Modelling problems in EPANET

[Tait]

Hi all,

I am very new to EPANET, and have recently been tasked with modelling an existing water supply network of a Municipal District. The town currently operates through one water treatment plant, where water is pumped from a well throughout the town in a pressurised system. 6 pumps are currently operational and are each designed to begin pumping when system demand reaches a particular level.

I have designed the network and applied controls for the pumping arrangements, but my analysis has consistently been returning errors despite many adjustments. In particular, my pumps are not turning on and performing when I would like. The results in my network tables seem to show fairly appropriate values, but it seems that at the start of each hour, or change in pattern, the pumps either can't deliver the required initial head, or exceed maximum flow. At a lower accuracy in my analysis, the system is also unbalanced at each hour.

I was wondering whether anyone could be of any assistance?

Thanks for the time and consideration,

Tait.

 

[hydrae]

I have found rule based controls to be tricky
each command needs to be checked for correct control
have you got the model to work as a simple network first?

Taking a look at your model I see all the elevations I see are at zero. Is the town truly flat?

It looks like you have several pumps stacked on top of each other, I suggest you build the pump station as truly exists, those short sections of pipes between the pumps help the program balance the flow, I find it easier to stretch out the pump areas and manually enter the pipe distances in the pump station. reservoirs and tanks operate in the model best when you match the actual piping configurations, if there is only one or two pipes connected the tank and then branches out into a header, set up the model that way. Take the base flows and manually turn on the correct number of pumps to get your system operating at a single point in time

Once you have the model working as a single point in time then you can engage
the controls, and very the demands

What actually calls the pumps? is it by a flow meter as you describe or a pressure setting?
I find most pumps are called on low pressure and are released on either low flow or high pressure. Set up the controls as the real world is


Hydrae

 

[Tait]

Thanks Hydrae,

I have followed your advice and made some adjustments of my own and things seem to be flowing. Thanks for the advice! The town is very flat, but not truly flat at all points. I have neglected elevations at this point, but to make the system more reflective of real conditions I will have to change that. The pumps are called on when system demands reach a particular quantity, and turn off when the value drops about 10% below this. I have tried to make the system as representative of this as possible.

I do however have one more question you might be able to answer for me. I was told that EPANET is not a dynamic model, in that changes in demand multipliers are not applied to the condition of the system at the previous stage, instead the program demonstrates how the system would operate under these conditions at a steady state. For example, when I apply a pattern that increases from 0.5 to 4, the program simply shows how the system operates under these two conditions; it does not consider that pumps need to apply a dynamic flow increase to build pressure from the previous time step to reach the current demand.

Contrary to this, in my system when I apply a significant change in demand, errors often occur where pumps can not meet their instantaneous requirements. If I continue the demand multiplier, or create a gradual increase to this demand, errors are avoided. Generally it seems to take an extra hour for the system to balance itself and generate what I know as the real-world steady state flow.

Do you know this to be a characteristic of EPANET, or is it possibly something in my system?

Thanks again for your time and help,

Tait

 

[hydrae]

yes EPAnet is not dynamic, it does not calculate accelerations just velocities.
In order to calculate the accelerations would require an order of magnitude increase calculating code and processor power. It only calculates points in time and it uses the last time step as a starting point, so the slow adjustment in changes in demand gives the better results. The slow adjustment in changes will also model the storage calculation better.

Hydrae