View Full Version : How to install two PRV's at same location?

December 8, 2015, 02:16 PM
Need to install a 4" and a 10" PRV in the same location. Does anybody have an idea how these should be graphically represented in the model?



December 11, 2015, 07:42 AM
As far as graphically, I usually just offset them parallel so they're easy to click on.

Hydraulically, you probably want to make sure they're separated at least by a link (i.e., don't connect them to the same node), and you probably would want the setpoints to be slightly different...at least a few psi, at least fictionally if they're not in actuality, so that they won't trip each other on successive convergence iterations. Adding minorlosses on the individual lines can also help achieve this as it will make one valve hydraulically favored.

Patrick Moore
December 11, 2015, 09:53 AM
Brian had a very nice response up above and covered many of the key points. I just wanted to post this for you as well to help guide you in your decision process as well.

With the increased detail being included in GIS systems, things like multiple PRV's at a location are being seen in models. While one can model multiple PRV's at a site, in order to make the model look just like the real system, this type of configuration can cause the potential for mode instability for any model that uses EPANET as its hydraulic analysis engine like InfoWater, H2OMAP Water, and H2ONET, so one should consider all options before adjusting the model as we often get cases of clients having problems getting models to converge when they use multiple valves. Some clients have fewer issues, others have more.

Typically larger PRV vaults have multiple valves (sometimes two, sometimes three) because one valve cannot handle the entire expected flow range for that PRV station. So multiple valves are used to handle the various flow ranges expected. The valves would be expected to be sized such that the operational flow range of each valve would just barely overlap which could result in something like a 2”, 6” and 16” valves to get flows from roughly 60 to 40,000 gpm. The 2 inch could supply flows from 60 to 600 gpm, the 6 from 550 to 6000, and the 16 from 4200 to 42,000 gpm. Each valve typically has a 1-10 range of flows like 60-600 gpm or 600 to 6000 gpm.

Usually the smallest PRV is set at the highest setting and will flow first, the next largest valve is then set 5 psi below the small valve and the largest valve set 5 psi below that valve. Thus, as the flow increases the effective downstream setpoint of the valve will effectively get lower as the flow increases through the vault. If one models the vault with multiple valves here are a few key considerations:

The multiple valves should not share the same downstream node. EPANET will issue an error if two PRV's share the same downstream node. It is often best to draw a valve vault using a "manifold" type layout as shown below in the example:

(click for larger image if necessary)


Assign each valve the same elevation (this way the pressure change is the only difference in the HGL of the given setting)
Assign each valve a minor loss coefficient.

This is key or 100% of the flow will be supplied by the small valve in the model if you don’t. The ML coefficient induces headloss through the valve such that at some point the internal headloss will exceed the head available upstream and cause the valve to run open in the model like a pipe with headloss governed by the ML headloss equation.
Minor Loss Headloss equation: HL_ml = K *V^2/2g where in US units g is 32.2 ft^2/s^2 and V is the velocity in feet per second (fps)

Assign the settings to each valve at least 5 psi apart. This is typical in the field and is about the closest in settings that the model can usually handle to get the best chance both valves don’t “fight” for control in the model.

The problem that can arise when you do put many of these into a model, is that two points of a fixed head value that are close together (think elevation plus the valve setting converted to feet) is a situation that is difficult to solve mathematically within the modeling software. The model tries to match each HGL exactly, but it is not always possible to do so and this leads to non-convergence in the model solution even after the maximum allowed iterations is achieved. As such adding a bunch of PRV’s in your model close together greatly increases the chances that two or more PRV’s could fall into this category and cause your model to no longer converge. While increasing the number of iterations and adjusting the advanced parameters in the simulation options helps, models with a high number of PRV’s , and especially models using multiple PRV’s in a single vault, can have an almost inherently high chance for non-convergence. This unfortunately will always remain in a model and while one can get it to run under one condition, the model may not converge under different conditions. This is often very frustrating to the modeler when this occurs. So in light of this, just understand this is a possibility that can occur with adding multiple PRV’s in each vault especially if it is done system wide.

One thing many modelers do is to use the largest valve but set it at the setting of the smallest valve. If one is concerned about the setting drop at higher flows, one could place a minor loss coefficient on the valve downstream pipe such that as flows increase the downstream pipe headloss can replicate the setting drop expected as the larger valves kick in. One can calculate a ML coefficient such that at the average flow range of the larger valve the headloss in the pipe increases 5 psi so that the effective downstream setting replicates that used in multiple valve situations. The main advantage of this, is that you eliminate model instability while being able to still replicate operation pretty closely.

Example. if a 6 inch valve is 600-6000 gpm the average valve flow would be 2700 gpm. If the downstream pipe was a 12 inch pipe what minor loss K would induce 5 ft of headloss at 2700 gpm? As it turns out this would equal a K of 12.67. This ML would produce 5 psi (11.55 ft) of headloss at 2700 gpm but would only produce roughly 1.6 ft of headloss at 1000 gpm. The downside can be that the higher flow headloss can be too much, but you can adjust which flows you use it for to get it best for the most likely high flow value. See below:
(click for larger image if necessary)

The other option you can consider is to use a general purpose valve downstream of the PRV (NOTE: It's probably best to separate the valve from the PRV by at least 1 node) where you can impart a headloss vs. flow curve such that at the flows for each valve go up, it adds the 5 psi headloss to account for the lower setting. This would generally exactly replicate the multiple valves but would not have the same challenges of instability during iterations that multiple PRV's would.

Good luck in your modeling efforts. We hope this helps you add the elements and understand the potential challenges of using multiple PRV’s in the model simply due to the way it solves the equations so you can make the best decision of what would work best for your model.

Pat Moore