Calibrating network model

[ZAID24]

Hi, does anybody know where I can find good literature on how to collect field data to be used to calibrate a distribution system model? I am working on a complex system with 7 pressure zones. I've found material on the process for calibration but not much on the data collection procedure itself.
Thanks

 

[BigInch]

A steady state calibration is relatively easy. With flowrates, pressures and temperatures at the inlet and outlet of each zone you should be able to bracket it in, but you'll be very lucky to get the flows corresponding to a known pressure exactly at the inlet and outlet of each zone without doing a lot of work and adding up all the flows from the take meters located in the zone itself. Basically your calibration accuracy will improve with each data point you can get your hands on.

What is difficult is calibrating a model to a system that is not in steady state. It is even more difficult calibrating a steady state model to a dynamic system. Both usually turn into a probabalistic calibration with emphisis on the "balistic" part. There you must get as many pairs of data as possible for as long a time history as possible and try to mesh them together to get something that makes sense. It often requires identifying several (or many) major steady states and making a corresponding model of each and then determining how they change between them to find a good time point to step into it.

BigInch-born in the trenches.

http://virtualpipeline.spaces.msn.com

 

[hydrae]

Zaid24
If the system has telemetry, that data is most prized, flow rates at sources and recordings of pressure in the system and levels in reservoirs will aid in predicting the flow and pressure in calibrating the model.
lacking such data, get yourself a pressure chart recorder, about $500 and place it in locations of choice for a week, at sources, at ends, at reservoirs.
To get flows, read the meters at the sources (aka master meters) on very frequent basis, during peak flow in summer once per hour is nice if you have the time else daily will also help
A tool that can also help is a strap on flow recorder, it wraps around the meter head and reads the spinning of the transfer magnets and records the flowrates. (about 2k)

Another method includes, once you have baseline data above, perform fire flow tests, measuring static pressure, residual pressure and flow. Your model should predict the responses to the flow test, be sure you know the base flow demands, source feed rates and tank levels during the tests.

Be sure you know how those 7 pressure zones react to each other, else unknown connections that only operate intermittently can skew your results.

Hydrae

 

[cvg]

fire flow tests should also be accompanied by corresponding information (if possible) of tank levels and pump on/off status. Record accurate time/date stamp information on all tests. If there are any large industrial users, also good to know what the typical usage is. Also, you will need to coordinate with the O&M folks to make sure there are no valves closed - or if so, determine which ones.

 

[niibii]

Testing of supply pumps and boosters, loss of head tests, pressure testing on fire hydrants, hourly flow monitoring of key trunk mains and large-demand water users. Also, you will need the storage levels of tanks or reservoirs. You can find more literature in the AWWA M32.