Small Pressurized Water Distribution System

Question

Hello, 
 I&rsquo;m trying to model a basic water distribution system for a small rural community. They currently have an existing pump and some water distribution pipes in place for just a few houses (30 or so) in the community. Our project is to extend this system to the entire community (100 households spread over 1 square mile). In my model, I would like to make the pump pressurize the system. I would like to determine whether the existing piping is useable, and what piping should be used for the rest of the system to achieve pressures between 10 to 80 psi at each house. 
 The pump has some problems and so the pump curve provided by the manufacturer is incorrect. The pump is at a depth of 170ft from the surface and we measured a flow rate of 37 gallons per minute at the surface directly from the pump. With this data, I was able to use the technique described at the link below to generate the pump curve: 
 http://communities.bentley.com/products/hydraulics___hydrology/w/hydraulics_and_hydrology__wiki/16879.estimating-a-pump-curve-for-a-model 
 With this pump curve I now have some confusion about how to model my demand. 
 My understanding is that there are three ways to model usage points in the system: (1) set a demand at each junction in junction properties (2) input an emitter coefficient at each junction in junction properties (3) connect a junction to a discharge to atmosphere and input characteristic pressure and flow to model the orifice. In the past, for another much simpler system that was only gravity fed I used the first method. I rationalized this because all junctions were more or less the same elevation and distance away from the tank located atop a hill. You can read more about the discussion about our past project here: http://communities.bentley.com/products/hydraulics___hydrology/f/5925/t/84702 
 For this new system, the junctions are definitely inequivalent. Recent flow rate data taken from three points in the existing system gives flow rates of 37, 8 and 24 gpm. Additionally, some community members report flow rates close to 1 gpm at the extremities of the current system. Hence in my model it seems very incorrect to input a fixed constant demand at each junction i.e. the first method would be incorrect to use. Is this reasoning sound? 
 It is also not possible to use the third method because many of the junctions do not exist yet (we intend to add them and want to model them first). Hence it will be impossible to determine the characteristic flow and pressure for them. 
 I attempted to use the second method. In order to determine the emitter coefficients of the existing junctions I used the existing flow rate data. My calculations are available on the spreadsheet at the link below for anyone to check or scrutinize. I will quickly outline the principle. We know the pump curve, flow rate and elevation at a junction. Thus I can calculate the head in feet via Bernoulli&rsquo;s equation or by doing a point to point model in WaterGEMS (both of these give reasonable agreement with each other). I can calculate the discharge coefficient C by the formula C = Q/(A*(2*g*H)^.5) where Q is flow, A is pipe cross-sectional area, g is gravity and H is head. I can then calculate the emitter coefficient k using C by the formula k = C*A*Sqrt[2*g]. 
 Spreadsheet with calculations: https://docs.google.com/spreadsheets/d/1RSBsPuBipl-xtQ4doBvX_RMIqGYKs96y7Itqe4k1jHI/edit?usp=sharing 
 My sources for formulas for discharge coefficient and emitter coefficient are below: 
 Discharge coefficient: http://communities.bentley.com/products/hydraulics___hydrology/w/hydraulics_and_hydrology__wiki/3443.modeling-reference-discharge-to-atmosphere-tn 
 Emitter coefficient: 
 http://communities.bentley.com/products/hydraulics___hydrology/w/hydraulics_and_hydrology__wiki/27399.options-for-modeling-an-outflow-that-varies-with-pressures 
 Finally, as a sanity check, I can input my flow rate at a junction in my WaterGEMS model of the existing system to see whether I obtain the flow rate I would expect (because I calculated my emitter coefficient from a known value of flow rate). Unfortunately, my flow rates are off by 15 to 20% or more when I calculate a separate emitter coefficient for each point (these errors are on the spreadsheet). When I use a single average emitter coefficient some of these errors are as high as ~200%. 
 My main questions are below:

Is my overall methodology and reasoning correct that it is incorrect to input fixed and equal demands at each junction?

Have I made any mistakes calculating emitter coefficients? Should each junction have a unique emitter coefficient? Or is there a single typical value I can use?

Should I be using another software to determine my emitter coefficients such as Bentley Flowmaster?

I have attached my WaterGEMS model at the link below: 
 https://drive.google.com/drive/folders/0B1HPXaOsy_b8ZXpzb0NNZkZ5cjA?usp=sharing

Tom Walski · Accepted Answer

There are a number of issues here. 
 
First, it's very difficult to accurately calibrate a model with very small pipes because the flow in any pipe can vary dramatically depending on who is taking a shower or running their sink at any given time. In these small systems, the demands vary widely with time of day. If you connect a pressure gage somewhere away from the tank, the pressure will vary significantly with time. It helps if you have a pressure recorder or data logger to look at the variations in pressure. 
 
When you say "flow rate data" is 37, 8 and 24 gpm, how are you measuring it? Is this the flow in a pipe or is it a nodal demand or the flow from the source? 
 
In systems like this, the demand is somewhat pressure sensitive (but no exactly like an emitter, unless customers just let faucets run all the time). I suggest you read up on our pressure dependent demand calculations. There you specify the flow you would expect if pressure isn't limiting and the model will calculate the pressure and flow when pressures fluctuate. 
 
For pipe sizing calculations, you would load the model for the demands you want to meet and size the pipes so that the pressure doesn't drop too much at peak demand. Also check a scenario with low demands to make sure the pressure doesn't get too high. 
 
Best wishes, 
Tom

Tom Walski · Answer

In a small system, a few water users can make the pressure vary quite a bit. The model will give you what the pressure should be for the demands that you loaded. You can compare that with the pressures you recorded to determine if the model is in the right range. 
 
You really should measure the actual flow at the pump station. In the US, you are required to have a flow meter at a pump station. If you measure at a user, you may be missing flows that are lost to leakage. A flow meter would be a good investment. 
 
You may want to also contact the manufacturer to obtain the pump curve when the pump was purchased and see how that compares with you measurement.