Energy management with variable speed drives, Part 3

Scott Sullivan walks through automation with VFDs and discusses the basics of automation, PID and questions from the audience.

By Plant Engineering April 27, 2023
Courtesy: Advanced Energy

Automation insights

  • There are two types of automation: Open-loop and closed-loop and open-loop is less precise than closed-loop.
  • Proportional-integral-derivative (PID) tuning plays a key role in energy efficiency.

Variable frequency drives (VFDs) provide effective speed control of ac motors by manipulating voltage and frequency. Controlling the speed of a motor provides users with improved process control, reduced wear on machines, increased power factor and large energy savings. The most significant energy savings can be achieved in applications with a variable torque load. Reducing a fan speed in a variable torque load application by 20% can achieve energy savings of 50%. For most motion control applications, reducing motor speed is often the easiest way to achieve large energy savings.

Scott Sullivan, a field service engineer at Electronic Drives and Controls or EDC, a certified system CSIA system integrator, gave a presentation on Energy Management with Variable Frequency Drives with CFE. Sullivan specializes in the application of variable frequency drive technology and onsite field service of ac drives. Since joining EDC in 2016, Sullivan has served on EDC’s field service support team performing repairs, preventative maintenance services, startups, training, and much more for ac and dc drives, PLCs and factory automation. Sullivan is a graduate of the University of Rhode Island with a bachelor’s degree in electrical and electronics engineering.

The presentation has been edited for clarity.

Open-and closed-loop control

There are two kinds of automation that you see. One’s called an open-loop control, and one’s called a closed-loop control. Open-loop control is a little bit of a retro name, it’s all a word that only makes sense in context of its counterpart. It’s like you hear someone say snail mail, well when there was just mail, it wasn’t a word. People called it mail, but once e-mail came about they called it snail mail because it was slower. Closed-loop control, you can see it makes a loop between a controller, a drive, a motor, a load, a feedback sensor and it loops back to the controller.

Open-loop control is very common. It’s not particularly precise, but if you don’t really care exactly about how fast a motor’s spinning, it’s usually the cheaper option for a lot of applications.

So if you have an exhaust fan that you just want to turn and you want to save 45%, but you don’t really care if it’s 46%, 47%, that’s probably the way to go. Closed-loop control is when you need something to be precise. If you say you need 45%, it better go 45%.

So I have a dash line here between the load and the feedback sensor, that’s going to indicate that we have a connection, but it’s not directly connected, like when the load is not directly connected. If I was measuring, say the temperature of a cooling tower, the fan is going to be spinning to cool down the water, but I’m measuring it from a temperature sensor, it’s not on the fan. So there’s a direct relation but there may not be a direct connection. The distributed centralized controls, these are big words, but I’ll break them down for you. When you’re trying to control a whole bunch of different things at once, there’s two ways to go about it. Distributed is that each individual motor, load and drive has its own feedback, has its own control, and they basically just control themselves. You tell what you want to do, everything is handled in its own individual capacity.

Most drives will handle this. Some drives even allow communication between a couple different drives. But for more than two or three, I don’t really recommend this, especially if you’re largely controlling the same thing. To use the same cooling tower example, let’s say you have three cooling tower cells and you have a drive on each one spinning the water or spinning the fan to cool the water. If you have three different drives making this calculation and you have three different towers running, they’re each going to try to cool down the water independently and they’re going to be fighting each other at some point.

Troubleshooting three things at once is less preferred. Because of that you may have centralized control. Drive, motor and load will all be connected in series, but there’ll be feedback going to a centralized controller. This controller will individually tell each drive how fast to go and when to turn on and off. This is greatly preferred for a large application. If you have like 20 or 30 fans or pumps in a MER somewhere, having them all control themselves is almost more trouble than it’s worth. Because of this, you’ll often see building automation systems or building management systems, that falls under the umbrella of the controller we’re talking here.

They’ll control every single drive independently and a lot of the smooth operation of your facility. But I got to go to through some terminology here so you understand what we’re talking about when we talk about automation. You may hear people talk about process variables or PVs. Process variable is what you’re monitoring and ultimately hoping to try to regulate. On examples, we got temperature, water level, pressure, tension, flow rate. Temperature would be a cooling tower. Water level could be like a condensate pump, pressure could be, I don’t know, anything based on pressure really, a compressor. Tension is something you only really see in industrial setting. It’s something’s moving through a line and you’re trying to control how much it’s pulled and taut.

PID’s role in energy efficiency

Flow rate could be pumps, could be air through an air handler, whatever you have. The other side of PVs or process variables are control variables (CVs). PVs where you’re monitoring, control variable is actually what you’re controlling. Using our cooling tower example from before, you’re ultimately looking at water temperature, that’s what you want to try to control. But when you’re spinning a motor, you’re not controlling water temperature, you’re controlling a fan speed, which in turn controls water temperature. So that’s what we’re looking at here.

All variables, they really only come in two varieties. Well over 90% of the cases you’re going to be doing motor speed. Motor speed is how fast this motor’s spinning. If you want to do more, you spin it faster, if you want to do less, you spin it slower. Sometimes you don’t care about speed and you just care about motor current. Motor current is how hard the motor’s working, it’s how much you’re pulling on something, how much you’re pushing on something, the amount of work you’re doing.

PID is how we control things almost exclusively. So proportional, integral and derivative: Those are what are commonly referred to as PID. If you use each one of these in series or as needed, they’ll allow you to control pretty much any process you want to automate.

Proportional is how fast you want the system to respond to a given stimulus. If you’re talking about air temperature in a building, this is going to change over a matter of minutes if not hours. So if you want to have a little bit slow of a reaction time, that makes sense. If you’re talking about tension pulling through a line, you need that to go pretty quick. If this tension goes too taut for even a second, you could snap the material. So you need to have this thing pretty much regulated to whatever you’re trying to do.

The problem is if you respond too fast, you may overshoot the setpoint you’re trying to reach. So say you’re warming up a building, I’m sorry, you’re cooling down a building. A building’s 80 degrees and you’re trying to cool it down to 70. If you make this thing respond too fast, it’ll get to 70, keep on going, hit 68, 69, 67 and then start coming back up to 70, overshoot it again, hit 71, 72, come back down, overshoot it 68 to 69, this is the kind of thing we refer to as ringing. If you need a system to respond fast and it ends up ringing like this around a setpoint, what happens is you have to use what’s called integral.

The integral uses the area over the curve to determine how much it overshot and it tracks down to flatten this curve out and get a nice steady setpoint. So in the case of this, there’s still going to be ringing. It’s going to be minimal, but there’s going to be ringing.

If you’re in a situation where you can’t tolerate ringing, then you’re going to have what’s known as the derivative. The derivative slows this down to compensate and get rid of any ringing at all. If you absolutely cannot tolerate this thing going over setpoints, but you want to run right at that tension, I don’t think you need a derivative. This will prevent you from ever going over and snapping the material. The downside is that it does this by slowing down the response time. So anytime you use the derivative, the whole system slows down and it takes longer to respond, but it is worth it.

So when I set these up, I try to do proportional as much as possible. If I can get away with just proportional and have very minimal amount of ringing, that’s it, I’m good. If I have ringing and I really need to respond still fast, I move on to integral and only when absolutely necessary, I’ll use the derivative.

In HVAC applications, you almost never see derivative, it almost never matters. I’m going to go through some examples for you. I talked about tension and I talked about pulling material through a couple of times, so I’m going to give you example of a winder, this is a very, very, very common industrial practice.

The general idea is you’re going to have material that unwinds, you’re going to do some kind of process to it. In this case, we’re going to be using an applicator to apply something to it. Once the applicator’s done, you’re going to be winding it backed up on a thing called the rewind. So as you spin these things and use the applicator, eventually that unwind’s going to shrink, the diameter’s going to decrease and the rewinds going to pick up the slack and start increasing. Now if you think back to math and you have linear speed versus angular speed when the diameter effectively changes the line speed of this thing changes.

So you got two options. I have load cells in this example to look at tension. You can have an operator on this line just sitting there tweaking dials all day, just all day speeding up one, slowing down the other, speeding up one, slowing down the other. And I’ve seen plenty of lines where the automation system has failed and that’s exactly what they have to do, just have a guy sitting there tweaking all day. Virtually you can put a controller on here and have each one separately running a drive. The controller will calculate the diameter of each one based on whatever thickness of the material you have and speed up and slow down the drives, and then turn the motors, and then turn the unwinds and rewinds as fast as you need.

This is an example of closed-loop control we talked about earlier. For instance, on the right, if you look at the unwind, between the unwind, the load cell, the controller, the drive and the motor, it goes a closed-loop back to the unwind and that’s your closed-loop control, same thing on the rewind section. If you do it this way, you can have your own operator load up the unwind and rewind, pull some material through, hit start, and then the machine will just go. Whenever it’s done, comes back, takes the rewind off, loads the new unwind and the process repeats, and it’s much, much more efficient.

So a tank, in an industrial setting, could be something you have maybe on a reservoir. Say you’re constantly applying or say it’s the applicator from previous example, you have a reservoir full of material that you’re trying to apply to align and that’s constantly draining because you’re applying it to material. At some point when this thing gets low, you want to pump it back up full so you can drain it again and the process would repeat.

In a HVAC setting, this could be domestic water. So you have a domestic water tank that you’re trying to pump up to the 14th floor. As the pressure builds, that tank fills up and as the water level decreases on, sorry, the water gets used on the 14th floor, it’s going to decrease and you have to keep pumping that tank up and down so your pressure remains constant. Again, this is just a simple example of an older way of doing things. It’s a tank with two float switches. What’s going to happen is you’re going to kick this thing on, it’s going to pump up the thing until it hits the top flow switch, turn it off, slowly drained unless the bottom one kick on until it hit the top and turn off until you go home for the day. This is fine, it works, nothing wrong with it, but this pump is going full speed. When you’re pumping this thing up, it’s going full speed.

As we discuss with energy savings, when you’re running full speed, you’re not really getting what you can and as far as energy saving’s bank for your buck. A more modern way of doing things, it’s pretty much the same, but instead of these flow switches, you have a level sensor. This tells you exactly how much is in the tank at any given time. That gets fed back to a controller, which gets fed to a VFD, which gets fed to the motor, which in turns pump and pumps the thing up and down. This is one example. Sometimes if you’re just talking about a distributed system, you can eliminate the controller and send it right to the drive. That’s certainly something you can do, as well.

PID used in an empty tank

So in this example, it’s pretty similar to the previous one. You have a almost empty tank, it’s time to kick it on. Little sensor picks it up, starts filling this tank up, go as faster, as low as you want. And if you notice right there, thing picks on, it rings a little bit, right there at the top, that’s going to happen. That’s where the I in PID comes along, there’s always some ringing. So depending on how much you can tolerate, how much ringing you need, usually on a water level, you can tolerate quite a bit. So you don’t really need to worry about derivative, but if you, like Scott, I’m filling this tank to the top, and if it goes over even a little bit, it’s going to start spilling out.

At that point I’ll be like, “Well, let’s start talking about the derivative.” Maybe it’ll be a little slower, but at least this is what will happen. So anyway, your PID does its thing, it slowly brings the water up to whatever level you need it at. Say it’s a 30-foot tank and you only want to fill up 20 feet, this would be 20 feet. So at some point this water’s going to start draining again. Well, I say water, it could be any material. If it’s the applicator, it could be paint, you’re trying to put on some kind of material so you can send it out. But at some point this is going to drain. When it drains, the sensor’s going to find out that it’s draining and kick this back on.

And when it’s filled up, ring again, then settle out and just repeat over and over again. Fill up, ring and then go to sleep. Eventually it’ll drain again, fill up, ring around and then go to sleep again. Because of this, you’re not really letting this thing drain all the way. You’re letting it drain maybe 10%, 20% tops. Because of this, you only got to run the pump 10%, 20% tops. This is as far as energy savings go, an amazing, amazing example. So I talked about cooling towers enough, might as well have an example of cooling towers. So this is an example of our graphics department going hog wild, but I love it. This is to say, two cell cooling tower example.

Let’s say it’s the middle of summer, you run this thing, your building’s at a good temperature and you’re running on one tower, you don’t really need anything else. Middle of the day heats comes, temperature starts heating up, you got to meet your setpoint. You would try to keep the building at 68 degrees, 70 degrees, 72, whatever, and one tower’s not cutting anymore. Your only choice on the older way of doing things is to kick on your second tower. You can kick it on, rebound, and kick it off again, and then kick it on, kick off again, if that sounds a lot like running the thing full speed and not getting energy efficiency, it’s because of it, it’s the exact thing we were talking about.

When you’re running a motor full speed, you are not saving any energy. You are running this thing and you’re wasting energy constantly. So a better example, putting things on at a drive. You can keep running with one tower, full speed, 60 Hz here in the United States, and then the second one you run with whatever you need. You only need to run this thing maybe a third speed, that’d be 20 Hz. So you keep running this, the one tower full, then you run the other one at whatever speed you need to maintain this setpoint. This is a significantly better way of doing things from an energy standpoint, but not still the best way because you’re still running one motor 100%, and when we run 100%, you’re not saving any energy. Two examples, instead of having one going full speed and one going slow speed, you just slow down both, whatever setpoint you need to go, you go. Both of these things will ramp up, both won’t ramp down and you can maintain your setpoint that way.

If at some point you reach six, seven o’clock at night and you don’t really have the demand for the heat anymore, you can shut one off and then just modulate the speed of the one tower. This is a very, very common example. Almost every cooling tower I see has drives on it. Honestly, it’s been maybe a couple years since I’ve seen cooling towers that still have old motor stator systems on it. So this is something that you’ll see almost all the time if you’re still on cooling towers if you have cooling towers in your facility.

Webcast questions, answers

Kevin Parker: Let’s start with, a dozen in ac drive, change frequency and voltage to change the speed, therefore it keeps the voltage to frequency ratio the same?

Scott Sullivan: Well, what you’re talking about is called volts per hertz. What’s going to happen is that at any given time, the speed of a motor is dependent on the frequency, but how much power you put behind this is dependent on the voltage. So what a drive uses is called volts per hertz. At a certain hertz setting, it’ll put out a certain amount of volts and at a certain higher setting it’ll put out a certain higher amount of volts. That way both your speed and torque are regulated. I didn’t really talk about that, but if you remember going back to when I talked about the carrier frequency, how much we gate the transistors on and off, that’s how we regulate the voltage. If we gate these things on 50% of the time, you get 50% of voltage, you gate it on 25% of the time, you get 25% of voltage. So that’s really how it’s done.

That’s the most common example, that’s called volts per hertz. Another way to do it is you can have what’s known as vector control. If you have vector control, you have some kind of speed feedback coming from the motor to the controller or the drive. In that type of situation, you’re going to be telling it to go 25% and that speed feedback’s going to tell you exactly how fast you’re going. If you’re only going 23%, it’s going to increase the voltage to try to make up the difference and get this thing turning 25%. If you’re trying to go 25% and you’re actually going like 28%, it’s going to decrease the voltage to try to keep you down to 25%. It’s a good question, I didn’t bring that up, but that is really heading back to the carrier frequency that we talked about earlier.

Parker: Drive terminals are often marked U, V and W. Where did this naming convention come from?

Sullivan: I’ve been asking myself that for six years. The best thing I got is that very commonly, the input terminals are marked R, S, and T, so alphabetically, the next three are U, V and W. That begs the question, why is the input marked R, S and T? Don’t know. A very good question. I could probably look into a little more and get back to you.

Parker: Could you please comment on 6 pulse drive versus 12 pulse drive and line and load reactors?

Sullivan: Okay, so the example I showed you earlier when we’re talking about the converter section, I had three diodes for the positive half and three diodes for the negative half. That’s a total of six, that’s a six-pulse input. When we do that, diodes tend to reject anything that doesn’t match its current flow. When you start doing that, you reflect that energy back onto the line. That’s what’s known as harmonics and it starts to become a problem. If you can’t tolerate any more harmonics in the building, what you can do is double the amount of diodes. Instead of having three per phase, you have six, and then that ends up being 12 and you count for positive and negative. You can go even higher and have an 18-pulse input. The more pulses you have, the more harmonics get kicked into a higher end of the spectrum and you end up having better harmonic response.

As far as line and load reactors, that’s going to be harmonics on the line side. If you have particularly dirty power in your facility and you start burning things up all the time, you’re probably going to put a line reactor on the drive. We didn’t talk about harmonics today, but that’s a good way to get good clean power out into the drive.

An output reactor typically is done when you worry about having a standing wave in power cables between the motor and the drive. A standing wave occurs is when the cable’s so long that it takes enough time when you get out there, anything that gets reflected back starts adding up to the incoming signal and starts causing peaks and troughs.

The peaks and troughs are power spikes and power dips. You really can’t have that. So when that happens, we put an output reactor on it to try to minimize that. The output reactor can usually be eliminated by keeping the run of wire short, like 50 to 100 feet. It’s going to depend on the manufacturer, but that’s usually what we try to do to eliminate need for an output reactor.

Parker: For PID control, do you have separate devices for each function or are they all incorporated in one device? Sounds like we may not be using derivative very often.

Sullivan: PID is called PID because they’re all packaged together. Any PID controller is capable of handling P, I and D even if you don’t need it, but they are all one controller. And if you don’t believe in using derivative, you’re properly right. You don’t need it. It slows down the system and if you don’t need it, you shouldn’t really be using it.

Parker: Is the drive controlling only air handling unit motor or is it correlated with the compressor?

Sullivan: Anything that spins a motor or anything that has a spinning motor on, you can stick a drive on it. And the example of an air handler, yeah, you can definitely, there’s definitely air compressor drives, but I do see compressors with motor starters a lot. One of the advantages of drives is they allow you to spin faster than 60 yards. Drives can chop up that frequency and put it out at any speed you want. You want to go 200 Hz, you can go 200 Hz. So sometimes I see compressors that have drives on them to spin overspeed, but since we were talking about energy savings and the primary energy savings is running slower speed, running overspeed really tanks for energy savings. So it can be done, I do see it, but more often than not the compressors are on starters and the air handlers will just be controlled by a drive and motor.

Parker: Harmonics was not part of your presentation, but you did talk about it just a minute ago, but we’ve got several questions about harmonics. So can you take a step back and just talk a little bit about the harmonics associated with VFDs in general?

Sullivan: Harmonics is a very complicated process, very complicated explanation. If I think back to school, we didn’t even start talking about it until my third year. Basically, anytime you start chopping up a signal, you’re going to create distortion. And if you think back to thermodynamics and energy, you can’t get rid of energy, it’s got to go somewhere.

So if the input of a drive is rejecting part of a wave, the only place for it to go is back on the line. Because of that, you really can’t have it. It’s going to cause power dips, it’s going to cause power spikes, it’s going to cause a problem. This is not limited to just drives, any non-linear load will cause this.