Energy engineers are forever looking to reduce chiller-plant energy use, typically relying on high-efficiency water-cooled chillers, "pony" chillers, variable-frequency drives (VFDs), premium-efficiency motors, low-approach/high-efficiency cooling towers, sophisticated control strategies, and the like. In most cases, they focus on increasing the efficiency of individual components and optimizing plants based on outdoor wet-bulb temperature, condenser-water temperature, and chilled-water and differential-pressure reset. In almost all cases, they try to minimize the amount of online equipment and stage chillers based on their ability to maintain chilled-water set point (or 95-percent full-load amps). The result typically is a modest improvement in overall chiller-plant efficiency.

Average annual central-plant efficiency can be determined using the following equation:

In this equation, central-plant energy use includes chillers, chilled-water pumps (primary, secondary, tertiary), condenser pumps, and cooling-tower fans. Figure 1 is a breakdown of typical comfort-cooling centrifugal chilled-water plants in Southern California.

Based on the author's experience, about 90 percent of water-cooled centrifugal central plants operate in the 1.0-to-1.2-kw-per-ton "needs improvement" range. Constant-speed equipment, decoupled pumping arrangements (mostly primary-secondary), significant mixing, oversized pumps, difficulty obtaining design chilled-water temperature, low delta-T, hunting, multiple flow and pressure problems, and systems that simply do not work typically are to blame.

To illustrate the detrimental effect constant-speed equipment has on average annual wire-to- water plant efficiency (kilowatts per ton), the following simple example is offered: A Southern California facility requires chilled water 24 hr a day, seven days a week (24/7). A central plant that employs a typical primary-secondary system with two 260-ton centrifugal chillers, two 10-hp constant-speed primary pumps, two 25-hp secondary pumps (with VFDs), two 25-hp constant-speed condenser pumps, and two cooling-tower cells with 10-hp fans is installed. Based on logged data from the energy-management system, the plant provides 960,000 ton-hours of cooling per year. The average annual contribution to energy use and kilowatts per ton by the lead constant-speed condenser pump alone is:

Thus, the lead condenser pump starts the plant at nearly 0.2 kw per ton. (Depending on the application, a properly controlled variable-speed condenser pump uses between 0.065 and 0.084 kw per ton.)

Demand-Based Control

In 2002, the author was introduced to (HPAC Engineering Editorial Advisory Board member) Tom Hartman's work with demand-based control.1 The theory behind demand-based control is the Equal Marginal Performance Principle (EMPP),2 which leads to the understanding that:

  • System optimization is achieved when marginal system output per unit input is the same for all individual system elements.
  • Individual-system temperatures and pressures are not directly related to optimization.
  • Adaptive methods of optimization are not practical in large, complex systems that operate dynamically.
  • Direct power-based control relationships provide a new direction for simpler and more-effective control of systems with all-variable-speed configurations.

Measurement and verification data from several ultraefficient all-variable-speed plants utilizing a licensed application of demand-based control show an average energy-use reduction of 54 percent over the last three years. The results point to a clear shift in the design, operation, control, and performance of central plants, the lessons being:

  • Every device must be variable-speed.
  • All central-plant components contribute to overall system performance. (A common control strategy is to achieve the lowest condenser-pan-water temperature possible, the thought being that the lower the temperature, the higher the performance of a centrifugal chiller. Although this can increase the performance of chillers substantially, it rarely optimizes entire central plants.) Minimizing or maximizing one component rarely results in an optimum configuration.
  • The amount of equipment online should not be minimized. Existing chiller and cooling-tower heat-exchange surface area should be utilized and new sequencing strategies incorporated.
  • Mixing should not be allowed.3 (All decouplers and three-way valves should be eliminated.) Every drop of chilled water should pass through a load before returning to a chiller.
  • Systems should be designed so that as flow approaches zero, head requirements do as well. (This can be accomplished with the valve-orifice-area method of controlling chilled-water-distribution-pump speed,4,5,6 which uses percentage of open valve-orifice area to determine required chilled-water flow rate. An advantage of this intelligent-iterative-control method is that as flow goes to zero, so does required head. If pressure is not allowed to drop in proportion to the square of flow rate, maximum energy savings cannot be achieved. This technique allows pumps to operate at their highest efficiency at all flows.)
  • Ultraefficient all-variable-speed primary-only systems are reliable and can be installed for the same cost as "standard" central plants.
  • Less-than-0.5-kw-per-ton average annual central-plant wire-to-water efficiencies are attainable. (Efficiency does not take into account reductions from water-side economizers.)
  • Demand-based control is much simpler than conventional control.
  • The EMPP works.

Case Study

An example of effective application of all-variable-speed operation to an existing chiller plant is the County of San Diego's North County Regional Center (NCRC), which consisted of:

  • 610,000 sq ft of air-conditioned space (courthouse, offices, and jail).
  • Three 575-ton centrifugal chillers (1998 vintage).
  • Four 20-hp constant-speed primary chilled-water pumps.
  • Individual secondary chilled-water pumps (with VFDs) at each building.
  • Two 850-ton cooling towers (originally with two-speed fan motors).
  • Four 60-hp constant-speed condenser pumps.
  • 24/7 chilled-water load.

The NCRC had a 1,725-ton constant-speed-primary/variable-speed-secondary chilled-water plant (the secondary pumps were installed in each build- ing) that had been in operation for about five years. Although the system had been designed by an experienced firm, sized correctly, and maintained properly, the plant experienced continuously low delta-T, significant supply and return chilled-water mixing, inefficient operation at part-load conditions, and difficulty achieving design chilled-water temperatures at the air handlers. Its average annual efficiency was 1.12 kw per ton.

In November 2003, all three-way valves (14 of the 58 total valves) and decouplers were eliminated, and all of the centrifugal chillers, cooling-tower fans, primary chilled-water pumps, and condenser pumps were retrofitted with VFDs. The demand-based-control sequence was programmed into the existing energy-management system.

With the system retrofitted into a primary/booster pumping arrangement (Figure 2), the speed of operating equipment is controlled by a direct-digital-control system, ensuring optimum efficiency at all times. Simple direct control algorithms coordinate the operation of chillers, pumps, and tower fans based on demand for cooling, which is determined by cooling-coil-valve position. Chilled-water temperature and tower leaving- water temperature float within preset limits (i.e., 40 to 45 F for the chilled water and 60 to 85 F for the condenser) to allow components to operate at their highest efficiency at all times. The demand-based control sequences, which replaced proportional-integral-derivative (PID) control, coordinate the operation of the condenser pumps and tower fans based on chiller power (kilowatts). The three electric chillers are sequenced on- and offline according to the "natural-curve principle," which sets a chiller-kilowatt threshold for sequencing, based on condenser and evaporator temperature.7 The distribution pumps are controlled using the valve-orifice method.4,5,6 Because the distribution pumps are in series with the primary pumps, a power/speed relationship optimizing the operation of the entire distribution system is maintained at all times. The downstream pumps always lead, while the upstream pumps adjust to match.

The retrofit was completed and commissioned in December 2003 at a cost of $423,700. Two years later, measurement and verification data indicate that the entire plant averages less than 0.5 kw per ton, saving the county more than $175,000 a year (Table 1). (Three years of pre-retrofit data were averaged to provide a baseline.) With a $205,447 incentive from the local utility, the simple payback was 1.3 year. Figures 3, 4, 5 and 6 show performance data from the central plant. Again, the kilowatts per ton take into account the energy use of all chillers, condenser pumps, primary chilled-water pumps, booster chilled-water pumps, and cooling-tower fans. The data were collected in 5-min increments using the building's energy-management system.


Conventional methods of improving chiller-plant efficiency tend to focus on increasing the peak efficiency of individual components. Additionally, because the energy performance of constant-speed chillers, pumps, and towers is maximized when components are operated as close to full load as possible, these methods generally involve the sizing and sequencing of plant equipment to fit a variety of load conditions while minimizing the amount of online equipment. This piecemeal approach needs to be changed. Measurement and verification data confirm that the energy efficiency of a chiller plant is improved most effectively by utilizing variable speed and optimizing the efficiency of the entire system in response to the requirements of the load served by the plant.

This improvement is maximized when equipment is operated at part-load conditions. Simple and reliable direct control algorithms that can be used to coordinate the operation of all-variable-speed chillers, pumps, and tower fans are available. Demand-based control through the EMPP is an effective means of operating an ultraefficient all-variable-speed chilled-water plant. Average annual wire-to-water central-plant operating efficiencies below 0.5 kw per ton are attainable.


The author wishes to thank the following individuals for their assistance in preparing this article: Tom Hartman of The Hartman Co., Anthony Roner and Eric Nyenhuis of Siemens Building Technologies Inc., Tom Shaw of Alpha Mechanical, and David Vaughan and Patricia Zeitounian of the County of San Diego.


1) Hartman, T. (2001, December). Ultra-efficient cooling with demand-based control. HPAC Engineering, pp. 29-32, 34, 35.

2) Hartman, T. (2005). Designing efficient systems with the equal marginal performance principle. ASHRAE Journal, 47, 64-70.

3) Hartman, T. (2002, April). All-variable speed chilled water distribution systems: Optimizing distribution efficiency. Available at http://www.automated

4) Hartman, T. (2003, September). Presenting intelligent iterative control: PID replacement for setpoint control (pt. 1). HPAC Engineering, pp. 13, 14.

5) Hartman, T. (2003, October). Presenting intelligent iterative control: PID replacement for setpoint control (pt. 2). HPAC Engineering, pp. 9, 10.

6) Hartman, T. (2003, November). Presenting intelligent iterative control: PID replacement for setpoint control (pt. 3). HPAC Engineering, pp. 9, 10.

7) Hartman, T. (2001). All-variable speed centrifugal chiller plants. ASHRAE Journal, 43, 43-51.

Ben Erpelding, PE, CEM, is engineering manager for the San Diego Regional Energy Office (, a state-funded, non-profit organization promoting energy efficiency. Over the last five years, he has performed more than 300 central-plant energy assessments, detailed measurement and verification analyses, and energy simulations. He can be contacted at