Pump System Curve Symantics

[BRIS]

We have a complex distribution system which is expected to experience a demand growth of some 50% over the next 10 years. The system has several sources of supply all from desalination plants which pump directly into the system through ground storage tanks. There are also a number of evaluated tanks. We are redesigning one of the pump stations that supplies the system.

My predecessor appointed a sub consultant and asked for a system curve expecting to get a curve of H v Q at the pump station in relation to the change in consumer demand on the system. I would also expected to get a series of growth curves showing system curves at different stages in the growth in demand.

What we have got is a curve of Q v H for the new pump station which is based on constant consumer demand but varying inflow from the other pump stations into the system. It takes a steady state situation and varies Q which automatically changes Q in from the other sources of supply (Qi = Qo) . It assumes that an increase in flow into the system from the new pump station is met by a reduction in flow from the other supply stations. The sub contractor argues that this is a system curve and that it complies precisely with the definition of a system curve - it is a curve of Q v H.

Yes my predecessor should have specified exactly what he wanted not just ask for a system curve.
My view is that on a distribution system the definition of a system curve is a plot of Q V H as seen by the pump station due to variations in consumer demand ?
This may be a case of semantics, but we are now in a position of having developed an operating philosophy on what we thought was a system curve of H v changes in consumer demand - not changes in contribution from other pump stations feeding into the system.
Interesting discussion point ?

 

[BigInch]

The short answer is, depending on the geometry of the piping system, supply, demand and pump station locations, the system curve might be the same, should it be based on supply or demand. Theoretically for each (steady state) condition, the demand balances supply, supply balances demand. As long as there is only one unique hydraulic scenario (number of pumps and pipeline segments operating and flows in each segment, etc.) which can meet the supply-demand condition at hand (which in a heavily networked water distribution system might be rather uncommon to say the least), the system curves will be equal, wether you say supply is the driving factor, or if you say demand is the driving factor.

If there are a number of different hydraulic permutations that can meet a given supply-demand scenario (pump stations 1 and 2 on, or pump stations 1 and 3 on, then you could see different heads at one or the other pump stations, even though flows remained the same, because when pump station 2 is operating it may affect the head at pump station 1 differently than if pump station 2 was off and pump station 3 was operating.

Considering your different supply flow assumptions,
It assumes that an increase in flow into the system from the new pump station is met by a reduction in flow from the other supply stations.

I would argue that, if there are a number of equally valid hydraulic design-operating configurations that can meet a given supply=demand scenario, the configuration that should be adopted to evaluate the head-flow system curve, or the performance of the entire system for that matter, should the one configuration that yields the lowest design-operating cost for that particular supply=demand flow scenario. It might be appropriate to consider some kind of a redistribution of supply flows from various other points, but I would say only if it had the potential to reduce the operating cost for that scenario.

 

[LittleInch]

BRIS,

I think in reality you're asking for something which is not possible to provide as there are simply too many possible combinations and flow patterns.

What I think you should have asked for is a number of simulations based on a set of criteria which provides the best case and worst case scenarios for current flows and then future years , say 1,2,3,5,10,15 based on some sort of future demand curve.

Given that in a network model ( this sounds like a water distribution system) with multiple inputs and probably even more multiple offtakes, there are a mind boggling number of permutations. What you need to do is create as accurate a network model as you can and then constantly run different combinations of inlets, flow outflows, new connections etc.

This is what network operators need to do on a continuous basis and predict where the network needs re-informant or additional flows to meet demands. A live system taking live data is often used by network operators to continuously model different operating parameters to have the lowest cost system. Unless you have a very stable set of flows that change very little over a day or week then you simply have too many combinations.

I'm not surprised your consultant gave you what they did - they had to work with something so chose a particular system they could model. What you describe is not possible to model as a simple system curve. IMHO.



Remember - More details = better answers
Also: If you get a response it's polite to respond to it.

 

[BigInch]

LI you can develop a Q-H system curve for any point in a system. It is true that for any given Q, H may vary widely depending on what the rest of the network is doing, however generally the variations in head will not be extreme for similar flows, as long as only reasonable operating permutations are considered. This is even more so for a complex water distribution network because the numerous pump stations and higher parallel paths that the water might move through typically reduce overall head losses for any large flow rate and also offer a number of possible paths of similar head loss for lower flowrates, thus extreme ranges of pressure-flow responses are reduced to a more average response, especially from those that are seen in simpler, very linear pipeline systems. If you discount any obviously unreasonable operating configurations, you can quickly see typically good characteristic relationships between head and flow for most points in a networked system. The more redundancy, the better the relationships hold.

 

[cvg]

as LI suggested, you need a number of scenarios.
those should probably include average daily and hourly flows, night time replenishment flows, fire flows. For redundancy, you may want to look at cases where other pumps are out of service. You can look at future flows, but don't go too far into the future since a lot can change. Finally, you probably also need to look at scenarios for upgraded transmission mains and additional reservoirs which almost certainly you have some of those planned as well. having done this type of modeling for a number of years, I can say that most large operators have a working hydraulic network model that is constantly used for modeling "what if's". This can become a full time job.

 

[Nick Wattss]

This really is very complicated issue. There is need to take look at various scenarios from water flow to every corner. Plan new network model and work progressively.