Modelling startup of VSD booster pumps in series - first pump already operating

Question

I am trying to model a 'changeover' in operation for a transfer system with a number of booster pumps in series (Pump Stations A, B, and C) to determine any transient impacts. 
 
 Initially, both PSA and PSC are operating at a given flow rate. The client's intended procedure for starting up PSB is as follows 
 - ramp down PSA and PSC to 'idle' speed (~50% speed) 
 - turn off PSC 
 - turn on PSB and ramp up to idle speed 
 - turn on PSC and ramp up to idle speed 
 - ramp up all pumps to full speed (at a higher flow rate than the initial conditions with only PSA and PSC operating). 
 
 I have all of the relevant speed-change patterns. My problem is that if I try to use the variable speed/torque pump type, although the pumps all turn on at the correct time, I don't get any "use" out of PSB, as both PSA and PSC go back to their original flow/head conditions - I can't figure out how to increase the flow and head to the new desired conditions. 
 
 As a workaround, I tried to run this scenario in two parts, first assessing the shutdown to idle speed (no problems) and then looking at a startup, but then again ran into problems with PSA when I tried to start it at an initial speed factor of less than 1.0 and ramp it up to a new flow condition alongside the others. So the THIRD thing I tried was to run PSA at full speed, ramp it down to idle in an initial (dummy) time period, and then start up the other two pump stations and start recording my transient conditions. Again, I could not get the full three-pump flow. 
 
 Finally I have tried to do a complete pump startup, with all three pumps off to begin with, ramping up PSA first to create the desired initial condition before starting up the other pump. But to do this I have had to artificially adjust the system with a downstream PSV to keep the pipe full and prevent the occurrence of vapour pressures across the highpoints, so that although I can make all the pumps start up in accordance with the patterns, the overall HGL is artificially inflated. 
 
 I hope this all makes sense - essentially I'm wondering whether what I'm trying to do (changing target flow rate of a speed-controlled VSD midway through a simulation) is actually possible? 
 
 Thanks, Liz

Scott Kampa · Accepted Answer

Hello Liz, 
 
Make sure that the model run time extends out past the timing in the patterns. In the model you sent, the run time is 18,000 seconds. This matches will with the "pump ramp up" patterns in the model, but not with the "full pattern" patterns that appear to match with what you stated you were trying to do in your original post ("ramp down PSA and PSC to 'idle' speed (~50% speed), turn off PSC, turn on PSB and ramp up to idle speed, turn on PSC and ramp up to idle speed, and ramp up all pumps to full speed"). 
 
When I use the "full pattern" operating rule and extend the run time to 30,000 seconds, the flows found in the three pumps are a little over 400 L/s at the end of the simulation. A good test to make sure that results are accurate is to set the model to the expected settings at the end of the model and compute a steady state simulation. When you do this, the flow through the pumps is a little over 400 L/s, so it appears to be in good agreement. 
 
So I would make sure that the model run time is correctly entered. If you only compute for 18,000 seconds with the "full pattern" operating rules, the flow at the end of the simulation is lower, though still matches the pattern; it just hasn't reached the end of the operating rule. I tested this in HAMMER V8i SELECTseries 4 as well and saw the same results. 
 
However, as mentioned earlier, everything moves so slowly that the transient impact is minimal. What sort of transient impact are you looking for? If it is just to see pump operation, it may be easier to model this as an EPS simulation in WaterGEMS. 
 
Note that if you are us a different pattern/scenario, please let me know. This is assuming the the user of the "full pattern" operating rules. If you are using another operating rule, I will need to know what those are. 
 
Regards, 
Scott

Scott Kampa · Answer

Hello Liz, 
 
We may need to see a copy of the model to getting a better understanding of the setup of the pumps and the model in general. However, there are a few things that you can look at to see if they help. First, for the pump that is initially off, make sure that the nominal head and flow are accurately entered. This head and flow is used by the program to assign the accurate four-quadrant pump curve used by the transient solver. 
 
Second, if the pumps that are initially one have a relative speed factor that is less than 1.0, you will need to make sure the pattern on the pump is correctly reflecting this. The following support solution has some information on this: communities.bentley.com/.../16427.initial-surge-occurs-when-the-initial-relative-speed-factor-is-less-than-1-0 
 
If this doesn't help with the issue, please provide a copy of the model files. There are two options for sharing your model files on BE Communities. If you would like the files to be visible to other members, compress the files into a zip file and upload them as an attachment using the ‘Advanced Reply editor’ before posting. If your data is confidential, you can follow the instructions in the link below to send it to us via Bentley Sharefile. Files uploaded to Sharefile can only be viewed by Bentley. 
 
 communities.bentley.com/.../7079.be-communities-secure-file-upload 
 
If you upload the model to Sharefile, please post here with the name of the file so that we know it is available. Thank you. 
 
Regards, 
Scott

Scott Kampa · Answer

Hello Liz, 
 
A smaller run time would help with troubleshooting, but it would also impact the results. For instance, currently PS B takes 720 seconds to reach half speed. This is longer than model HAMMER runs, and the transient impact of such a slow buildup of speed will likely be pretty small. 
 
A better option may be to model these as separate events, since the time scales suggest that is what they are, at least in turns of transient time scales. For instance, model a case where PS C turns off. In a separate model, model PS B turning on, and so on. That would eliminate the need to force the pumps to operate in a single long model run and still will likely give you viable results for the individual events. 
 
Regards, 
Scott

Scott Kampa · Answer

Hello Liz, 
 
I tried a time step of 1 second and the whole model run computed faster. There is not much in the way of transients, but in looking at the flow results for the pumps, the pumps do seem to turn on as expected based on the operating rule. I tried both the operating rules the model was saved with and with the "full pattern" operating rules that seemed to match your description above. 
 
So you know, HAMMER V8i SELECTseries 5 has a new feature in the Transient Results Viewer that allows you to see graphical results that were previously only available in the Detailed Report. For pumps, speed is one of these, and it is a good indicator that the operating rule is working as expected. If you cannot upgrade to SELECTseries 5, you can still see pump speed results in the Detailed Report. 
 
Regards, 
Scott