Chiller COP and kW/ton: Why Flow Measurement Matters

In most commercial buildings the chiller producing chilled water is the single largest electrical consumer. Yet in many facilities nobody knows how efficiently the machine actually runs: the controller shows leaving water temperature, setpoint and alarms, but not how much cooling is being produced. Efficiency metrics such as COP, EER and kW/ton depend on exactly that invisible quantity. This article explains how the calculations work, which measurements are mandatory, and what can safely be concluded on a site where flow is not measured.

COP, EER and kW/ton: Three Ways to Express Efficiency

COP (Coefficient of Performance) is the ratio of cooling capacity to electrical input power: COP = Q / P, with both terms in the same unit, usually kW. A higher value means a more efficient machine. kW/ton, common in North American catalogs, inverts the ratio and gives electrical power per ton of refrigeration. Since one ton of refrigeration equals 3.517 kW, kW/ton = P / (Q / 3.517) = 3.517 / COP, and here a lower value is better. EER expresses cooling in BTU/h and power in W, and relates to COP through EER = 3.412 · COP.

Calculating Cooling Capacity: Flow, cp and ΔT

All three metrics contain cooling capacity, and capacity is calculated from the water passing through the evaporator: Q = ṁ · cp · ΔT, where ṁ is mass flow, cp is the specific heat of water and ΔT is the difference between return (entering) and supply (leaving) water temperature. For plain water, with volumetric flow in m³/h and ΔT in kelvin, the practical form is Q [kW] ≈ 1.163 · V̇ [m³/h] · ΔT [K]. In glycol loops both density and specific heat change, so the coefficient must be recalculated from the fluid data sheet.

The formula makes one thing clear: ΔT alone is not capacity. The same ΔT at half the flow means half the capacity. In variable primary flow systems, with pumps staging on and off, or with a bypass left open, flow changes constantly. That is why true COP and kW/ton require flow measurement; any calculation that simply assumes design flow is, at best, an estimate.

How Sensor Accuracy Amplifies ΔT Error

ΔT is the difference between two separate sensors, so their uncertainties combine. If the errors are independent, δ(ΔT) = √(δT_in² + δT_out²). The relative uncertainty of capacity is approximately δQ/Q ≈ √((δV̇/V̇)² + (δΔT/ΔT)²). The critical term is ΔT in the denominator: as ΔT shrinks at part load, the same absolute sensor error turns into a much larger relative error. When ΔT halves, the temperature-driven relative error doubles. A measurement chain that looks acceptable at full load can produce meaningless COP values at low night-time load.

• Select entering and leaving sensors as a matched pair and calibrate them together against the same reference; the offset between the two matters more than absolute accuracy.
• Use immersion thermowells with thermal paste; a sensor taped to the pipe surface is influenced by ambient temperature.
• Place sensors where mixing is complete; a reading right next to a bypass or a common header junction is misleading.
• Log power, flow and temperature with the same timestamp and the same averaging interval.

Part Load, IPLV/NPLV and the Effect of Condenser Temperature

Chillers spend most of the year below full load, so a single full-load COP does not represent real operation. The AHRI 550/590 standard summarizes part-load performance as IPLV: efficiencies at four load points (full, three-quarter, half and quarter load) are combined with fixed weights. For metrics where higher is better, such as COP, IPLV = 0.01·A + 0.42·B + 0.45·C + 0.12·D; for metrics where lower is better, such as kW/ton, IPLV = 1 / (0.01/A + 0.42/B + 0.45/C + 0.12/D). NPLV applies the same weights at project-specific design conditions instead of the standard rating conditions. Because the weights represent an average climate and building profile, real seasonal efficiency departs from IPLV whenever your own load distribution is different.

The second major driver is the condenser side. The pressure difference the compressor must overcome, known as lift, grows with the gap between condensing and evaporating temperatures. When condenser entering water or outdoor air gets warmer, when the coil fouls or when fans underperform, condensing temperature rises and more power is drawn for the same cooling. Raising the chilled water setpoint moves the evaporating temperature up and reduces lift. For this reason a COP figure quoted without load and ambient conditions cannot be compared with anything.

What Can Be Said Without a Flowmeter? Proxy Indicators

Many existing plants have no flowmeter on the evaporator loop, and adding one requires pipework. Absolute COP cannot be calculated in that case, but some proxy indicators still reveal trends. The most useful is the kW/ΔT ratio: if the pump runs at constant speed in the same arrangement, flow is roughly constant, capacity becomes proportional to ΔT, and kW/ΔT moves in proportion to electricity per unit of cooling, which is the inverse of COP. The ratio is only meaningful for comparing a machine with its own history under similar outdoor and load conditions; it cannot compare two different chillers or periods with a different pump arrangement. Because the ratio explodes as ΔT approaches zero, low-ΔT and low-power rows must be removed from the analysis.

Checklist for Reliable COP Measurement on Site

1. Install an electromagnetic or ultrasonic flowmeter on the evaporator loop with the straight pipe runs the manufacturer requires.
2. Position a matched, calibrated entering and leaving sensor pair close to the flowmeter.
3. Connect an energy analyzer to the chiller feeder that covers compressors, fans and auxiliary loads.
4. Record condenser entering water or outdoor air temperature on the same time axis.
5. Group results by load point and ambient condition; never decide on a single instantaneous value.

Chiller Data and the Setpoint Reset Engine in Argus EMS

Argus EMS reads chillers such as Rhoss and York YVAA over Modbus: operating status, evaporator entering and leaving water temperatures, setpoint and alarms are monitored. For units whose controller does not provide power, electrical power comes from the energy analyzer matched to that chiller. Because the sites have no flowmeters, Argus EMS does not calculate absolute COP or kW/ton; for efficiency it uses the kW/ΔT proxy to follow the trend of the same machine. On units where condensing temperature is available, the approach of condensing temperature to outdoor air is also tracked as an early sign of heat rejection problems.

The platform's chilled water setpoint reset engine combines the ASHRAE 90.1 outdoor air reset curve with ASHRAE Guideline 36 Trim & Respond logic and makes a decision every fifteen minutes. The target is derived from the outdoor air curve, limited by humidity-driven ceilings and an upper bound, and never falls below the base setpoint; while there is evidence that the machine holds its setpoint comfortably, small trim steps can climb above the curve. Guards come first: if telemetry is stale or frozen, or the unit is in local mode, nothing is written; an open fault alarm, a ΔT collapse or leaving water failing to track the setpoint pulls the setpoint back to the base in a single move; short cycling and manual intervention lock any increase. If the last increase clearly worsened kW/ΔT, the engine steps back. A new unit is first observed in shadow mode, which logs decisions without writing; in active mode commands are sent signed and the written value is verified by reading it back.