Fast turnaround on individual gas and liquid simulations can be crucial for critical operational or planning decisions, however complex transient simulations for large systems can be slow. Surge analysis studies can be slow too, due to the necessarily small time steps and knot spacing for those use cases. Standard techniques for speeding up transient simulations using parallel computation generally require expensive low-level software rewrites. We present a parallel method that can be applied with little-to-no modification of existing software, and demonstrate both liquid and gas pipeline examples showing the new method to be computationally equivalent to their single-threaded counterparts. The method is applicable both to desktop and cloud machines, and scales to arbitrarily large pipelines.
Rapid turnaround of computational studies is essential for timely decision making. Such studies might involve completing many individual simulations in parallel - but often informed decisions hinge on getting an individual large transient simulation to run quickly.
“Large” in this context may mean geographically large. Some representative really large gas transmission examples include Petrochina’s West-East pipeline at 5224 miles (8707 km)1, GASUN at 2993 miles (4989 km)1, TransMed at 1485 miles (2475 km), and Rockies Express at 1007 miles (1679 km). Some liquid transmission pipeline examples include the ESPOOP at 2914 miles (4857 km)1, the Kazakhstan-China pipeline at 1679 miles (2798 km)1, and Keystone 1853 miles (3088 km, as completed). If one considers entire gathering and transmission systems as a whole, the scales increase even further. For example, TC Energy’s NGTL line has 14,745 miles (24,575 km) of pipeline.
Simulating systems of this scale inherently require a lot of computation - especially when compressor station modeling and transient thermal pipeline to soil calculations are required. This can lead to long run times, yet there is often a need to complete a particular individual simulation in a timely fashion in order for preemptive decisions to be made before it is too late.
At the other scale extreme, certain pipelines can be geographically small yet lead to computationally intensive simulations. Surge studies in liquid lines may involve relatively few km of pipe, yet may run slowly because of the extremely small knot spacing and time steps necessary to accurately model high speed surge relief valves and the sharp wave fronts and reflections. This is another category of problem that can benefit from running individual simulations faster.