I was wondering if someone can suggest how I can arrange and simulate private pump stations in my network.
I have a relatively small WDN of 505 households and which has been skeletonized to 30 nodes. A brief survey of the area found that about 30% of the households have private pumps which they use to fill their tanks when pressure is low. I am using WaterGems at the moment.
What I want to investigate is the effect of these private pumps on the performance of the distribution network and most especially how the use of the pumps affect other customers that have no pump. At the moment, the only information of the pump that I have are the Hmax and Qmax which I will use as a general representation for the pump curve. I am considering using the PDD function but I have a few questions.
The questions are:
1. How do I set up the pumps? (since I am only interested in effects of the pumps do I still need to put in the tanks or is there an alternative?)
2. Do I need any controls for the pump?
3. Can anyone suggest how to locate the pumps to represent 30% of the households?
Any other added suggestion will be highly appreciated
I know there might be insufficient information but I can explain further if need be.
Apologies for the delay in response.
If you only need to assess the instantaneous impact on pressure that these private pumps have on the rest of the system, you could consider running a steady state simulation with the pumps modeled as junctions. You would enter a demand on the appropriate number of junctions, based on an assumed maximum number of running pumps. For example if there are 30 nodes total and ~10 of those 30 have pumps, and those represent 505 * 30% = ~150 pumps, how many of those 150 pumps will be running simultaneously? If you wanted to make a conservative assumption on that, you could add an estimate of the respective pump flow rate as a demand on those nodes. For example if you assumed a maximum of 50% of the pumps running at the same time, you would add the demand on 5 of those 30 nodes. The demand would be equal to the flow rate of the respective number of running pumps, and the program would calculate the pressure in the network accordingly. Using the same example of 50% simultaneous running, if each individual pump has a flow rate of 1 L/s and the 5 nodes represent 75 running pumps, then you would place a demand of 150 / 5 = 15 l/s
As far as which nodes would be used to represent as pumps - since you are modeling 505 households (~150 with pumps) as 30 nodes, you may want to spread them out evenly. You could also set up a few scenarios to check the sensitivity to that distribution (try one scenario where they are evenly distributed, one scenario where most are on one side of the model, one scenario where most are on the other side of the model, etc)
However, if you need to get more detailed to see for example the impact that these private pumps have on the water level of the source, you may need to run an EPS (extended period simulation). This would allow you to assess the impact from the change in water level and the pumps turning on or off. However, you may not be able to get accurate results with this method unless you explicitly model all of the individual pumps, so that you can model all of their respective controls and tanks. The pump controls would be necessary to accurately model the rising and falling of the respective tank and its corresponding impact.
Therefore if a steady state is adequate, that might be the best approach.
Others on the Community may be able to provide further input.
Jesse DringoliTechnical Support Manager, OpenFlows ProductsBentley Communities Site AdministratorBentley Systems, Inc.
Answer Verified By: Sushma Choure
Most engineers who work with water distribution do not get into the details of individual house pumps. You never know exactly which pumps are running. The best you can do is estimate the demand over time and spread it evenly among the model nodes. At any given time, that estimate will be wrong but over time, it is better than any other estimate.
Pipe sizing is usually based on a reasonable worst case peak demand. If you want to model the pumps, just put a demand on nodes such that the sum of the demands equals the flow into the system. As the system gets smaller, the peak to average flow ratio increases. You might want to look at the factors that are used by the pressure sewer pump companies like e-one to get a handle on a reasonable number of pumps to be running at once and multiply it by the flow from a pump to see if the value is reasonable.
Concerning where to place these demands, the worst case is when the flowing pumps are far from the source.
There are some nice theoretical methods to assign a probability of certain flows in a system, such as the Poisson Rectangular Pulse model but you need data to calibrate that sort of model and I doubt there is sufficient data for a system such as yours. With that data you can estimate the probability that certain pumps will be running at any time. That's probably too much work for what you need to do.
Is this a system that has 24/7 water supply or are there significant times when the pipes run dry?
Do these pumps pump directly in line or does water flow into an ground level tank and then is boosted?
I'm assuming that the Hmax and Qmax values are the rated capacity of the individual house pumps. However, since I assume they are centrifugal pumps, this is no guarantee that they will actually produce that flow.
For the houses with pumps, do they pump directly into the water uses or do they pump to rooftop tanks? If they pump to rooftop tanks in system with intermittent supply, the worst case is when flow is restored after being *** off and virtually all of the pumps come on at once.
As you can see there is a lot of judgment involved in designing this sort of system.
What if the situation as you mentioned above "does water flow into an ground level tank and then is boosted?"
In other words, if the water flows in to a domestic tank under network pressure without involvement of any pump and that water is pumped to a rooftop tank subsequently.
How to model a this kind of a situation?