Best way of Modelling Wells

[MSeoane]

Hi Everyone,

I have some doubts about what is the best way to model water wells in an epanet network.

I have seen them modeled in many ways that I will enumerate below:

Reservoir + pipe + FCV
Reservoir in dynamic level + pipe + pump
Junction with total head and negative demand (you need a "fake" reservoir at some point of the network in this case to make it work).

What do you think is the best way to model a waterwell?

Regards,
Matías Seoane
Civil Engineer

 

[fel3]

It depends. If you have a variable speed well pump, then a reservoir plus a FCV can work alright. This can even be used for constant speed pumps, but less accurately.

However, changes in groundwater levels over time and extremely high demands (e.g. MDD+FF) can make pumps operate below their normal output pressure and this can result in situations where the model doesn't match reality. I have a seven-well model like that right now that I inherited from a good engineer that I used to work with at another company. His model works fine for normal demand conditions, but to match field measurements during extreme conditions, he had to arbitrarily lower his fake reservoir levels. It is cumbersome, but in the absence of good data on the actual wells and well pumps, it was the best approach.

That being said, the absolute best method is to model each well and well pump as they really are. That means you need the pump performance curve, static groundwater level (=reservoir), groundwater drawdown curve, column loss curve, etc. The pump, drawdown, and column curves are then combined to create a new, composite pump performance curve. If your pump is variable speed, then you can work the similarity calculations to come up with several more parallel curves for lower speeds.



==========
"Is it the only lesson of history that mankind is unteachable?"
--Winston S. Churchill

 

[77JQX]

I have used methods 2 & 3, never method 1. If I have a pump curve and pumping water level in the well, I use method 2. I think this is the best method if the information is available, and if the model matches real world data.

If I only know the well flow, I generally use method 3. I just use a negative demand, not sure what you mean about "total head" and "fake reservoir". One needs to be careful with this method because the model can return stupidly high pressures with this method, particularly if there's no where for the water to go (maybe that's what you mean about fake reservoir?). To get around this limitation, sometimes I'll fake method 2 with an assumed pump and drawdown.

 

[hydrae]

I have always done this using method 1.
The reservoir is at static hgl in the well.
The pump is set to the correct elevation with a fully modeled pump curve.
The FCV models the well performance as determined by the draw down curve.
Input new FCV curves to model long term changes in the draw down or other influences on well performance.
yes, you do need a short pipe between the reservoir and the pump otherwise EPAnet can get upset.

I find the negative demand the third method would be cumbersome since the pump and well performances are variable.

Hydrae

 

[MSeoane]

Thanks to everyone for the answers.

77JQX: What I tried to say in case 3 was to put as the elevation of the node with negative demand the "real elevation" plus an extra elevation to represent pressure.
Fake Reservoir (I've read it also as silly reservoir) is when you put a reservoir with same head of the well node and connected by a pipe with a checkvalve to avoid reverse flow. I've seen this in some epanet examples made by some people.

This thread really helped me to identify the options available.