Complex Pipeline Network Simulation

NextGen is a natural gas pipeline simulation application developed by Gregg Engineering that lets users create hydraulic simulation models of their complex pipeline networks and perform both steady state and transient simulations.

Simulation models are used to:

  • Design facilities
  • Estimate capacities
  • Determine how to operate the pipeline

The simulation models can be operated in the same manner as you would operate the real pipeline by putting flow into the system at various supply locations, pulling flow out at delivery locations, opening and closing valves, turning compressor units and off, controlling pressures with regulators and so on. Simulation runs are used to predict pressures, flows and other parameters and are typically used from a design perspective to design pipelines and pipeline facilities, and from an operational perspective to estimate capacities and determine the best ways of operating the pipeline.Obviously, in order for pipeline simulation models to be effective and useful, they must accurately portray what the real pipelines do in real life.

Simulation models must be accurate to be useful

There can be several issues that can cause a pipeline simulation model to deviate from the real world pipeline.For example the internal roughness of the pipe may not be properly set in the model, or there may be elevation changes, obstructions, bends or other features in the pipeline that are simply not reflected in the computer model. Pipeline diameters and other user entered data in the model may differ slightly from what is in the real pipeline.

Model inaccuracies can be caused by:

  • Incorrect model configuration
  • Bad telemetry
  • Liquid hold up

There may also be inaccurate telemetry of operating data such as pressures and flow rates which can make its way into the simulation models and cause model inaccuracies. One of the biggest culprits in gathering systems is the buildup of liquids in dips or low elevation areas of the pipeline which reduces the cross sectional area of the pipe that the gas travels through. This in turn leads to pressure drops that are substantially higher in the real pipeline than those predicted in the simulation model. These inaccuracies are highly volatile and can change from day to day as liquids build up and slugs of liquid then get pushed out by pressure build up or by running scrapers through the lines. Simulation Models that are inaccurate, are just not very useful and in some cases can be quite costly. For example if an inaccurate simulation model is used to calculate capacities, the model can grossly overestimate actual pipeline capacities which can lead to unplanned curtailments, or putting wells on line that the system is not capable of handling. On the other hand, underestimating capacities leaves unused capacity in the pipeline and reduces profitability. Inaccurate simulation models can also lead to improper operational planning in terms of compression needs and line pack management, which can result in unused capacity in some cases, and unexpected curtailment in others. And when designing additions or expansions to existing pipelines, inaccurate models can even lead to installing wrongly sized pipes and either building unnecessary facilities, or not building enough facilities to handle planned capacity.

What all of this points to, is that it is absolutely essential that pipeline simulation models accurately reflect real pipeline conditions. In order to accomplish this, simulation models must be properly tuned and calibrated as often as possible.

Model accuracy is achieved by calibrating pipeline efficiencies

This is typically done by modifying the pipeline efficiencies in the various pipe segments throughout the model, so that predicted pressures and flow rates more closely match up with those found in the real pipeline. This is a fairly straightforward process for straight line single pipe shotgun systems, and an engineer can usually make a series of simulation runs and by trial and error globally modify pipeline efficiencies in various areas of the pipeline until predicted pressures and flow rates match up with actual data.

This process becomes much more difficult for more complex transmissions systems with multiple line services, or highly branched gathering systems where telemetry can be sporadic. And it can become almost impossible for very large highly interconnected distribution systems.This calibration process can also be very time consuming and take up to several days to complete and in some cases, calibrating the model is so difficult that it is simply not done at all or only a half-hearted attempt is made to properly calibrate the model.

The solution: Gregg Engineering’s AutoTune Module

Over a period of several months, Gregg Engineering staff worked to develop a series of optimization algorithms to help users calibrate their models and this was then packaged into NextGen’s AutoTune Module. The development of AutoTune was a very ambitious undertaking, and the development team faced a number of challenges. The first optimization algorithm worked well with straight line transmission systems, but did poorly with branch type gathering and highly networked distribution systems. A second optimization algorithm was developed to deal with branched gathering systems, and a third optimization algorithm was customized for highly networked distribution systems. Each of these techniques worked moderately well with the specific type of system they were designed for, but it would take several hundred iterations and in some cases several thousand iterations to calibrate all of the pipeline efficiencies on a pipe by pipe basis and to significantly reduce the overall error between predicted pressures and flow and actuals by about 80% for actual pipeline systems this was tested on.

The problem is that there is no such thing as a theoretically perfect branched gathering system, or straight line shotgun system, or networked distribution system, and that in fact all pipeline systems actually contain some degree of branching, some portions of straight line pipes, and some degree of interconnected networks.So rather than trying to pick which optimization algorithm would work best for the particular system being calibrated, we decided to try a combination of all three optimization algorithms operating in sequence regardless of the type of system.

Three layers of highly sophisticated optimization algorithms are used to recalculate individual pipeline efficiencies on each and every pipe segment in the model so as to best match up the model with real world conditions.

The results were not just good, but phenomenally good. So much so that at first we questioned whether the results were even valid, thinking that perhaps we had introduced some weird bug into the software. Whereas before it would take several hundred to several thousand iterations to get an 80% reduction in overall error, with the new combined methodology, we were getting an 80% reduction in overall error on the very first iteration!

By the time we completed 10 iterations, we were getting 95% to 99% reductions in overall system error, and in most cases the pipeline system is fully tuned and calibrated after 20 iterations.The methodology worked just as well for highly branched gathering systems as it did for highly interconnected distribution networks.

95% to 99% reduction in system error within 10 iterations!

By looking for lines with very low efficiencies, we can also identify areas of the pipeline that most likely have liquid hold up, and would be good candidates for running a scraper to push out liquid hold up.

Lines with liquid hold up can be easily identified, and scraper runs scheduled properly

Note that AutoTune should always be re-launched after any scraper runs in order to recalibrate the model since pipeline efficiencies in those lines can improve significantly after a scraper has cleaned them out.With AutoTune, pipeline efficiency calibration studies that might not have been possible otherwise, or might have taken an experienced engineer several days to perform, can now be performed in minutes.

With AutoTune, models can be recalibrated weekly or even daily

AutoTune is very easy to use, and all you have to do is import actual field data (pressures, supplies, deliveries) from an Excel spreadsheet into NextGen, and press the Run AutoTune button – the program takes it from there and does the rest.Because of the fast turnaround, companies that may not have recalibrated their simulation models at all, or only infrequently, can now easily do so weekly, or even daily.