epanet fcv valves

[maxeng84]

Hello
I'm modelling a water distribution network with Epanet2 and I've a problem with an FCV valve. I've choosen the setting of the valve (8 l/s), and the status (none). The simulation lasts 24 hours. Everytime the valve is active and the flow is limited in 8 l/s, the program computes an headloss of about 7 meters on the valve but this value overestimates the real headloss I have in it (about 3 meters when the valve is closing and flow is limited in 8 l/s). Because of this, the computed pressure of the FCV controlled district is underestimated of about 4 meters and I'can't find a way to change the headloss and compute the real headloss. Can anyone help me to fix this problem,please?
Thanks
Max

 

[77JQX]

Is see two potential issues: First, the minor loss coefficient for the valve may be too large. Check the manufacturer's literature for the model and size of the valve being used. The second issue is that the model may be right for the loss across the FCV - there may another error in the model that keeps the downstream pressure in the model too low, and the loss across the valve is what it hydraulically solves to at the 8 l/s flow (due to the low pressure on the downstream side of the valve).

 

[hydrae]

A control valve set to control flow will do just that, and if the valve is achieving the flow desired, then the minor loss coeffienct should not be a factor since that would only engage if the valve was unable to achieve its set point with the pressure differential available.
you have a set point of 8 l/s and only if the computed flow is 7.9 l/s or less then the minor loss engages along with the other head losses and then the model ignores the flow control function


Where is the error? upstream or downstream, from your question I will assume downstream
What is the source of pressure control downstream of the valve? both in your model and in the system
tanks and or reservoirs?
other control valves with a target pressure
or demand and emitters

if demand, then that would make pressure vary widely according to changes in demand
if emitters this will have a self dampening effect but would still vary

Otherwise I would then focus on the source of pressure control to correct to the expected pressure

Hydrae

 

[maxeng84]

Thank you guys, you both helped me to understand the problem with my model.
I've found a solution that might fix it, but maybe you could tell anyway what you think about it. Infact, the distribution network i'm studying (i call it N1) is not independent, but is connected to another network (called N2, not modelled) which serves N1; so the pressure in the modelled network N1 is controlled by three reservoirs which represent the boundary conditions (the other network N2 should be modelled too but in this moment i can't do that). The boundary conditions are setted starting from some head measures in the points of connection between the two networks (N1 and N2). We wanted to connect the network N1 to another one (N3, modelled too), and that's the meaning of the FCV valve (to prevent the tank of N3 to get empty). As i connect N1 with N3, the distribution of the flow completly changes in N1, and at the same time of course, also the boundary conditions with N2 should change. The problem is that i can't foresee how the boundary conditions in the connection between N1 and N2 can change because of the new connection between N1 and N3.So the first idea i had, and it's probably too wrong, was to keep the same head boundary conditions in the connection between N1 and N2. So, to balance the network, the program computes an headloss in the FCV valve in the connection between N1 and N3 when the flow exceeds 8 l/s, according to the head setted in the boundary conditions with N2, and overexstimates it (about 6 meters instead of about 3 meters). So, the solution i've found is to make some iterative little changes to the head boundary conditions with N2, till the distribution of the flow can balance the network with an headloss of about 3 meters on the FCV valve when the flow exceeds 8 l/s. Obviously the problem is not simple, because there are too many variables in the boundary conditions and it's not easy to foresee how they can change. But i think that this can be a good solution to fix my problem: infact, now the pressure is not underestimated anymore in the FCV valve controlled district, and there are no problems in the other part of the network N1 connected with it.
What do you think about it?? Any suggest?
Hope I was clear.
Thank you again!

 

[hydrae]

I would make simple models of the unmodeled networks with their major supplies, demands and pipes all grouped together
say N2 has a head controlled source which is 2 km away from the FCV through a network 4 800mm pipes, the pipes would be grouped as a single 1600mm pipe, and split that pipe into 2 parts with the local demand centered between the FCV and source.

how are N1 and N2 connected?
also where you are measuring head near the FCV, are the velocities high, this could input error into your measurement due to velocity depression

so the FCV is connected between N1 and N3, right?

 

[77JQX]

How do you come up with the required 3m pressure differential and 8 l/s flow between N1 and N3? Are N1, N2, and N3 existing distribution systems (not just models) that you made measurements in?

I think you're on the right track with looking at the boundary conditions, although my first suspicion was that it was a boundary condition in N3 that resulted in the low pressure.

I'd also add that in our water system, it can be hard to resolve differences of less than 5 psi (3.5m) between the model and the real system. Greater accuracy can require greater effort both in the measurement (how accurate are your gauges and flow meters?) and in setting up the model (for minor losses, individual pipe C factors, spatial and temporal distribution of demands, etc.). Does the 3m discrepency affect the conclusions of your study? If not, that may be the limit of accuracy without expending substantial additional modeling effort.

If you're happy that N3 is accurate, then I agree with hydrae that the first place to put in extra effort is in modeling your supply (boundary conditions with N2). I tend to shy away from skeletonizing (representin a group of pipes as a single pipe) for three reasons: (1) drawing the actual pipes is often quicker than computing the equivanent single pipe, (2) modeling the actual pipes allows me to compare model results with field results, and (3) drawing the actual pipes gets me closer to having finished model (for N2 in this case).

 

[maxeng84]

Thank you again

I'll try to explain in details how N1 works.
Actually the network N1 is just connected with N2. the connections are 3, in 3 different points with different measured values of head. 2 of them are completly free, even without a CV valve, and serve the biggest part of N1: so head in N1 completly depends on N2 demands and head; the other one conncetion between N1 and N2 is controlled by a CV valve, but in facts it's always closed because of bigger head in N1 in the point of connection. N1 is also connected to another network ( called N4 and not modeled) with a partialized valve which limits the flow in about 5 l/s (in order not to create pressure deficiencies in N4) and this connection serves a little part of N1.
With this actual conncetions, network N1 has some head deficiencies both in winter and in summer at the rush hours, because the N1 pipes have small diameters and many headlosses. So we thought we could connect N1 with the near network N3 (this network is completly modeled and calibrated), which can serve a part of N1 (with max 8,5 l/s). This new connection actually does not exists, i'm designing it with a new pipe. The pressure this new pipe can guarantee is about 5 meters bigger than the actual pressure in N1 at the point where the new connection should be realized. This estimate comes from the difference between the calculated value of pressure in the point of connection between N1 and N3 taken from N3 model simulating a 8,5 l/s new demand for N1, and the real measured pressure in N1 at the point of this new connection.
Then i've made a new model which includes both N1 and N3, with the new pipe which should connect the two networks.
In order to prevent N3 tank to get empty, we decided to limit the flow between N3 and N1 with the FCV valve I was talking about. To limit the headloss in the FCV valve the builder suggested me a model of valve with just 3 meters headloss when the flow is limited in 8,5 l/s. As i keep the same actual head boundary conditions for N1, the N1-N3 model gives me a 6 meters headloss in the FCV valve and gives me not satisfying pressure results and underestimates the pressure in FCV controlled part of N1 (less than 3 meters accurancy is important for me to justify the new pipe in project). For this reason i thought i could make some iterative little changes (max 2-3 meters) in the other head boundary condition (with N2 and N4) in order to have new balances of N1, till the headloss in the FCV centers about 3 meters (the real headloss). Of course i can't say that new boundary conditions i've found are the perfect ones, but the system works, and the results now are satisfying.
As you can understand, it's hard to optimize this problem, and obviously the best solution can't prescind the model of N2 and N4 too.
But i can't do that: N1 and N3 belong to a municipality, N2 and N4 belong to a different one...the problem has too many variables and i think that i can't model 4 networks (even with a simplified model) just to design one new pipe...i add that N2 is much bigger than the other netwoks (50,000 people for N2 vs about 10,000 for each other one)...it's too expensive for me!!!
For this reason i think the solution i've found can be enough for me.
The problem is not simple, what do you think?
Thank you for your attention!
Max

 

[hydrae]

Ahh...

I believe you have properly value engineered your time to answer the question to your satisfaction.
I also think the multiple networks as you describe them are vulnerable to high pressure during low flow, and low pressure under high flow, since it sounds like they are using restrictions within the networks to control the pressure

Hydrae