Frequency drive running on agressive PI

I have a question about the current reaction.

My actual-value is an underpressure sensor 0 to -1 Bar.

The Setpoint is -0.25 Bar.

At that point, the drive is controlling a frequency around 40 Hz.


When I choose for a slow PI regulation:
P = 0,2
I = 600s
Deadband = 0,7%
Then I get a smooth current consumption, even when the actual value has peaks, the frequency is allmost constant en so is the current. It's stable around 80 Amps.


When I choose for a good and fast PI regulation:
P = 1
I = 2
Deadband = 0,5%
Then the frequency I unstable between 1 Hz, so not that unstable, but the current acts very nasty, also with high peaks when the actual value changes. It's unstable around 70 till 90 Amps.


I know this because I made trends in WInCC

The second PI is the better one for keeping the actual value good on the setpoint, but the current is not so constant.

My question is, which one is the best for the motor... ?

And wich one is the best for the frequency drive ?

I assume that the variable speed motor is driving a vacuum pump creating the negative pressure. If it is, then I don't see any need for deadband and would tune the system with it set to zero.

It would be interesting to see what P and I values you get when you do an autotune. That would likely be too soft or conservative so you could tighten up from that point.

Do you have the drive Max and Min Speeds set wide enough so the setpoint controller can work as it should. Are the ACCEL/DECEL regulators inside or outside the control loop. On many drives, they are inside the control loop where they do not belong. In that case, you need to set both accel and decel time to zero and retune.

Try these and let us know what you've got.

This is with the agressive PI regulator: P = 1 and I = 2:




This is with the slow PI regulator: P = 0,2 and I = 600s




I do not have an autotune function in S7 I think. The second one is not bad, but doesn't regulate fast and doesn't regulate very good to setpoint.

Is PID the solution instead of PI ?

I need a way to compensate or ignore the peaks in the black (Bar) line. These are the cause of an other machinepart in the plant.

Maybe a way to calculate where the peaks are and PI regulation between the peaks or something ???

Okay

This is an example when it doesn't regulate the way we want, the underpressure we want is -0,25Bar, and it's around -0,25 Bar. But as you can see, the other machine distubes this rugulation more offen now, it depends on what other process happens in that machine.

But can I make it stable in some way: Little current peaks and -0,25Bar good regulation ?

Hello,

By looking at these screenshots I would say that you have problem with your presure(vacum) transmitter. I can not claim that since I do not know where it is applied but anyway you should not have that kind of signal.
We once had a problem with cosistency transmitter and look at screenshots before and after intervention:

I would agree with Krcedinac. What is causing the repeating waveform on the vacuum signal? Is that actually happening or is the sensor feeding back something that really isn't there.


If the vaccum signal is real you need to pick you poison. If you want close setpoint tracking you need to run high gains. High gains mean high currents. You can probably help yoursel some by running a mid-range P (something like 0.5) and a much lower I (something like 50s). Most of the current deviation is due to the proportional term.

However, I don't think the current values you are seeing will really cause a problem. Unless you see something in your mechanical system that says the current spikes are causing a problem I would use the tuning values that give you the most favorable process results.

Keith

The sensor is measuring well, because it's doing actually the same as a manometer on the pipe. Sometimes we loose underpressure because of a valve that opens, it's normal. But we were searching for a compromis between 0,25 underpressure setpoint and no current peaks.



I've tried to smooth the sensor is S7. So in the folowing trendings, you are seeing (the black one) the actual sensor input, not the smoothed one, but you can see that the frequency is smoothed because of the smoothing before the PI regulation, and the peaks are less big. The setpoint of 0,25 is kept well, so we will evaluate this for 1 week and if it's okay, install on all cavuumpompsystems

kamenges said:

I would agree with Krcedinac. What is causing the repeating waveform on the vacuum signal? Is that actually happening or is the sensor feeding back something that really isn't there.


If the vaccum signal is real you need to pick you poison. If you want close setpoint tracking you need to run high gains. High gains mean high currents. You can probably help yoursel some by running a mid-range P (something like 0.5) and a much lower I (something like 50s). Most of the current deviation is due to the proportional term.

However, I don't think the current values you are seeing will really cause a problem. Unless you see something in your mechanical system that says the current spikes are causing a problem I would use the tuning values that give you the most favorable process results.

Keith

Click to expand...