Appendix 5G: Large air-conditioning systems

The space-cooling savings that one might suppose have been most thoroughly mined out, because they're the basis of a sophisticated global industry, are in the conventional (refrigerative) air-conditioning systems that render conventional large buildings more or less habitable. About 10-20% of the total capital cost of a major commercial building is its HVAC (Heating, Cooling, and Air-Conditioning) system—especially the cooling part. The central water-cooled chillers at the heart of almost every big building complex today are massive, precise, expensive machines that must be ordered years ahead. The building is constructed around them. They run for decades. They and their smaller, even less efficient cousins use one-sixth of all electricity in the United States. On hot summer days, air-conditioning uses the full output of about 200 thousand-megawatt (billion-dollar or more) power stations—about 43% of the nation's entire peak electric load. Residents and businesses in the city of Houston in 1982 paid $3.3 billion for cold air—more, The Wall Street Journal remarked2, "than the gross national product of 42 African nations." In East Asia in the early 1990s, air-conditioning was adding about 25,000 to 50,000 megawatts of new peak demand per year.

One might suppose that these air-conditioning systems' design has been pretty well perfected over the past 70-odd years. Well, not exactly. In fact, one of their most basic design assumptions turned out to be fallacious. In the 1980s, Professor Sam Luxton, an iconoclastic mechanical engineer at the University of Adelaide in South Australia, and his colleagues became curious about how air actually gets cooled and dried by being blown through a cooling coil. The traditional cooling coil consists of many closely spaced sheets of copper, penetrated by tubes that weave back and forth carrying chilled water. Textbooks following 1921 laboratory findings by air-conditioning's inventor, Willis Carrier, say that the airflow is turbulent, full of puffs and eddies, and that as the air cools, the water condenses in a thin film. Luxton's group built a special wind tunnel to check. Surprise! Actually the airflow was almost completely smooth (laminar), and the water beaded up into a dense sweat of little droplets covering the cold plates. Mr. Carrier, it seems, had misinterpreted his lab data, and every air-conditioning company in the world, including his distinguished namesake, had been misdesigning its coils ever since.

Armed with actual observations, Luxton realized that it would make more sense to turn the coil around sideways: to make it shallow instead of deep, space the plates widely instead of closely, and above all, move the air through it slowly and the coolant quickly, rather than the other way around. Why? First, because it takes hundreds of times less energy to move water (per unit of heat transferred) than to move air. Second, because slow airflow leaves the little droplets to bulge out from the plates, extending the surface area over which heat transfer can occur. Blow too fast and you smear out the droplets into a film. Blow harder and you blow them away, eliminating their advantage.

This redesign into a "low-face-velocity/high-coolant-velocity" coil sounds simple and obvious enough, but its effect is profound. The pressure drop from forcing air through the redesigned coil drops by 95%. This decreases fan size, capital cost, and noise. Dehumidification per unit of cooling improves by nearly one-third. The coil costs the same, but the smaller fan costs less, so total capital cost goes down. And the more you reduce the airflow—to cool the room on days cooler than the very hottest—the better the coil dehumidifies. This matches cooling needs, maintaining excellent comfort year-round and far outperforming conventional coils. Furthermore, when combined with the very low-friction air-handling system and efficient fan already described, the smaller fan and lower airflow needed for the redesigned system often fall so low that the conventional "silencer" is no longer required: at most, residual noise can be electronically cancelled with inexpensive "antinoise" devices, such as are becoming common in high-end Japanese refrigerators and cars. But removing the silencer still further reduces the pressure drop against which the fan must fight, making it still smaller, cheaper, quieter, and less energy-consuming.

Next step: Make the air filters bigger. Less friction. Smaller fans. Less capital cost. Less energy. Slower airflow. Less noise. Even less need for silencers and antinoise. Less friction. Smaller fans. Less energy. Less noise. Also better filtration. Filters last longer—their lifetime increases as the inverse square of air velocity. Less maintenance. (Same thing in cleanrooms, only more so.)

The new cooling-coil design is only part of a whole family of interlinked innovations marshaled by Eng Lock Lee, the negawatt wizard of Singapore. Together, they let him air-condition big buildings and electronics factories in some of the world's muggiest climates—Singapore has 84% relative humidity year-round, and ranges from hot to broiling—using about one-third the normal amount of air-conditioning energy: 0.61 kilowatts per ton3 of cooling for the entire system, vs. a local norm of about 1.75. Among Lee's main innovations:
  • Whenever exchanging heat in a refrigerative system's successive loops—between room air and chilled water, chilled water and refrigerant, refrigerant and condenser water, condenser water and outside air—use many times the normal amount of metal surface area between the two fluids, so that their temperatures approach within 1–2 F° of each other, rather than the traditional 10 F°. Why? Because copper is a lot cheaper than kilowatt-hours. This is simply another rule-of-thumb that's very far from an economic optimum.
  • Spin the chiller's impeller at exactly the optimal speed, not the arbitrary one set at the factory. This costs nothing extra—the gearbox costs the same no matter what its ratio—but saves about a tenth of the giant chiller's energy.
  • Rigorously wring friction out of all the fan and pump systems. Just as in Jan Schilham's pumping system, with its 92% energy savings, achieve order-of-magnitude savings in all pumping and air-handling. Make the ducts—just like pipes—bigger, smoother, and straighter to reduce the air friction.
  • Make the cooling towers, which dissipate the building's heat to the outdoors, big. Run their fans slowly. Oversize their fill. Again, the strategy is bigger heat-exchange surface, closer approach temperature, and slower airflow. Run these and the indoor fans at variable speed, optimized to current conditions precisely measured in real time.
  • Use premium, right-sized motors, and the other improvements described in Appendix D to run all the devices.
  • Usually split your chiller load into several units of different sizes, so that no matter what the load, you can meet it with a machine, or a combination of machines, running at very nearly optimal efficiency.
  • For the spare chiller, which runs only rarely and briefly while a main chiller is being repaired, buy a second-hand unit, even if it's inefficient. It'll hardly ever run, so the extra cost of a new spare unit is better spent improving the devices that will run more than 99% of the time.
  • All the fanpower ends up heating the air being moved: if it didn't get hotter, it would have to go around faster and faster to reflect the added energy, and it doesn't. All the pumping similarly makes the liquid hotter. Most of that heat ends up inside the building. Therefore, the more efficient all the equipment is, the less heat gets added to that already present from the windows, roof, lights, etc.; so the less the equipment has to be sized to carry away its own waste heat, recycled around and around; so the smaller it can be, the less it will cost, the less energy it will use, and so on in circles.
  • Having made all these improvements, make everything exactly the right size, based on actual measurements in the field—not too big. Now that they're smaller, go back and count the smaller amount of heat they release into the building, then make them correspondingly smaller. Repeat several times until the convergent series of savings dies out. Thus the smaller and more efficient the cooling components get, the smaller and more efficient still they can become. It's just like the way weight savings snowball in a Hypercar (Chapter 2/Reinventing the Wheels).
  • Control everything with hand-calibrated, top-flight sensors and powerful graphical software so the operator is instantly notified if anything needs a tune-up, and can immediately see what to do (Chapter 5/Building Blocks).
The results are clear. Lee's supply fans use not the Singaporean "good-practice" norm of 0.60 kilowatts per ton, but 0.061—90% less. His chilled-water pumps use not 0.16 but 0.018—89% less. His condenser-water pumps use not 0.14 but 0.018—87% less. His cooling towers use not 0.10 but 0.012—88% less. His systems provide better comfort, are more reliable, are easier and faster to build, take up less space, and usually cost less to construct.4 These savings come not from any sorcery but from the rigorous practical applications of accepted engineering principles that most engineers have carelessly stopped applying, because they got into the bad habit of substituting hoary component-based rules-of-thumb for critical thought and whole-system engineering.

1 Cler, G., Shepard, H., Gregerson, J., Houghton, D.J., Fryer, L., Elleson, J., Pattinson, B., Hawthorne, W., Webster, L., Stein, J., Davis, D. & Parsons, S. 1997: Commercial Space Cooling and Air Handling Technology Atlas, E SOURCE, Boulder CO,

2 Burrough, B., "In Houston, The Ubiquitous Air Conditioner Makes Tolerable an Otherwise Muggy Life," Wall Street Journal, p 31, 21 Sept 1983.

3 A refrigerative ton, or 12,000 BTU/hour, or 3.52 thermal kilowatts, is a rate of cooling, named for the coolth provided by melting one short ton (2,000 lb) of ice—the traditional way of buying cooling—in 24 hours.

4 His chillers use about one-third less too—not around 0.60–0.75 but originally around 0.50 and lately in the low 0.40s—but that's the source of only a fifth of his total savings; the other four-fifths are in the normally neglected "auxiliaries" that actually use two-thirds of the system's total energy.

To make comments or report problems with this site, please contact

© All rights reserved. Published by Rocky Mountain Institute.
2317 Snowmass Creek Road  |  Snowmass, CO 81654-9199  |  Ph: 970.927.3851


Powered by Intrcomm Technology's SMC