Dynamic Modelling of Flare Networks

This is a repeat of an older post with some added information

There is a lot of interest in modelling flare networks dynamically these days. Although the HYSYS Dynamics pipe segment doesn't model all phenomena rigorously, you can do a pretty good job with it. To be clear where the limits are, first some things you cannot do:

• The pipe segment is lacking the momentum balance. So, you should not use it to evaluate what happens in the really short term after opening a relief valve. HYSYS Dynamics cannot model shockwaves that traverse the pipe when the relief valve pops open
• The HYSYS pipe segment does't account for kinetic energy. The lack of kinetic energy term will affect the temperature results, but the deviations shouldn't be too larger. I have no direct comparison of dynamic cases, but steady state cases indicate it should be within a couple of degrees C.
• The pipe segment cannot model any phase slip effects

But typically, the interest in a dynamic model of a flare system is around the build up of back pressure and how you can use staggered opening of blowdown valves to use the capacity of the flare system as efficiently as possible. Velocities are usually high and liquid amounts are usually not high. So in many cases the liquid is fully dispersed in the gas. The phenomena we are interested in are short term, but not so short term that shock waves have a significant impact.

There are a couple ground rules you need to respect when building your model.

Use the HYSYS pipe segment and remember the following:

1. HYSYS Steady State has no network solver (except Aspen Hydraulics), so it may be just as easy to start  in dynamics right away
2. Start with a case where you actually do have flow from all sources. It doesn't need to be the full flow, but some flow for sure
3. As the dynamic solver is more prone to failing once in a while, make sure you frequently save things
4. Build up the network a couple of pipes at a time
5. The "model holdup" option was inactive per default in pre-V8.x versions, so make sure you check this for all pipes in older models
6. Use the "Full Range Churchill" method
7. Use "one DP calc/segment"
8. Lump your fittings as much as possible
9. Keep you segments long enough
10. If you want to model the heat loss, best do so from the start. Make sure temperatures are behaving before you move on to model your next couple of pipes.

Items 8 and 9 merit some more explanation:
Do NOT try to put all the detail you have into the model. lump all the bends and tees and whatever else into a single fitting. A swage is a particularly nasty one to have. The problem being that a swage that enlarges the pipe diameter increases the pressure, so HYSYS Dynamics will translate that into a negative k-value and when your flow goes away from the flow where it was calculated, all sorts of odd things can happen.

You need to make sure that the residence time of the gas in a segment is less than the time step of the pressure flow solver. For example, if the maximum velocity in a pipe segment is 100 m/s and that segment is 10 m long, the residence time is 0.1 s. So, the time step needs to be less than 0.1 s in this case. You can see that if you start putting a bit of pipe of only 0.1 m long in your model, you should really reduce your time step to 1 ms. And that will slow down your model tremendously. If you do not respect this limit on the time step, you risk seeing unrealistic instabilities in your model.

If you want an accurate model of the flare system, you'll also need an accurate model of the relief valve. You should use BPT-PSX. This is a commercial extension for HYSYS (and Petrosim) that uses the HEM method and properly accounts for and updates all the derating factors. So, you don't need to discard all cases where the back pressure rises above the maximum allowable back pressure, the valve will account for the reduction in flow due to excessive back pressure.