Artificial intelligence (AI) policy: ASHRAE prohibits the entry of content from any ASHRAE publication or related ASHRAE intellectual property (IP) into any AI tool, including but not limited to ChatGPT. Additionally, creating derivative works of ASHRAE IP using AI is also prohibited without express written permission from ASHRAE.

logoShaping Tomorrow's Built Environment Today

Engineers’ Dialogue: Higher Efficiency Chilled Water Systems Without VFDs

By Jeffrey Joe Newcomb, P.E., ASHRAE Member

“I’ve been thinking, maybe all this rush to put everything on a VFD or ECM  is a little overboard,” mused Old Sam.

”I’m not sure where you are headed with this,” replied Fred.

Fred was visiting Old Sam at the Little Egypt Retirement Village and they had both had a warm piece of pecan pie a la mode while looking out the window at the river.

Sam said, “There can always be too much of a good thing.”

Fred eyed the very few pie crumbs left and said “Always?  How about pecan pie?”

“Don’t be impertinent, youngster!” Sam scolded. “ I’m talking about chillers and pumps and the like. You know, VFDs do take some power to operate and they get less efficient below their design load. A 60 Hp VFD is only 95% efficient at 25% load and it drops substantially after that. The fine art of Engineering is choosing the best alternative.”

“Ok,” Fred said. “What is ‘best’?”

“What I am talking about is the system producing the most cooling for the least energy at the same price point.” Sam went on—“Last I checked, VFDs and ECMs were not free. I read an article not long ago that said a constant-speed chiller that cost the same as a chiller with a VFD built in was 13% more efficient at full load than the VFD chiller.1 

“Even as the load dropped, the constant-speed chiller did better than the VFD equipped chiller as long as the condenser water stayed the same temperature.”

“Ok,”  Fred said, “but the condensing temperature does vary, so shouldn’t a plant have one or two VFD chillers in a VPF arrangement?”

Sam smiled, “Like a lot of things, it depends. I’m sure you have seen the layouts with two VFD chillers in series to extend the turndown of a VPF system and also side-step the sudden load change issues that come with starting a second chiller in a VPF system. The problem is  you have to start at a really high pressure drop in the evaporators to get some reasonable turndown. They take chillers that could operate at 5 ft of head and make them operate at 20 ft of head, just so they reduce the flow down to 20%.”

Sam went on, “You know, there is a way of taking that 5 ft head chiller and using it to feed a load that is 10% or even 5% of the chiller capacity.”

“Now Sam,” Fred said. “You can’t drop the flow that far or laminar flow will cause real problems with a chiller — no manufacturer will provide a warranty operating there.”

“I never said the chiller would operate at 5 or 10% flow," Sam said. “The chiller will always operate at 100% flow.”

Sam took a sip of his coffee, enjoying Fred’s baffled look—“Not all that long ago, primary-secondary pumping was a real breakthrough. The reason it was a breakthrough is that we could decouple the flow of chilled or heating water on the production side from the flow on the consumption side. A constant flow could be maintained through a chiller while the flow varied in the rest of the system.  Non-condensing boilers could be maintained at a high temperature while the heating water temperature to the loads was reduced with a mixing valve to give better control and longer control valve life. Pumped flow decoupling was a breakthrough. Just imagine if we could decouple the chilled water consumption from the production of chilled water.”

“You mean like the chilled water storage tanks used to shift peak chilled water production into the night hours?” Fred asked. “Those are pretty expensive and not very useful in areas that do not have a big differential between on-peak and off-peak energy rates.”

“There are other reasons for chilled water storage tanks.” Sam reminded Fred. “Buffer tanks also store chilled water and are commonly used as a big flywheel for small chilled water systems. Piped properly, buffer tanks can supply chilled water for a short period of time when the chiller stops producing chilled water. Maybe I should call what I am thinking of a balance tank because it balances the difference between chilled water production and consumption. I have been looking at this for a while and figured out I would only need a 3,000 gallon balance tank for a 3,000 ton system.”

“How is that even possible?” Fred asked.

“Let’s take a step back first,” Sam said. “Chillers used to take a long time to start.  Now you can get chillers with a quick start so they can produce chilled water in 5 minutes.  That really is pretty incredible. We can use a small balance tank to decouple production and consumption in a chilled water system.

“Here’s what I mean,” Sam said and started sketching.

Old Sam's Sketch. 

“I can put the balance tank between the chilled water supply and return pipes in a location between the chiller and the load. I also put a shutoff valve on my chiller that closes whenever the chiller is off. The water in the pipes to the balance tank reverses flow direction depending on whether the tank is being charged or discharged. It is an ASME tank, so does not need any pumps to get water in and out of the tank.”

“In normal operation, the tank is  charged by the chiller. When the tank is fully charged, the chiller shuts off and the tank feeds the load. You know a very similar flow diagram appeared in an ASHRAE Journal a long time ago.”2

“Let’s put some specifics to this to take a better look at it and then we can scale it up. Say my chiller is 200 tons at 16 deg delta T.  That is about 300 gpm at 1.5 gpm/ton. Five minutes storage for this system is 1500 gallons. Let’s be safe and put in 10 minutes storage and go to 3000 gallons. That is a stock buffer tank size from Wessels.

“Here’s how this system would operate. Assume the load is only 20 tons, about 10% of our 200 ton chiller, which is 30 gpm. When started, the 200 ton chiller will feed the load 30 gpm and the balance tank the other 270 gpm. It will take a little over 11 minutes to fill the balance tank and then the chiller shuts down. The balance tank will then feed the load for about 100 minutes before the chiller needs to start again.  Almost a 2 hour cycle.

"At 40% load (120 gpm), it takes 16.7 minutes to fill the tank and 25 minutes to empty the tank. Down to a 41.7 minute cycle time. At 50% load, it takes 20 minutes to fill the balance tank and another 20 minutes to empty it. That 40 minute cycle time is as low as it gets. As the load rises, it takes longer to fill the tank and the overall cycle time goes up. The worst case is 1.5 starts per hour. Not too bad for a chiller.”

“Looks like you used up your 3,000 gallons of storage on a 200 ton chiller. You don’t have any left for your 3,000 ton system., Fred observed.

“If I select my chillers right, I don’t need any more storage,” Sam countered. “When did it become a law that all the chillers in a plant had to be the same size? Loads vary, and not by even increments. Why do the chillers have to be sized in even increments?”

Fred was surprised—“Common increment chiller sizes increase system reliability. The pump for one chiller could serve another, if they are the same size. But you cannot do that if the chillers are different sizes.”

“Ah, redundancy is a good thing, but sizing to the expected loads is also a good thing and they are not mutually exclusive in a large plant and there are different ways to have redundancy. I would choose two 200 ton chillers, two 400 ton chillers, and two 1,000 ton chillers—all of which would be constant-speed chillers with constant speed pumps to save operating cost on the chillers and the pumps.”

Fred observed, “That is more than 3,000 tons.”

“I use as much redundancy as is needed," Sam stated. “Let me describe how this system will work throughout the load range. I start out with one 200 ton chiller running till the load rises above 200 tons, then start the second 200 ton chiller and one runs continuously while the other one cycles. Above 400 tons, we start a 400 ton chiller, shut down one 200 ton chiller and cycle the other 200 ton chiller. Above 600 tons, we can either start the second 200 ton chiller or run the two 400 ton chillers with one cycling. From there, we have lots of options that operate chillers to have between 0 and 200 tons more capacity than the actual load.”

Fred said, “Do you measure the flow and direction through the balance tank to figure out which chillers to run?”

“BIngo!” Sam crowed, “You do understand this.”

“Well, I do notice that this system requires going back to a primary-secondary pumping arrangement.” Fred said.

“Of course,” Sam agreed. “Trying to move water through both the production and consumption sides of the system with one pump would require that pump to run all the time. Like chillers, pumps save the most energy when they are off.”

“So this is what you meant when you said you would decouple the production of chilled water from the consumption of chilled water.” Fred observed,  “You actually turn a chiller and its pump off after refilling the balance tank and let the tank take the load for a while.”

“Yes,” Sam said. “It gets better. You can feed a whole campus this way by putting the secondary pumps in the campus buildings and pumping the chillers into the mains. That really makes the chilled water mains between buildings part of an extended primary loop. The chillers do not have to all be in the same building and the balance tank or tanks can be located remote to the chillers, too. Connections to buildings are really simple, too. No fancy bridges that can end up causing low delta T instead of solving it.”

“You’ve got this all planned out,” Fred said.

“Ha!” Sam exclaimed. “I can’t take credit for this. Some ideas have been around for a long time.  You should study old articles by a man named Gil Avery.  Way back in 2001, he wrote an article with a diagram very close to what I am describing for a campus system.2 I don’t agree with everything he wrote, but he was really quite good at hydronic stuff.”

“I am guessing you have to keep the pressure drop in the extended primary loop small. How do you do that?” Fred asked.

Sam yawned.  “It's almost bedtime. I will have to enlighten you about that another time. But I will give you a hint—civil engineers solved that problem a long time ago! See you next time.”


  1. Sullivan, S. 2014. “Comparing Constant-Speed And Variable-Speed Centrifugal Chillers” ASHRAE Journal 56(12):64-66.
  2. Avery, G. 2001. Figure 7. “Improving the Efficiency of Chilled Water Plants” ASHRAE Journal 43(5):14-18.