Total Plant kW Optimization: Finding the Sweet Spot Between Chiller, Pump, and CT Fan Power

Every chiller plant has at least five power-consuming subsystems. Each has its own optimisation lever: VFD speed for pumps and fans, setpoint reset for chilled and condenser water temperatures, staging logic for chillers. Individually, each lever delivers savings. The problem is that optimising any one in isolation can worsen total plant performance, because the subsystems are coupled through shared thermodynamic variables.

The coupling variable is chiller lift: the difference between condenser water supply temperature and chilled water supply temperature. Lift determines chiller COP. Anything that changes CWS or CHWS, on either side of the plant, changes lift, which changes chiller power. Pump speed, CT fan speed, and setpoint reset all flow through this single variable.

This is why subsystem-by-subsystem savings estimates require a second-order correction: the interactions between subsystems. It is why a plant operating at excellent chiller kW/TR can still show poor IKW/TR, and why low delta-T at the coil level creates penalties that propagate beyond the CHW pump term alone.

This post introduces the concept that ties these together: the sweet-spot curve. Total plant kW, plotted against pump speed and setpoint combinations, has a minimum, though the curve in practice is not perfectly smooth. Chiller staging transitions create step changes, VFD minimum frequency limits create exclusion zones below 30–40% speed, and centrifugal compressor surge limits constrain the accessible lift range. The sweet spot exists within these constraints. It cannot be found by optimising any single subsystem alone, because moving one variable shifts the curve that every other variable sits on.

Why subsystem optimisation fails: the chiller-pump-CT power tradeoff

The KW_Total equation decomposes plant power into five terms. What it does not show is that these terms are not independent. They are coupled through three causal chains, each of which creates a tradeoff between one subsystem's savings and another's penalty.

Chain 1: CW pump speed, condenser water temperature, and chiller lift

Reducing CW pump speed saves pump energy via the affinity law (P ∝ N³). But lower CW flow means the same heat rejection load is carried by less water, so the condenser water temperature rise increases. CWS rises. Higher CWS means higher lift. Higher lift means higher chiller kW.

To isolate this tradeoff, hold CT fan speed constant and vary only CW pump speed. Using the condenser-circuit chart data from a representative plant: at 100% CW pump speed, pump power is approximately 30 kW and chiller power is approximately 155 kW (total: 185 kW). The absolute values are plant-specific; the curve shape and the existence of the minimum are not.

At 65% pump speed, pump power drops to approximately 8 kW (a saving of 22 kW), but CWS has risen by roughly 4°C for this plant, and chiller power has increased to approximately 180 kW (a penalty of 25 kW). The total is now 188 kW. The plant is worse off despite a 73% reduction in pump power. The 22 kW of pump savings has been more than consumed by 25 kW of chiller penalty.

At a moderate 75% pump speed, the numbers look different: pump power drops to approximately 13 kW (saving 17 kW), CWS rises by roughly 2°C, and chiller power increases to approximately 165 kW (penalty of 10 kW). Total: 178 kW. Net savings: 7 kW. This is closer to the sweet spot, where the pump savings still exceed the chiller penalty.

The quantified relationship: each 1°C increase in condenser water supply temperature raises chiller compressor power by approximately 2.5–3% for centrifugal chillers at part load. Screw chillers show lower lift sensitivity (typically 1.5–2.5% per degree) due to their positive-displacement compression mechanism, but the direction of the tradeoff is the same. At a chiller operating power of 155–165 kW, that is 4–5 kW per degree for a centrifugal. The pump engineer who does not account for this will overshoot the optimum.

Chain 2: CHWS setpoint reset, compressor lift, and secondary pump flow

Resetting CHWS upward (say from 6.5°C to 8.5°C) reduces chiller lift by 2°C. Each degree of CHWS reset reduces compressor power by 2.5–3% for centrifugal chillers (somewhat less for screw types), yielding 5–6% total at 2°C reset. On a centrifugal drawing 165 kW, that is approximately 8–10 kW of savings.

For centrifugal chillers, the accessible reset range is bounded on the low-lift side by the compressor surge line; below a certain lift at a given load, the chiller's internal controls will limit further reduction regardless of the BMS setpoint. The practical CHWS reset range for centrifugals is typically 1.5–3°C. Screw chillers do not have a surge limit and can accept wider reset ranges.

But higher CHWS reduces the coil's heat transfer capacity. The log mean temperature difference between air and water narrows. In a 2-way valve plant, valve opening increases modestly. In a 3-way valve plant, bypass flow mixes warm return water into the supply header, collapsing delta-T more aggressively.

Either way, the secondary pump must deliver more flow to maintain the cooling load. The ΔT degradation model shows that the actual coil ΔT decreases as CHWS resets upward, and the flow ratio (design ΔT divided by actual ΔT) determines how much faster the secondary pump must run. The pump kW penalty follows the cube of the speed increase.

A further constraint in hot-humid Indian climates: raising CHWS reduces the coil's dehumidification capacity. The reset range must account for latent load, not just sensible cooling, a constraint explored in detail in the subsequent post on CHWS setpoint control logic.

For a plant with 3-way valves, the CHWS reset range is severely limited: raising CHWS causes uncontrolled flow increase, ΔT collapse, and a secondary pump kW increase that can partially or fully offset the chiller savings. With 2-way modulating valves, the reset range is wider and the penalty is more manageable. Valve infrastructure determines how much of the sweet-spot space is accessible.

Chain 3: CT fan speed, approach temperature, and CWS

Reducing CT fan speed saves fan energy (P ∝ N³). But lower fan speed means less airflow across the cooling tower fill, which means the tower achieves a wider approach to wet-bulb temperature. CWS rises. Higher CWS means higher chiller lift, higher chiller kW.

The dynamic CWS setpoint formula is: CWS = WBT + Target Approach. The approach target for Indian conditions is typically 2.5–3.5°C. At full fan speed, the tower can achieve tight approach (2.5°C). At reduced fan speed, approach widens (4–5°C or more). The chiller penalty from a 2°C approach widening is approximately 5–6% of compressor power.

The optimisation objective for the condenser circuit is: minimize KW_Chiller + KW_CW_Pump + KW_CT_Fan. Not any one of these.

The sentence that frames the entire post: Each of these three chains describes a correct optimisation of a single subsystem. Each, pursued in isolation, worsens total plant kW.

The sweet-spot curve: how total plant kW varies with setpoint and speed combinations

Plot total plant kW on the y-axis and pump speed on the x-axis, while sweeping CHWS or CWS setpoint as a parameter. The resulting family of curves has a distinctive shape.

As pump speed decreases from 100% toward 60%, total kW initially falls: pump savings dominate because the affinity law delivers large power reductions for modest speed decreases (at 80% speed, pump power is already down to 51% of rated).

But as speed drops further, the chiller penalty accelerates because each additional degree of CWS or ΔT degradation costs 2.5–3% of chiller power, and the chiller is the largest single load in the plant (55–65% of total). The total curve bends upward. The minimum of the total curve, the sweet spot, is the operating point where marginal pump savings exactly equal marginal chiller penalty.

Now add the second dimension. At +0°C CHWS reset (design setpoint), the sweet spot occurs at one pump speed. At +2°C reset, chiller kW is lower (reduced lift) but ΔT is degraded, so the pump penalty starts earlier and the sweet spot shifts.

At +4°C reset, the chiller savings are larger, but in a 3-way valve plant the pump kW increase may fully offset the benefit. The sweet-spot curve shifts with every setpoint change.

The same multi-dimensional structure exists on the condenser side. At a given CWS approach target, there is a CW pump speed that minimises the sum of chiller, CW pump, and CT fan power. Change the approach target and the curve shifts. Change the wet-bulb temperature (which changes hour by hour and season by season) and it shifts again.

The actual sweet spot is not a point on a two-dimensional curve. It is a point in a multi-dimensional space defined by: CHWS setpoint, CWS approach target, secondary pump speed, CW pump speed, CT fan speed, and AHU fan speed.

The minimum total kW exists at a specific combination of all six variables for a given cooling load at a given ambient condition. Change the load or the ambient, and the minimum moves.

Identifying the sweet spot for a specific plant requires a parametric analysis: sweeping setpoints and speeds across their feasible ranges while computing total plant kW at each combination. This can be done with a physics-based simulation tool during design, or with controlled step tests during commissioning, where one variable is adjusted while monitoring the impact on total plant power.

This is why fixed setpoints cannot find the optimum. A CHWS of 6.5°C, a CWS of 32°C, and all pumps at full speed may be near-optimal at design load on the hottest day of the year. On every other day, at every other load, it is some distance from the minimum, and that distance costs energy.

It is also why single-subsystem VFD retrofits capture less than the full opportunity. Adding a VFD to the CW pumps allows movement along one axis of the space. The pump finds its own minimum, but that is not the plant minimum, because chiller, CT fan, and secondary pump setpoints remain fixed.

The retrofit captures pump-axis savings but misses the interaction savings from simultaneously adjusting the other variables.

Part-load behaviour: why the tradeoff matters most below design capacity

Most commercial buildings in India (hotels, malls, offices) operate at 50–70% of design cooling capacity for the majority of their operating hours. Hospitals tend toward higher average load fractions, but the principle holds during shoulder seasons and night-time.

The load profile follows an approximate bell curve centred on the typical operating load. In the SOLACE simulation model, the spread varies by building type: hotels at σ ≈ 15%, malls at σ ≈ 20%, and data centres at σ ≈ 8%.

The significance for total plant kW optimisation is that the sweet-spot argument bites hardest in the 50–70% load band, for two reasons.

First, at full load (90–100%), there is little room to optimise: all equipment runs near rated capacity, and pump speed reductions are limited by minimum flow requirements. At very low loads (below 30%), total kW is low in absolute terms and hours spent in this band are few.

In the 50–70% band, the chiller has unloaded (its internal part-load efficiency peaks at approximately 70% of rated load for screw and centrifugal types, then degrades at both higher and lower fractions). But in a fixed-setpoint plant, the pumps and fans have not unloaded; they continue to draw rated power regardless of load.

The divergence between the chiller's operating point and the auxiliary equipment's operating point is at its widest. This is where the sweet-spot minimum is farthest from the fixed-setpoint operating point, and where the absolute kW gap is largest.

Second, the hours spent in this band multiply the per-hour savings. A building operating at 60% load for 3,500 hours per year with a 30 kW gap between fixed-setpoint and optimised operation is losing 1,05,000 kWh annually. At ₹9.5/kWh, that is approximately ₹10 lakh per year from this load band alone.

The connection to IKW/TR is direct. A plant tracking IKW/TR continuously will see it fluctuate with load: rising as load drops (because auxiliary power stays fixed while cooling load decreases, pushing up the kW/TR ratio).

The sweet-spot strategy minimises IKW/TR across the full load profile, not just at design point. The goal is to flatten the IKW/TR curve so that part-load efficiency approaches design-point efficiency, rather than degrading as load decreases.

From physics to practice

The physics described above defines the control strategy required to deliver whole-plant savings. Not fixed setpoints chosen during commissioning and never revisited. Not single-subsystem VFD retrofits that capture one axis of a multi-dimensional optimisation. A load-band strategy that adapts every setpoint and every speed to the current operating point.

In practice, a chiller plant manager (CPM) encodes this in a band table: at each load level (20% through 90%), the CPM holds a strategy row specifying CHWS setpoint, CWS approach target, secondary pump speed, CW pump speed, CT fan speed, and AHU fan speed.

At 90% load, setpoints stay conservative and speeds stay high to protect capacity. At 50% load, CHWS resets upward to 8.0–9.5°C, CWS approach widens to WBT+5–6°C, pump speeds drop to 60–80%, and CT fan speeds drop to 40–75%.

Each row is a different sweet-spot solution for that load condition. As the building's cooling load moves through the day, the CPM shifts between bands, tracking the minimum of the total kW curve in real time.

The band table is not a substitute for feedback control. It is the feedforward component that positions the plant near the sweet spot for each load condition. Feedback loops (ΔP reset, approach temperature control, return temperature monitoring) handle fine adjustment within the band.

Without the feedforward, a purely feedback-driven system would need to search for the minimum in real time, a process that takes hours given the thermal mass and response times of chiller plant equipment (rate-of-change limiters at 0.5°C per 15 minutes are typical), by which time the load has shifted and the minimum has moved.

The constraint infrastructure matters. If the secondary pumps lack VFDs, the CPM cannot modulate pump speed, and that axis of optimisation is locked at 100%. If the plant uses 3-way valves, CHWS reset is clamped because raising CHWS causes ΔT collapse and pump flow increase that offsets the chiller benefit.

The sweet-spot space narrows with less capable equipment. The band table encodes what is physically achievable, not what is theoretically ideal.

The interdependency premium, explored in a subsequent post, quantifies what this integration is worth: the savings from optimising all five subsystems together exceed the sum of optimising each alone.

That gap is not marketing; it is the shape of the sweet-spot curve. A piecemeal approach moves along one axis at a time, finding local minima. An integrated approach moves through the full space, finding the global minimum.

The minimum of the total curve is the only number that matters. Everything else is a partial optimisation masquerading as the answer.