Every week I become ever more certain that a great collective difficulty we have in the air-conditioning world is one of visualisation. For some time now, I have been rattling on about one of our difficulties being that our business is all about handling heat. We cannot see heat, although we can see many of its results, such as a glowing piece of red-hot metal. Furthermore, the heat we can’t see is being carried in air, which we equally cannot see. Part of the heat is latent heat, or humidity, which again we cannot see until condensates. Over recent years, this emerging appreciation has done much to help me set better direction in the devising of my various technical information materials. I contend that, without a clear understanding of these ‘invisible’ vital components of refrigeration and air conditioning, no person can have better than a half-baked understanding of the theory which dictates the functioning of our chosen Industry.
Along with the ‘invisibility’ of heat, we have another important example of these invisible hurdles. It is one we have carried over from last month and into our discussion today: our visit into the realm of control theory. Over the years, I have encountered all too frequent repeats of the industry mantra, mumbled by a person designing a control loop: ‘I don’t really know the details of what is to be achieved, but beyond a doubt this advanced controller will sort out everything!’ I have seen proportional controllers installed in the leaving water of a chiller, simply because the device was cheaper than a full-blown P + I controller, with the justification being: ‘Its proportional band can be set right down to 0,5K. And, surely, 0,5K variation in leaving chilled water temperature is as good as any person could wish for’. It simply doesn’t work that way. You have to know precisely what it is you intend to achieve, and equally importantly, the absolute finer details of how you plan to achieve the intended outcome. Only then will you have a good chance of success.
Incidentally, the story just related in the previous paragraph is all too often, true. It leads directly to the untimely death of some costly compressors, and draws me to a realisation of how critically important it is that the person preparing the design can generate a detailed understanding of the ‘how?’, and translate that into a well set up control installation. This is often relegated to the back burner, dismissed inadequately with ‘justification’ that if a thermostatic controller wants more cooling, it will ask for more cooling. If it wants less cooling, it will ask for less cooling. This oversimplified approach is perilous.
I find it invaluable to retrieve some of this control technology of a departed era, filed in memories far away in my brain, to explain a degree of ‘how’.
Recently, a group of us started out with a group of four scroll compressors, with these collectively serving a single air handler. As we proceeded, it soon became apparent that we should start at the beginning. We said the plant served by those four fairly significant compressors could be serving a room of several hundred square metres; something which raises issues of its own. We tapered back to the control of a smallish room, served by direct expansion cooling, with this sourced from just two scrolls of modest size. We looked at switching differential of the individual steps, and then studied the overlapping the switch points, rather than programming in a distinct full sequence. The latter often appears ‘logical’ to some at first thought. In part one, published last month, our rationale was exposed to a trifle of more extensive logic which would lead to a better outcome in terms of limiting the temperature swings in the room. At that point, we ran out of space, and made a promise to continue from that departure point at another convenient time.
The control of applications such as chilled water fan-coil units, particularly when the area served by the fan-coil unit was relatively small, was common in years gone by in spaces such as single offices, or a group of offices on the same zone. To give a better idea, all offices would have east, or maybe west exposure. The eastern zone peaks not too long after sunrise, possibly at 10h00. It tapers off as the sun crosses over at noon, and then the west starts intensifying in load. This may peak at 15h00 or later. This indicates two totally different load profiles. They must be handled in a way that permits the matching capacity offered by the plant with these load profile differences. Identification of zones for individual treatment needs to extend way beyond those two just mentioned.
Figure 1 shows the control and water-side elements of such a system. The thermostat in this case is indeed itself the controller. The room had been designed to hold a constant temperature of 22ºCelsius – the temperature input into the program that calculates cooling load for design purposes. Typically the design parameters would allow room, for around 2K operating span. Our example has been shown as a three-way water flow control system. If handled correctly, this assures chiller flow rate will remain constant. Justification favouring three-way control is that, as loadings shift, there are no resultant varying flows in the system as a whole. You won’t have pressure differentials which swing wildly by the hour, and the system will not easily degenerate into being a nightmare. People at the end of any given pipe run won’t be infuriated by having chilled water to their office fan-coil unit dwindle to nothing every time the weather gets too hot.
To demonstrate proportional control, alongside the controller in figure one, you will see a settings breakdown indicating that, at 22,5ºC the three-way control valve will be wide open to providing flow to the cooling coil. This will be the case when the room is experiencing the kilowatts of load anticipated by the cooling load calculating program for a peak summer day. An important thing to note with proportional control is that it really is proportional. The fully opened control valve won’t push the room back to 22ºC. For so long as the load remains at 100 percent of design, the valve will remain fully open to flow. Please grasp this issue, central to the understanding of proportional control. It is so easy to miss the point.
As load reduces over a period of time, this fact would be ‘understood’ by the controller in terms of the room temperature tending to drop slightly. The controller’s response to this will be to trim back on the valve’s opening to the point eventually where half of design flow will now be admitted to the cooling coil. If the design hydraulics are correct, it will be the other half that will pass through the bypass arrangement. As an outcome, flow in the branch feeding the fan-coil unit in question will consistently remain at 100 percent of design, and the water distribution boat will not be rocked at all by load shifts. The proportional controller will bring things about for this situation to take place when the room temperature is sitting right upon 22ºC, given our present settings. Most likely, the commissioning technician would have settled on 22ºC as the set point value, with the proportional band being spread equally about this. This however is a purely arbitrary decision on the part of the commissioning technician.
We extend this rationale to when the room temperature has dropped to 21,5ºC. This is at the bottom of our design proportional band, to which the controller also has been set up. The command signal now being issued to the three-way valve will be for it to be fully open to bypass -- fully closed to flow through the cooling coil.
Sensitivity and proportional band
The proportional band of the system we described was 1K. That should be small enough to please the most fastidious of tenants. But, will it perform suitably? The most stable plant of all is one that is operating at full load and full capacity. The ride will be smooth. Control quality starts to be revealed when load diminishes. As this happens, the control system must compensate for a diminishing load by trimming back on the capacity being made available to the fan-coil unit. Given the cooler weather, the nature of the outcome now could be that the valve reduces chilled water flow rather rapidly and be too warm. The valve will compensate by opening, perhaps too rapidly. There could soon be produced a rush of cold air, resulting in it becoming too cold. In a flash, there develops an up-down temperature swing— the dreaded ‘hunting’.
Figure 2 illustrates a fellow on a pogo stick. This is a parallel which can serve to illustrate a hunting system. Suppose the plant in figure one had electric heating as well. This would be set to bring in the first bank of heating at a little below the temperarture at which the cooling valve had been driven fully closed. Say the first heating had been programmed to come in at 21,0ºC and the final step to come in at 20,5ºC. The very fact that this energy is available on call to act in this opposite direction would hugely increase the potential for hunting. If the electric heating circuit were isolated at times when cooling was required, this downside would be avoided. But the prime objective here has been to illustrate the nature of hunting. The pogo stick, given the available extremes of full heating on one hand and full cooling on the other and therefore being caused to bounce like a wild thing, serves to illustrate this well. Going back to before the ‘heating’ issue was added, to when only cooling was available, with the proportional band being just 1K.
The standard remedy is to broaden the proportional band, so as to add stability to the operation. Increase this to perhaps 2K. You would increase the proportional band perhaps just enough to make the sub-system stable under a range of operating conditions. If you were to fatten the proportional band to perhaps 4K, the system if working on chilled water alone would certainly be as stable as a house, but its operation would be unacceptably sluggish—too cold in winter; too hot in summer. The magic lies in finding the elusive ‘happy medium’.
Modern electronic controls beat those of my day by having electronic temperature sensing devices of extremely small thermal mass, which are far better able to keep pace with the temperature swings, and to therefore transmit accurate and meaningful information to the controller. But we have had a somewhat different reason for travelling this route. That reason has been to equip us for a discussion on chiller control.
Benefit of a proportional band
It is rather easy to say modern electronic controls have other functions, notably the ‘integral’ function, which was comparatively rare in days past, which can effectively limit the proportional band. Why do we even need to talk of a proportional band? The fact of the matter is that, if it is a burning hot summer day, and design conditions are 22ºC ± 2K, it permits us to run the building at 24ºC. For people coming in and going out to the burning hot streets, the 24ºC room temperature may indeed be more comfortable than a lower temperature. More importantly, it requires less energy to air condition a building to 24ºC in hot weather. Furthermore, during the swing from 22ºC to 24ºC, which will take place over a couple of hours as load peaks, the building itself contributes a sizeable chunk of stored, short-term cooling. Think of thermal storage integrally present in the structure. On the upswing, an amount of heat will flow into the bricks, concrete and furnishings instead of having to be accommodated there, and then by the refrigeration capacity. Think proportional control.
Refrigerant liquid feed control
Recently I presented a SAIRAC class with my course, ‘Refrigeration Explained’. The subject was liquid feed control, and resulted in an exercise of comparison between thermostatic expansion valves (TEV) and electronic expansion valves (EEV), along with those confounded restrictor controlled feeds of the Dark Ages. While we concluded that TX valves were the head and shoulders of the fixed restrictor family of liquid feed control devices, in that they contribute a physically varied size opening and make a reasonably legitimate effort to offer ordered control, TX valves are archaic in terms of stable and economical performance when measured against EEVs.
An important contributory issue is illustrated in Figure 3. This compares the sensing bulb of a TX valve with a typical electronic temperature sensor as may be applied with the control loop of a TEV. The objective of any temperature sensing element of such a control loop is to keep very close track of the suction line temperature. For this provides one component of the data for determination of the measure of superheat present at the moment in that line. When I have measured suction line temperature directly at the exit of a coil with a highly responsive instrument, that temperature has danced around with verve of a dragon fly. While the nimble electronic sensor has an excellent chance of keeping abreast of this, that fat and clumsy sensing bulb of a TX valve would be better compared with a dancing elephant!
The outcome of this is that the TX valve becomes bewildered as to what is going on as soon as conditions start to bounce about in a lively way; something it finds all too easy to do. It is being confounded by having established what the suction line temperature was three minutes back and integrating that with an instantaneous evaporating pressure measurement to determine the instantaneous superheat value. Garbage in = garbage out.
On the other hand, the electronic control component of an electronic control valve installation detects temperature changes within a second or two of an evaporating pressure shift having taken place. So, in the case of a TX valve, it has to be set to control at something coarse like 6K of superheat, this militates against achievement of best evaporator capacity, and even then the dice of producing stability are loaded rather badly against it. On the other hand, the sensors of an electronic control valve probably ‘know’ within a second or two the exact value of the prevailing suction superheat. The control loop can define this reliably down to about 0,5K. Furthermore, the full electronic evaluation and response are effectively completed instantaneously. And in a fraction of a second, the inner valve will have been regulated to precisely meter the flow of refrigerant so as to exactly match the practically instantaneous liquid feed requirement. The difference is similar to comparing night with day.
Water chiller control
Having stuck our noses into some diverse issues of control, we will shift to yet another—control of a multi-step water chilling machine. We will assume the four scrolls of our ‘part one’ discussion, instead of cooling air, are instead serving a water chiller, as in figure four. On this, you will see two control options have been entertained. One is to have the control system measure the entering water temperature, and the second is for the arrangement to measure the leaving water temperature. If these ever become confused, be prepared to carry compressor after compressor out to the junk yard.We should have sufficient space today to discuss just the entering water control option. Figure 4 makes it very clear where the sensor for this option must be sited.
The rationale adopted here is that our chiller has been designed for entering water to be at 12ºC with a leaving water temperature of 6ºC at these being at a given flow rate of water. These constitute a somewhat common choice of temperatures for a water chiller on air conditioning duty. The kilowatts of refrigeration capacity incorporated into the machine will have been based on this information.
Where do we commence in setting up our logic regarding entering chilled water control? We refer to the chiller design conditions. Assume as above that the chiller had been ordered to deliver x l/s of water at 6ºC while being fed with an equal flow of return water at 12ºC. This has been spelt out with Figure 5. We are to assume the chilled water ‘flow’ pipework spreads to serve the various fan-coil air conditioners distributed through the building, and the ‘return’ line gathers up water from these fan-coil units, and channels this back to the chiller.
From here, we will make the assumption that water is leaving the chillers at a constant temperature of 6ºC. Therefore, if we were to measure 12ºC at the returning water nozzle, we could look bright and say “This building is operating at full load”. However, if we were to measure just 9ºC, Midway between 12ºC and 6ºC we could make a profound statement: “This building is operating at half load”.
Our next step of rationalisation is to tell ourselves that if the building is operating at full load, our chiller must respond by functioning at full capacity. On the other hand, given the measured return water temperature of 9ºC, our logic would tell us we require operation of just two of our four compressors. We fill in the other gaps by saying that if water is found to be returning at 10,5ºC , that is indication that 75 percent of cooling capacity is required. We would respond by seeing to it that three compressors run. By this same token, 7,5ºC returning water would have us operating just one compressor. Of course, a return water temperature of 6ºC would inform us that load is zero, and therefore none of the compressors would presently be required to operate. When we come back to thinking how this plays out with the four-stage chiller of figure four, we will see it is the temperature sensor on the left about which we are speaking. Please be very clear on this. The sensor on the right in figure 4 cannot be allowed to play any part in today’s discussion. Note very specifically the case we have made today applies expressly to the left side controller; the one that examines and responds to variations in the entering chilled water temperature as a proxy for load.
This provides us with a fundamental target of our objective. But things are still somewhat ropey. You would be completely justified if you were to challenge our earlier statement which assumed leaving water temperature would obligingly sit at a beautifully constant 6ºC. While it served adequately for the initial rationale, which was false, by what logic would you have challenged the statement? Think again of those four compressors collectively providing 6K of cooling to the prescribed water quantity. How much cooling is each providing? The answer is 1,5K to the overall effort.
Another question: When one chiller is stopped, what will happen to the leaving water temperature? When robbed of the 1,5K cooling effect, leaving water temperature will increase by 1,5K. The counter-question: What happens to leaving water temperature when an additional step is brought into operation? It will drop by 1,5K. We must have a periodic wiggle of at least 1,5K taking place in the leaving water. Furthermore, these responses will require a few seconds to come up on the radar of a leaving water sensor.
Think of how we have arranged things. These ‘blips’ taking place periodically in the chilled water temperature as it leaves the chiller will start out on a circuit through the building. It will increase the capacity of each fan-coil unit by a trifling amount. The chilled water circuit possibly comprises various lengths of piping in the loop to each fan-coil unit. So there is a blending effect on top of a flywheel effect. This all contributes to stable control. Stable control wins you first prize.
Figure 6 addresses the remaining gaps. One issue to which we must return is that of switching differential. You will remember this came up in ‘part one’, along with the ‘wise’ staging of the switch points. As we lack the means to animate our illustrations, you will have to provide that bit yourself. Get a ruler, and place it across the temperature scale on the left of figure six. Move it up and down between 12ºC and 6ºC, while watching the switching operations identified at the right. These aspects constitute ‘extras’ we must incorporate into our thinking as we reach this grand level.
Put your ruler across the image at 11ºC. Compressors 2 to 4 will be running. What about compressor one? Will it be running or will it be taking a snooze? To answer that question, you would have to know whether the temperature had risen from somewhere under 10,5ºC to reach the 11ºC mark, or whether it had dropped to there from fractionally above 12ºC. Had the temperature risen to reach 11ºC, the compressor would be standing, for the reason that the temperature must first rise through 12ºC to cause the first contactor to engage. A quick read of this might make it sound like I’m talking nonsense, but it is not so. This is an important area of understanding. Dwell on Figure 6 with your ruler until you have fully grasped these subtle but critical control issues spelt out there.
Short cycling of individual compressors
There is a further flaw in this thinking which we must weed out. What I have described is ideal for a single reciprocating compressor with three stages of capacity control. Cooling controls now are achieved by cycling four independent scroll compressors. I am assuming four discrete refrigeration circuits for this description. Bad news for a scroll is for it to be called on to do a short run of around a minute, followed by a stand period before the next equally short run. At each start, a compressor of this type will pretty quickly throw some of its oil along with the discharging hot gas. It may take several minutes for the displaced oil to finish wending its way back to the compressor per the suction line. If we hit on a sequence of short runs, lubricating oil will be slung out to the evaporator in each duration, but will be denied the opportunity to creep back to the crankcase. After a number of such starts, the affected compressor could find itself in trouble through a lack of oil. Bearing and impeller scoring will be a likely outcome. This of course can lead to dismal operation, often followed all too soon by total failure.
The electromechanical type of controller of Figure 6 would be stuck to provide an answer here. It would have been fine for the four-step reciprocating compressor. What we need is a carefully crafted rotating cycling pattern of the operation. This will be planned to keep a compressor running for a respectable period of time after each start. Rather than have me knot up the words, figure seven illustrates the objective. A green tick represents a fresh start activity, while a grey tick indicates continuing operation of an already running compressor. Similarly, a red X indicates a compressor being freshly stopped, while a grey X signifies a standing compressor remaining idle. Please think carefully through these issues if you have any desire to make yourself an ‘informed’ person on control issues. We will return next month with ‘part three’ of this series as we paddle more and more around the control pond.