Chiller Staging Point and Part-Load Efficiency

The plant load at which running two chillers uses less power than running one, and whether the staging setpoint in the controller matches it. The crossover exists because chiller efficiency is not monotonic in load. Many centrifugal machines are most efficient somewhere between 40 and 70 percent load rather than at 100, because condenser water is colder at part load and the lift is lower -- so two machines at half load can beat one at full. Below their best point efficiency falls off again, and eventually one machine at a moderate load beats two machines running badly. Screw and scroll machines behave differently, which is why the curve is entered at four part-load points here rather than assumed. The auxiliaries are what people leave out and they frequently move the answer. Starting a second chiller starts its chilled water pump, its condenser water pump and often a tower cell, and those add power before the compressor does anything -- so the second machine has to save more than its own auxiliaries before it pays for itself. On a plant with constant-speed pumps that penalty can push the crossover well above where the compressor curves alone would put it, which is exactly the case a curve-only comparison gets wrong. The crossover is found by searching the load range rather than by solving, because the curve is not monotonic and a solver would find the wrong root. The practical output is the comparison against the setpoint actually programmed: a plant staged at a habitual fraction of a machine that crosses over somewhere else is running the wrong number of machines for a large part of the season, and that is a setpoint change rather than a capital project. This compares one machine against two on an ENTERED kW/ton curve interpolated linearly between the given points, plus a flat auxiliary power per machine started. It does not model condenser water temperature, tower staging and fan power, the effect of variable-speed pumping, minimum run times, demand limiting, or thermal storage; it does not use IPLV, which is a weighted single number useful for comparing machines and not for staging them; and it does not address the flow and bypass constraints that staging also has to satisfy. The chiller manufacturer's part-load data at the actual condenser conditions and the plant's controls engineer govern.

one machine's power = load x kW/ton at its percent load plus one set of auxiliaries; two machines = load x kW/ton at half that percent load plus two sets; the kW/ton curve is interpolated linearly between four entered part-load points, and the crossover is found by SEARCHING the load range because the curve is not monotonic.

The chiller staging comparison as central-plant practice writes it, with the kW/ton curve ENTERED at four part-load points because efficiency is not monotonic in load and centrifugal, screw and scroll machines differ. It does not model condenser water temperature, tower staging and fan power, variable-speed pumping, minimum run times, demand limiting or thermal storage, and it does not use IPLV, which is for comparing machines rather than staging them.

A curve the reader supplies, interpolated and searched.

Estimate. AHJ and licensed professional govern.

Field names used by the API: machine_tons, plant_load_tons, kw_per_ton_100, kw_per_ton_75, kw_per_ton_50, kw_per_ton_30, auxiliary_kw_per_machine, staging_setpoint_pct, one_pct, two_pct, one_machine_kw, two_machine_kw, crossover_tons, setpoint_tons

• The curve is entered machine type and condenser conditions change it entirelythe chiller manufacturer's part-load data
• Auxiliaries are a flat per-machine figure variable-speed pumping and tower staging are not modelledthe plant's controls engineer
• Not IPLV a weighted single number is for comparison, not for stagingAHRI 550/590