PID level control of sequential pumps

drbitboy said:

Yes it's essentially correct.

CV_n above will be equal to CV_n-1.

Click to expand...

i.e. If Kp=1, Ki=0, Kd=0, and for example SP = 8.5 SP(n-1)=PV=PV(n-1) = 10
I suppose CV(n-1)=0, so CV=CV(n-1)=0

MaxK said:

i.e. If Kp=1, Ki=0, Kd=0, and for example SP = 8.5 SP(n-1)=PV=PV(n-1) = 10

Click to expand...

In that case, if SP_n-1 was 10 on the last update, thenWhether it is plus or minus depends on whether the PID is direct- or reverse-acting.

CV_n-1 becomes, at the next update, n, the Bias term in the "Positional" form of the PID equation:

MaxK said:

Possibly off topic
I Please to excuse me for such noob question:
PIDE formula:
CV = CV(n-1) + Kp * delta_E + Ki*E*delta_t + Kd*(E - 2*E(n-1) + E(n-2))/delta_t
is it correct?

If Kp=1, Ki=0, Kd=0, and for example SP = 8.5 PV=PV(n-1) = 10

What is the value of CV ?

Click to expand...

Undetermined unless you set it to some value almost like a bias. The incremental form of PID doesn't work unless there is an integrator gain.

skyfire said:

Now, if I could redirect everyone back to the original questions, I would appreciate it.

1. Is it worthwhile to use a PIDE block here over a regular PID block, considering I will need to spend a little time familiarizing myself with it and what useful options should I look for (since there are so many)? I am leaning towards using the PIDE just on the basis of rate change limits and bumpless handover, but that brings my next question.

2. What are recommendations between using the PIDE's built in rate of change limiter vs using an RLIM external to it? Any practical difference? My initial plan is to use both, with the internal limiter set to be a little faster for more responsive control and using an extra RLIM on pump start that brings it up to speed a little slower before bypassing it.

3. From a control strategy overview, any high-level recommendations on setting up the control? My plan is a single PIDE block (potentially tuned to PI), when adding pumps ramp them up at a fixed rate until their speed matches the PIDE CV speed (which will be falling due to the new pump) and then hand them over to the PIDE control which will be scaled to the number of running pumps.

Click to expand...

1. PIDE has an advantage when using adaptive gains:



2. Use the PIDE rate limiter as an overall limiter and use a separate limiter when ramping on/off

3. When bringing on a new pump:

• start at minimum speed
• wait for a period of time to allow the PID to adjust
• ramp to PID CV

Peter Nachtwey said:

Undetermined unless you set it to some value almost like a bias. The incremental form of PID doesn't work unless there is an integrator gain.

Click to expand...

It will just stay at the previous CV.

Process Variable tracking can prevent this issue when transitioning from manual to auto.

Peter Nachtwey said:

Undetermined unless you set it to some value almost like a bias. The incremental form of PID doesn't work unless there is an integrator gain.

Click to expand...

Not quite; CV_n-1 is, in effect, both bias and accumulated I-term (Σ K_i E Δt).

The incremental form is mathematically the same as the absolute form, and it works fine without an integrator gain. There may be roundoff in the sixth decimal place.

skyfire said:

My plan is a single PIDE block (potentially tuned to PI), when adding pumps ramp them up at a fixed rate until their speed matches the PIDE CV speed (which will be falling due to the new pump)

Click to expand...

Not necessarily.


Do you know what the ramp rate limit will be (%/minute or %/s)?

Do you know what the static head of the lift station is?

Generally you don't need to ramp up any slower than 10-15 seconds from minimum speed to 100%. Fast ramp to min speed in 2-3 seconds to avoid binding the pump impeller if there's any rag in the product (you'd be amazed what people flush...). The OPs minimum speed of 20% seems a bit low... failing any advice from the design guy, my limits are 30Hz min to 50Hz max (we are a 50Hz country) which is 60%.

I think OP is unlikely to give you the full dentists office worth of teeth you so desperately require DrBitBoy. I have a project coming up in a couple of months with both a tank level control where the intent is to keep stable outflow, and a seperate lift station for recycle water flow back to the plant inlet. I'll have access to real plant operational data and all the design details. I'll post them here and you can go wild.

Saffa said:

...I'll post them here and you can go wild.

Click to expand...

... with bated breath ...

drbitboy said:

Not quite; CV_n-1 is, in effect, both bias and accumulated I-term (Σ K_i E Δt).

The incremental form is mathematically the same as the absolute form, and it works fine without an integrator gain. There may be roundoff in the sixth decimal place.

Click to expand...

It requires cv(0) be set to a good value initially.
All it does is add complications.

Peter Nachtwey said:

It requires cv(0) be set to a good value initially.
All it does is add complications.

Click to expand...

tradeoff is that anti-windup complications are eliminated. six of one ...

To exclude further flooding of the topic. Let me draw some conclusions:

Possible options level control of sequential pumps:

1. Turning on the pumps by levels (of course with dead zones)
pros: an algorithm understandable to anyone
cons: scientific and technological achievements are not shown

2. P-control (of course with dead zones)
The result is identical to point 1
pros: it is absolutly clear that the work was performed by a highly qualified specialist (used a PID controller)

3. PI controller (of course with dead zones)
pros: level control (bring the level to the setpoint)
cons: the need for a specialist who at least superficially understands the operation of the PID controller, knows the tricks of the I-component and knows how to deal with them.

I agree with @Peter Nachtwey in essence, but disagree with the wording. Therefore, since on this forum I have already been convicted of being a rude, toxic beast, I will give my wording:

In level control systems, in the absence of strict requirements for maintaining the set level (which is the most common case), the use of a P-regulator will be sufficient - this compensates for the difference in flows and stabilizes the level, but not at the set value. The use of a PI controller will allow you to bring the level exactly to the set value, however, the use of the I-term requires minimal knowledge, which in 99.999 ...% of cases, specialists tuning PID controllers do not have.


P.S. I consider the stability of the function CV = CV(n-1) + Kp * delta_E (taking into account the conditions of real systems) mathematically unprovable. i.e. using PIDE P-only is unsafe

MaxK said:

P.S. I consider the stability of the function CV_n = CV_n-1 + K_P * ΔEn (taking into account the conditions of real systems) mathematically unprovable. i.e. using PIDE P-only is unsafe

Click to expand...

Actually, as long as Δt is small enough, it is quite provable.