(440e) Managing Flow Reversals in the Optimal Design and Operation of Pipeline Networks

Diego C. Cafaro, Demian J. Presser, Diego J. Trucco, Renzo O. Piccoli, Ignacio E. Grosmmann

Pipelines are among the most efficient and reliable means of transportation for liquid and gas products across extensive supply chains. However, pipeline networks comprise significant investment costs that are only justified when they are planned to operate at reasonably high utilization levels over long periods of time. That is why the optimal design, planning and operation of pipeline networks have received increasing attention from the research community for more than 40 years [1]. With production and demand patterns changing fast for most industries, usually under uncertain and unexpected scenarios, building efficient pipeline networks has become a certainly challenging task. One of the most interesting strategies that pipeline operators may apply after an abrupt change in the production-demand balance is flow reversal. Reversing the flow of a pipeline segment has the aim of using the same transportation infrastructure to make products flow in the opposite direction. Reversing pipelines can be particularly useful to achieve more economical network designs.

Interesting examples of flow reversals are shown in [2], where highly integrated water pipeline networks are built and operated to supply, gather, treat and recycle water streams for unconventional gas production. In these examples, nodes with high demand of water turn into production nodes after a few months, making it necessary to reverse the flow direction in order to gather and process the so called “flowback” water. Another interesting example is presented in [3], whose focus is on the optimal design of pipeline networks to gather unconventional oil and gas. If the development strategy is uncertain (driven by oil and gas prices), the optimal solution suggests splitting oil and gas facilities, each in its corresponding productive region. In a real context, company focus on oil and/or gas often changes over time, yielding mobilization of rigs and fracturing equipment from one region to the other. Nevertheless, centralized processing facilities and pipeline networks cannot be relocated. This implies that oil and gas flows may be subject to reversals if the same pipelines are used to send production streams to any of both regions, under alternating development strategies. Similar examples can be found in supply chain networks alternatively storing and withdrawing hydrogen [4], methane [5] or carbon dioxide [6] (usually underground), involving extensive pipeline networks with bidirectional segments.

Despite being a relevant problem, the active implementation of flow reversals has not been formally addressed in the optimal design and operation of pipeline networks. In fact, none of the previous works have accounted for actual capital and operational expenditures associated with flow reversals, never assessing to what extent their complexity can be justified. Converting a pipeline segment into a bidirectional transportation resource requires large capital investment in additional equipment, namely pumps, compressors and valves. Moreover, every time the pipeline flow direction changes from direct to reverse mode or vice-versa, time- and cost-consuming tasks need to be performed. Previous contributions on water pipeline networks [2] have assumed that, due to its relatively minor importance when compared to pipeline investment decisions, the operation of pumps and valves in either direct or reverse flow modes could be left aside from the scope of optimization models. However, for gas and multiphase flows, managing flow reversals is particularly relevant.

This work presents a generalized optimization framework for pipeline network design and operation assuming no predetermined number of echelons. A set of generic nodes need to be connected in order to gather streams and send products to processing facilities for demand fulfillment. In any of these nodes, facilities for merging, splitting, storing, separating and/or processing flows should be installed to make the product flows be ready for delivery or use. Flow direction may be reversed in any pipeline segment over the time horizon, but in contrast to previous contributions, changeover times and additional capital and operational expenditures due to specific pieces of equipment are explicitly accounted for. The optimization model is divided into two parts: topological decisions and fluid dynamics constraints. On the one hand, the optimal topology of the pipeline network is addressed by means of 0-1 variables accounting for the installation of a pipeline segment of certain diameter between a pair of nodes. Pipeline interconnections are usually selected from a superstructure of alternatives. Material flows (continuous variables) and balances (linear constraints) are also included in this part of the problem. However, the transportation capacity of a pipeline is not fully determined by its diameter and length. Sizing pipelines also implies the use of fluid dynamics equations to predict the pressure drops, which are highly nonlinear, both for liquid and gas streams, yielding complex relations between mass flowrates and pressures.

By including pressures as decisions variables, flow directions and rates can be optimally handled along the time horizon to make better use of the pipeline transportation capacity. Moreover, when a pipeline is reversed it is necessary to enforce the flow to be zero if the difference of pressures is negative (the fluid moves in the opposite direction), yielding disjunctive representations of the material flows at every period. As a result, the optimal design and operation of pipeline networks with flow reversals leads to a mixed-integer nonlinear programming (MINLP) formulation that is computationally challenging. The sources of complexity are basically two: the combinatorial nature of the network design problem, and the nonlinear relationship between time-varying flowrates and pressures. To solve the problem to global optimality we propose decomposition strategies and tightening algorithms that systematically add fluid dynamic constraints to a reduced set of segments and directions in the network. By means of these novel optimization techniques we are able to assess the impacts of flow reversals in pipeline networks of real size.

We illustrate the capabilities of the optimization approach by solving the design, planning and operation of a shale gas gathering network. We address a combined oil and gas production region with 72 wellpads organized in 8 rows. There are four potential capacities for processing facilities and five alternative pipeline diameters, from 10 to 24 inches. Every row is a potential location for a junction and/or processing facility. The time horizon is discretized into 60 bi-monthly periods, yielding an MIQCP (mixed-integer, quadratically constrained) model comprising more than 25,000 equations, 34,000 continuous variables and 23,000 0-1 variables, which is solved using GAMS-GUROBI 9.5. By allowing flow reversals we seek to enhance the utilization of existing infrastructure (pipelines and facilities) taking into account the steep declining slope of shale productivity curves. Given a well development plan, the goal is to optimally determine the number, size and location of gas processing facilities; pipelines diameters and interconnections; the need for flow reversals; and timing for the installation and expansion of pipelines, facilities and equipment. Results show that the net present cost can be reduced by more than 11 million USD (2.6 %) if flow reversals are allowed. Interestingly, only three of the pipeline segments in the network are reversed once along the time horizon. Savings are mainly realized from the use of smaller diameters in certain segments and postponing investments in longer pipelines. Ongoing work seeks to evaluate the importance of flow reversals when field development plans are uncertain.

References

1. Mah RSH, Shacham, M (1978). Pipeline Network Design and Synthesis. Adv in Chem Eng 10:125-209.
2. Cafaro DC, Grossmann IE (2020). Optimal design of water pipeline networks for the development of shale gas resources. AIChE J 67. https://doi.org/10.1002/aic.17058
3. Montagna AF, Cafaro DC, Grossmann IE, Ozen O, Shao Y, Zhang T, Guo Y, Wu X-H, Furman K (2022). Surface facility optimization for combined shale oil and gas development strategies. Optim Eng . https://doi.org/10.1007/s11081-022-09775-8
4. Zivar D, Kumar S, Foroozesh J (2021). Underground hydrogen storage: A comprehensive review, Int J of Hydrogen Energy 46:23436-23462.
5. Knudsen BR, Foss B (2017), Shale-gas wells as virtual storage for supporting intermittent renewables. Energy Policy 102:142-144.
6. Presser DJ, Cafaro VG, Cafaro DC (2022). Optimal Sourcing, Supply and Development of Carbon Dioxide Networks for Enhanced Oil Recovery in CCUS Systems. Comput-Aided Chem Eng. 49: 493-498.