Honest reflections on control strategies that sometimes help and sometimes hinder performance.

Geppetto wished that he had a bit more control of his unpredictable creation.

When I was an undergraduate I found, to my surprise, a one-to-one correlation between the subjects I enjoyed most and the ones I found most difficult. Chief among the challenges was the concept of dynamics. I was fascinated by the way that this, at least for me, was not intuitive and how instinctive attempts to improve performance of a control system could easily make it worse.

For example, if a system seems to be slow to respond to changes and is constantly “hunting” it might feel appropriate to speed up its response to change, but this will just increase the overshoot. The correct adjustment would be to slow the reactions and allow it to approach the desired value gradually. In one case I was faced with a large variable air volume air-conditioning system with 12 steps of capacity control and was permanently ramping up to full load and back down before switching off and restarting. Stable operation was achieved by increasing the time delay between steps of loading and unloading from 45 seconds to 75 seconds.

Refrigeration also has its fair share of these non-intuitive systems. A float valve on a liquid refrigerant storage vessel will open as the liquid level falls, but one of the consequences of this action is to raise the pressure in the vessel slightly. If the liquid is frothy because of gas bubbles mixed in, then the immediate response to the rise in pressure is for the liquid level to drop, increasing the opening of the make-up valve. A thermostatic expansion valve where the degree of opening is adjusted by the temperature of the gas at the evaporator outlet relative to its pressure is also difficult to model due to inherent time delays built into the mechanism.

Over many years the control systems in refrigeration have become more and more complex, with motorized electronic expansion valves, variable speed fans and many options on compressor capacity control becoming common. Now that we have a great array of control options, often built into sophisticated electronics including auto-tuning of proportional, integral and derivative control, it’s easy to lose sight of the odd behaviors of cooling systems. This is unfortunate because the current global shortage of electronic components means that many systems are having to be operated “by hand” as a short-term expedient while awaiting parts for new builds or service replacements.

Proportional control varies the size of corrective steps taken depending on how far from the setpoint the measured signal sits. Integral control helps to home in more accurately on the setpoint by noting how long the measured signal has been away from the setpoint and nudging it in the right direction. Derivative control adds a further adjustment by considering how quickly the measured signal is changing and making an adjustment to the size of the corrective step accordingly. On paper this can be a fantastic benefit in reducing overshoot, but it is difficult to do well in practice, particularly in refrigeration plants when the system characteristics are changing all the time. I remember a wise old hand telling me “Don’t you worry about derivative control—it’s not for the likes of you.”