Written by SIMA Intelligent Buildings
Most chiller plants are managed as if “efficiency” were a single setting: a chilled-water setpoint, a condenser-water setpoint, or a pump DP setpoint.
That mindset is the root problem.
A chiller plant is not one machine. It is a system of machines, and every major component in that system has an efficiency curve—meaning its efficiency changes depending on where it operates.
- Chilled water pumps: efficiency shifts with flow, head, speed, and where you sit on the pump curve.
- Condenser water pumps: same story—plus they directly affect heat rejection and chiller lift.
- Shell-and-tube heat exchangers (evaporators & condensers): performance is a function of approach temperatures, fouling, tube cleanliness, and flow regime.
- Cooling towers: performance is a function of wet-bulb, airflow, water flow, fill condition, and control strategy.
- Chillers: efficiency changes dramatically with lift, load, and how capacity is controlled and staged.
The key idea for this series is simple:
Plant optimization is not about optimizing one piece of equipment. It is about operating the entire plant at the point where the resulting overall plant efficiency curve is best.
In practice, that “best point” moves all the time—because the drivers that shape each curve are constantly changing: weather, load, control actions, equipment condition, and even sensor quality.
The “Plant Efficiency Curve” concept
If you plot plant kW/ton (or kW/MBH) against operating conditions (load, wet-bulb, flows, setpoints), what you really have is a moving target. Your plant has a “best achievable” region at any given moment—but you only hit it when all the underlying variables are coordinated.
That’s why traditional operations often settle into “good enough” instead of “best achievable.”
The factors that change each equipment’s efficiency curve
To build the argument cleanly across the series, I’m going to number the major factor categories that shape plant performance. Later, we’ll connect them back to how optimization is actually achieved.
- Load profile and part-load behavior Plants rarely operate at design load. Most of the year is part-load, where staging and control decisions dominate performance.
- Weather and heat-rejection limits (wet-bulb) Cooling tower capability—and therefore condenser water temperature—moves with ambient conditions.
- Hydronic conditions: flow, head, and valve behavior Two-way valves, DP control, bypasses, and valve authority can push pumps to inefficient operating regions and create low ΔT.
- Heat exchanger health and approach temperatures Fouling, scaling, tube condition, and flow regime change approach temperatures and force higher lift and higher energy.
- Chiller lift and capacity control method Lift is one of the strongest drivers of chiller kW/ton. Capacity control type (VFD/IGV/etc.) and staging logic can help—or hurt.
- Setpoint strategy and sequencing (coordination across assets) Optimizing a tower setpoint without considering fan kW, pump kW, and chiller lift is how you “win locally” and lose globally.
- Maintenance condition and degradation over time The curve you think you have is not always the curve you actually have. Dirt, wear, and drift shift performance.
- Instrumentation quality: sensors, metering, and data integrity You cannot optimize what you cannot measure reliably. Bad sensors create “fake reality,” and operators end up controlling noise.
- Operational constraints and risk limits Minimum condenser water temps, freeze protection, differential pressure limits, tenant comfort, process loads, and reliability constraints narrow the feasible operating region.
If you’re reading that list and thinking, “No one can continuously track all of that in real time,” you’re exactly right.
Why this naturally points toward AI
Humans can manage a few variables well—especially when the plant is stable.
But chiller plants are not stable systems. They are dynamic, multi-variable, constraint-bound systems where the optimal point shifts continuously.
That’s the core premise we’ll prove through this series:
When optimization depends on monitoring and coordinating dozens of moving factors across multiple efficiency curves, continuous optimization is structurally better suited to AI than to manual human tuning.
Typical End Suction Centrifugal Pump Curves
In the next post, we’ll start with the most misunderstood lever in many plants: pumps—and how system curve can change trough the operation affecting the pump’s operating point efficiency.
If you want, comment with your plant’s approximate size (tons) and whether you run primary/secondary, variable primary, or constant flow — We will tailor some examples as the series progresses.
Written by SIMA Intelligent Buildings
Articles in this Series:
