comprises of models of well flows, the pipeline network and facilities. Pressure flowrate relationships are quite important in natural gas systems and the simplest relationships for flows in wells and pipelines are nonlinear. Additionally, the network transports multiple qualities of gas at any time and it may be interesting to determine a preferential routing in the network to achieve the desired blend at the LNG plants. This results in bilinear constraints in the model. Hence the overall infrastructure model is nonconvex.

Operation of the SGPS is based on operational and contractual rules that have evolved along with the system over its lifetime. The contractual and operational rules dictate the customer requirements, including demand volumes as well as gas quality, that is defined by the heating value and species thresholds. Also, these rules govern the transfers between contracts when a particular contract cannot fulfil its demand. Several entities own the facilities of the SGPS. Hence this framework also lays out rules for the use of shared facilities. Any planning tool has to take into account these complex rules for the results to be acceptable for implementation on the network. An accurate mathematical representation of these rules is quite a challenge and necessitates the use of binary variables and constraints.

Important decision variables in the model include the production share from each well or field, the corresponding flowrate and pressure distribution in the network and the status of the inter contractual transfers. The objectives may include maximizing delivery of dry gas, maximizing/minimizing an economic metric or maximizing production of by-products (Liquefied Petroleum Gas (LPG) or Natural Gas Liquids (NGL)).

The final model is therefore a relatively large-scale nonconvex MINLP with several hundred continuous variables. It is necessary to use global algorithms to solve the problem reliably and robustly. We use a global branch-and-reduce algorithm ([1]) to solve the problem. It may be necessary to customize the algorithmic heuristics using physical insights from the problem structure to be able to solve larger and more refined models of the system.

References:

[1] Mohit Tawarmalani and Nikolaos V. Sahinidis. Global optimization of mixed-integer nonlinear programs: A theoretical and computational study. Mathematical Programming, 99(3):563?591, April 2004.