A chiller plant is the single largest energy consumer in most large buildings, and it is where the widest gap opens between the design on paper and the plant that actually runs. The chillers themselves are reliable, well-engineered machines; what goes wrong is everything around them — how they are configured, how they are staged, the water temperatures they are fed, and how they were piped and commissioned. This is a tour of the whole plant: the configurations you can choose, the design pitfalls that quietly waste energy for twenty years, and the installation problems that show up on day one.

1 · The plant is a system, not a chiller

A central chilled-water plant is a loop of loops. Chillers make cold water; primary (evaporator) pumps push it through the chillers; secondary (distribution) pumps send it up the building to the air-handling coils; and on the hot side, condenser pumps and cooling towers reject the heat to atmosphere. Get any one of those wrong and the chillers — however good — cannot deliver. This article is the plant-level companion to the source-to-terminal walk-through in the HVAC module of the MEP course; here we go deep on the plant room itself.

Water-cooled central plant — condenser side (hot) · chillers · chilled-water side (cold) cooling towers CDWP CHILLERS primary decoupler secondary building — AHU / FCU coils blue = cold supply · red = warm return · the decoupler splits the production loop from the distribution loop
Original schematic of a primary–secondary (decoupled) water-cooled plant. Cooling towers and condenser-water pumps (CDWP) reject heat; chillers and constant-speed primary pumps make the cold water; variable-speed secondary pumps distribute it; the short decoupler line hydraulically separates the two.

2 · Chiller types — the first fork

Before configuration comes selection. The dominant machines are vapour-compression chillers, split first by how they reject heat and then by compressor type[1]:

TypeWhere it fitsEfficiency & notes
Air-cooledSmall–medium loads, no water available, simpler O&MLower efficiency (rejects to dry-bulb air); no cooling tower or condenser water
Water-cooledLarge plants, district coolingBest efficiency (rejects to wet-bulb); needs towers, condenser pumps & water treatment
Scroll / screw (positive-displacement)Small–mid capacity, good at part loadRobust, modular; screw common 200–800 TR
CentrifugalLarge capacity (500 TR–several thousand)Highest full-load efficiency; VSD centrifugals excel at part load
AbsorptionWhere waste heat / cheap gas existsThermally driven, low electrical use, lower COP; niche

For a large building or a campus, the answer is almost always multiple water-cooled centrifugal (often VSD) chillers, and the rest of this article assumes that plant. The number and size of those units is the first real design decision — and the subject of the first chart.

3 · Plant configurations — how the loops are arranged

The same chillers can be piped in fundamentally different ways, and the choice sets the plant's efficiency ceiling for its whole life[2][3]:

4 · Sizing & staging — the first design pitfall

The most common plant mistake is made before any pipe is drawn: oversizing. Loads are estimated conservatively, safety factors pile on safety factors, and the plant ends up sized for a peak that occurs a few hours a year — then spends its life crawling at part load, where a badly-staged plant is least efficient. The defences are honest load calculation, sensible diversity, and splitting the plant into the right number of chillers so it can follow the load in efficient steps, with N+1 redundancy for reliability[2][4].

Why unit count matters A plant built from one or two huge chillers has coarse steps: at 20% load a single chiller must run at 20–40% capacity, deep in its inefficient zone. Split the same capacity into four or five units and the plant can run two or three of them near their efficiency sweet spot instead. More, smaller chillers cost a little more to buy and pipe, but they hold high efficiency across the part-load hours where the plant actually lives.

5 · Interactive: staging & part-load efficiency

This is a plant of fixed total capacity, split into N equal chillers, staged so each running unit sits at a sensible part-load ratio. Drag the number of chillers and the operating load, and watch the plant's efficiency (kW/ton — lower is better) across the whole load range. The sawtooth is staging: each dip is a chiller being added at just the right moment.

Plant efficiency (kW/ton) across the load range
A 1,000 TR plant split into N equal VSD-centrifugal chillers. The curve is plant kW/ton vs load; the marker is your operating point. Lower kW/ton is more efficient. More chillers = finer staging = flatter, lower kW/ton at part load.
Total plant capacity fixed at 1,000 TR, split into N equal units.
Current plant load as a share of total capacity.
Chillers running
2 of 4
Each at load
90 %
Plant efficiency
0.50 kW/ton
Input power
225 kW

At 45% load a four-chiller plant runs two units near 90% each — close to the efficiency sweet spot. Drop N to 1 and the same 45% load forces one chiller to lug at 45%, and the low-load hours get worse still. The gain from more units is largest exactly where the plant spends most of its time: the middle and bottom of the load range.

6 · The Low Delta-T Syndrome — the classic plant disease

If a chiller plant underperforms, this is the first suspect. Chillers are selected for a design temperature difference between return and supply chilled water — commonly around 5.5 °C (10 °F). The flow a chiller needs follows directly from load and ΔT[3][5]:

\[ Q = \frac{\dot{q}}{\rho\,c_p\,\Delta T} \quad\Rightarrow\quad Q\,(\text{L/s}) = \frac{\dot{q}\,(\text{kW})}{4.187\times\Delta T\,(^\circ\text{C})} \]

The trouble is that real coils often return water colder than design — fouled coils, wrong control valves, three-way valves, dirty filters, or simply part-load conditions. When the return is colder, the actual ΔT falls, and because flow is inversely proportional to ΔT, the flow required to move the same cooling explodes. Worse, in a primary–secondary plant the swollen secondary flow exceeds the primary flow, pulling warm water backward through the decoupler; chillers see a diluted, cool return, cannot load up, and the plant is forced to start another chiller to make flow it does not need — burning chiller and pump energy to deliver the same tons.

7 · Interactive: the Low Delta-T flow penalty

Set the plant load and its design ΔT, then drag the actual ΔT down to mimic a degraded plant. The curve is flow versus ΔT — a hyperbola that shoots up as ΔT collapses. Read how much extra flow, and how much extra pumping power, the low ΔT costs you for exactly the same cooling delivered.

Chilled-water flow vs. temperature difference
Flow Q = load ÷ (4.187 × ΔT). The blue marker is the design point; the green marker is the degraded (actual) point. As ΔT falls below design, flow — and the pumping power that rides the system curve (∝ flow³) — rise sharply.
Cooling delivered (1 TR = 3.517 kW).
The return-to-supply temperature drop the chillers were selected for.
The ΔT the plant really achieves — drop it to see the penalty.
Design flow
122 L/s
Actual flow
168 L/s
Extra flow
38 %
Pumping penalty

An 800 TR plant at a design 5.5 °C needs about 122 L/s; let the ΔT sag to 4.0 °C and it needs ~168 L/s — 38% more water for the same tons, and roughly 2.6× the pumping power on the affinity law. That is before the decoupler backflow forces an extra chiller online. Protecting design ΔT — good coil selection, two-way control valves, clean coils, correct sensor placement — is the highest-value thing you can do for plant energy.

8 · Pumping — where variable speed pays back

Distribution pumping is the other big energy lever, and it turns on one physical fact: pump power falls with the cube of speed (the affinity laws). A constant-speed pump that is throttled to reduce flow saves almost nothing; a variable-speed pump that actually slows down saves dramatically, because at half flow it draws roughly one-eighth of the power[2][6]. Since a plant spends most of its hours well below peak, matching pump speed to load — the whole point of primary–secondary and variable-primary-flow layouts — is where much of the annual saving lives.

9 · Interactive: constant-speed vs variable-speed pumping

The same distribution pump serving a typical cooling day, run two ways: constant-speed (flat, full power all day) and variable-speed following the load with the cube law. The gap between the curves, summed over the day, is the energy a VFD saves — the core of the variable-flow business case.

Distribution pump power over a cooling day
Constant-speed runs near rated power all day; variable-speed power ≈ rated × (flow/rated)³, with flow tracking the cooling load down to a minimum. The shaded gap is the daily energy saved.
Distribution pump power at design (peak) flow.
Lowest flow the system allows (min chiller flow / minimum circulation).
How hard the day runs — scales the whole load profile.
Constant-speed
1,800 kWh/d
Variable-speed
kWh/d
Energy saved
%
Annual saving
MWh/yr

A 75 kW distribution pump left at constant speed uses ~1,800 kWh a day whatever the load. Put it on a VFD and, because the day averages well below peak, it draws a fraction of that. The curve here is the idealised affinity cube law (no static head); a real loop with some fixed head saves a little less, but the saving is always large. The cube law is why variable flow is not a refinement but the default: the pump you slow down is the cheapest chiller-plant upgrade there is.

10 · Installation & execution problems

A perfect design still fails if the plant is built badly, and chiller plants are unforgiving of site shortcuts. The recurring execution problems fall into a few families[1][4][7]:

10.1 Piping & water

10.2 Cooling towers & condenser water

10.3 Machinery room, electrical & controls

The number one execution failure If you remember one thing from this section: flow verification and commissioning. A plant that was never balanced, never had its control sequence tuned, and never had its ΔT verified will underperform its design forever — and the symptoms (extra chillers running, high pumping energy) are exactly those of the Low Delta-T Syndrome. Most "the chillers are undersized" complaints are really un-commissioned plants.

11 · The design & installation checklist

The one-line summary Chillers rarely fail; plants do. Size for the load you actually have and split it into enough units to stage efficiently, pick the configuration on purpose, defend the design ΔT above all else, let the pumps slow down — and then build and commission it properly, because an un-flushed, un-balanced, un-tuned plant will waste energy for its entire life no matter how good the design looked on paper.

References & standards

  1. ASHRAE. Handbook — HVAC Systems and Equipment (liquid chillers, cooling towers, central plant, hydronic pumping).
  2. Taylor, S.T. Fundamentals of Design and Control of Central Chilled-Water Plants. ASHRAE self-directed / ASHRAE Journal series (staging, variable primary flow, pumping).
  3. Taylor, S.T. (2002). Degrading Chilled Water Plant Delta-T: Causes and Mitigation. ASHRAE Transactions — the Low Delta-T Syndrome.
  4. ASHRAE. Handbook — HVAC Applications (design, commissioning, testing-adjusting-balancing).
  5. Kirsner, W. (1996). The demise of the primary–secondary pumping paradigm for chilled water plant design. HPAC Engineering.
  6. ASHRAE. Standard 90.1 — Energy Standard for Buildings (chiller efficiency, IPLV, part-load); AHRI Standard 550/590 (water-chilling packages, IPLV).
  7. ASHRAE. Standard 15 — Safety Standard for Refrigeration Systems (machinery-room ventilation and refrigerant detection).
  8. CIBSE. Guide B — Heating, Ventilating, Air Conditioning and Refrigeration; and district-cooling design guidance for large-ΔT, series-counterflow plants.
#ChillerPlant #HVAC #Cooling #CentralPlant #ChilledWater #VariablePrimaryFlow #PrimarySecondary #LowDeltaT #DeltaTSyndrome #Staging #PartLoad #IPLV #CoolingTower #CondenserWater #VFD #PumpingEnergy #ASHRAE #DistrictCooling #MEP #Commissioning #BuildingServices