View Full Version : Model two adjacent reservoirs

Cliff Dredge
May 29, 2016, 07:04 PM
I'm trying to model two adjacent reservoirs which are identical size and elevation joined by a 100mm pipe about 2m long. My model seems to be doing funny things. I'm pretty sure it's the short length of connecting pipe, any ideas how to get around this?

I realise it would be easier to just model it as one double the volume reservoir, which I'll probably do but wouldn't mind knowing how to get around current scenario.

Cliff Dredge

Patrick Moore
May 31, 2016, 01:06 PM

Due to the mathematics used by EPANET to solve the head and flow equations, EPANET can often exhibit unusual behavior when two tanks are very close to one another in the model. Generally what happens is that the model has large flows between the tanks that flow back and forth in subsequent iterations simply due to trying to balance slight differences in the tank head by inducing large flows as this is the only way to get the head in each tank to match. In the next iteration the situation reverses and the high flow returns to the 1st tank as it now has a lower head than the other tank. In the next iteration it reverses and on and on the process goes.

The best and easiest way to resolve this is to simply model both tanks as an equivalent tank.

To do this use the following equations
A_new = A1 + A2
A = pi/4*D^2
So, A_new = pi/4*D_new^2 and A1 = pi/4*D1^2 and A2 = pi/4*D2^2
so A_new = A1 + A2 now can be written as pi/4*D_new^2 = pi/4*D1^2 +pi/4*D2^2
Divide all by pi/4 and rearrange you are left with D_new = sqrt( D1^2 +D2^2)

We have seen clients have better success with two tanks if they use very small hydraulic timesteps in the Time Setting Options, like in the 5 minute range, but given the mathematics of the equations it is often easiest to simply use equivalent diameter tank for both of the individual tanks. Please note if you decide to try the smaller hydraulic time step, make sure you reduce the Rule Timestep and quality timestep to 1/10th the revised hydraulic timestep, as this is a guideline recommended by EPANET.

Patrick Moore