Details: Consider a PID controller with pre-saturation on the output. We wish to implement integrator anti-windup, whereby once some saturation level is reached, the integrator stops integrating. I have seen two different basic policies at this point (lots of variants of these two).
* In one case the integrator accumulator is cleared, i.e. the integrator is reset and held at 0 value until we come off of saturation.
* In an alternate case, the value of the integrator is held constant, but no new input comes in, until we come off of saturation.
I guess my question is, whether there is any consensus on when it is advisable to do one versus the other.
Dear Daniel,
Indeed I share with you the same idea and I think that a form of this consensus must exist. Here are links and attached files to help you elucidate this issue.
-DB Pensions – Abnormal Vol
https://volatilityviewpoint.wordpress.com/category/db-pensions/
Best regards
Article Improving Accuracy and Linearity of Low-Cost Flow Meters
Article An optimal anti-windup strategy for repetitive control systems
Conference Paper Optimization of Static and Dynamic Anti-Windup Compensator u...
Article Control design methods with anti-windup mechanism for distur...
Article A novel anti-windup PID control scheme for greenhouse climate
Article Anti-windup Fractional Order $$\textit{PI}^\lambda -\textit{...
Dear Mohamed-Mourad,
Thank you for all the references. I appreciate you pointing me in the direction of what others have said on the subject.
Danny
In the literature, it is possible to find more than two anti-windup strategies. Particularly, I prefer avoiding the saturation of the integrator and therefore I use the back-calculation technique. The integrator reset can produce jumps in the control signal and the constant holding mechanism has a slow response if the control signal oscillates near the saturation value. In general, both are simple to implement but in my experience they are not given satisfactory results.
Visioli's book (see below) dedicates the whole Chapter 3 to study many anti-windup strategies.
Best regards
Adrian
PS: Some books that include the topic are
A. Visoli, Practical PID Control. Springer Verlag
K. Aströn and T. Hägglund, PID Controllers: Theory, design and tuning. Instrument Society of America.
K. Aströn and T. Hägglund, Advanced PID Control. Isa
M. Johnson and M. Moradi, PID Control, Springer Verlag
I use the latter approach. It worked so well (OK, for the particular plant I was dealing with) and it is so simple/obvious, I wondered why there are so many papers written on this topic.
I think it relates to the conditions that caused things to go into saturation. A large setpoint change would mean that the error history stored in the integrator were not at all related to the new setpoint. In that case, one would do well to zero out the integrator. On the other hand, going into saturation in a regulator mode would imply that the error values stored in the integrator were most likely relevant and so holding the old value constant would make more sense.
- Badges
- Science topic
- Similar topics
- Engineering
- Control Theory
- Control Systems
- Control Systems Engineering