DP Control of a pump

Hello all,

I have a hot water network (district heating) that is driven by VSDs trying to maintain the Differential pressure of the network at a certain SP. Generally the process is very slow. The DP is collected at a remote location (worst DP branch) and it is transmitted to the PLC controlling the pumps via GPRS. The Actual DP is collected every 30-60 seconds and it is defined by the user.

My concern is has to do with the control method I should follow.
If I were to choose a PID control then my concern is that the PV is collected every minute and so I guess that would not work as it is not real time data.

The second method I am considering, that has been used by other system integrators is some sort of step control.
I have a DP setpoint, the actual DP, a deadband and an evaluation time. I do a comparison to check if the actual DP is outside the DP+/-Deadband zone. If I am below the DP SP then I add step (say 5%) to the Pump SP. If I am below the DP SP I subtract the same step from pump SP. If I am within the deadband zone, I do nothing. This calc is repeated cyclicaly at a time set as the evaluation time (say 30sec).
This method is reported to have good results at similar installations.

What are your thoughts about this? Can you think of a better way?

Although it does sound nifty (and you Sai others have done it) I find myself wondering what you.aregonna do if u are way out of setpoint. you have to calculate amount of error and adjust the pump more depending on error. but not too much. to.make it work well you will need to spend some time manually tuning it.

What you describe is floating control, and I have found it much superior to PID in many applications. The tuning is more intuitive and the stability is better. You process looks like a good candidate. If you are worried about the large error problem raised by DwSoFt you can add proportional response.

Most processes don't need (and can't achieve) zero error. Therefore adding tolerance won't hurt operator acceptance. Most operators would rather have stability.

Tom that is exactly what I am doing. Thank you.

if you are far away of setpoint, you maybe want do larger corrections. Couple different steps with different errors makes your "step control" little bit faster.
Eventually it won't be much different than PID-controller which is tuned to P-controller with long sample time.

I have integrated in the logic a half step and a double step function.

If I am far away from SP, it doubles the step. If I am a reasonable distance from SP I use single step and if I am close to SP I half the step.

The only other thing I could suggest, which would be getting halfway back to a PID-style control, would be to check your error and then adjust by a set amount based on that error. So instead of having three stages ("double step", "single step" and "half step") you have a basic calculation that works out you are x units below SP, so change the control output by x*MagicNumber %. Where MagicNumber is a number you pluck from thin air and then adjust until it responds the way you want

I guess at the end of the day, that's basically a P-only representation of a PID controller, only much simpler to look at!