Chiller Plant Loading Analysis

This report analyzes chiller plant operations at the Redstone Technology Campus for calendar year 2025. The plant consists of 7 water-cooled chillers with a combined design capacity of 9,850.0 tons, housed in the central plant. Data was collected at approximately 5-minute intervals from the BuildingLogiX Data Exchange (BDX) platform, comprising 101,664 observations across the full year.

The plant most frequently operated with 3 chillers running, accounting for 29.8% of the year. The median plant load was 2,318 tons (23.5% of design capacity). The peak instantaneous (5-minute) load reached 7,999 tons (81.2% of design), while the peak sustained (1-hour average) load was 7,970 tons (80.9% of design). The fleet accumulated a combined 2,520 start cycles across all 7 chillers.

• Plant is oversized for typical demand. Median load is only 2,318 tons (24% of the 9,850-ton design capacity). The plant runs 3 chillers most of the time and rarely needs more than 4. Five or more chillers were required for only 5.4% of the year.
• Chiller_2 is disproportionately loaded. It spends 28% of its running hours above 90% capacity, 3.1x the fleet average. When it’s maxed out, 3 other chillers are typically running, meaning capacity exists to redistribute load. This suggests the sequencing logic favors Chiller_2 rather than balancing across the fleet.
• Significant short-cycling problem, especially in shoulder seasons. The fleet logged 999 cycles under 60 minutes, 133 under 30 minutes, and 48 under 15 minutes. Chiller_6 (533 starts, 47% short-cycle rate) and Chiller_7 (457 starts, 51% rate) are the worst offenders. January and February show the highest short-cycling rates, with Chiller_7 reaching 72% in January. This indicates the staging deadband is too narrow for shoulder-season load fluctuations.
• Low delta-T syndrome is pervasive and originates on the distribution side. The median system delta-T is only 5.2°F, and 48% of running hours fall below 5°F. Only 2% of running hours achieve a healthy delta-T of 10°F or above. Low delta-T is typically caused by load/distribution-side conditions: three-way valve bypass at AHUs, oversized or poorly controlled secondary pumps, or fouled coils. The impact cascades back to the plant by degrading chiller efficiency, driving up pumping energy, and forcing premature chiller staging as supply temperature rises.
• Chiller_4 is heavily underutilized. Only 983 hours of runtime (11% of the year), yet it carries a 62% average load when it does run. It appears to be a swing/backup machine rather than part of the normal sequence.
• Startup loading is generally acceptable. Median startup load across the fleet is 30–40%, meaning chillers are ramping into load rather than slamming on. Hard starts (>80% at first sample) are rare.
• Little to no data visibility on distribution and end loads. Beyond the central plant, metering drops significantly at the distribution level (no flow/BTU data) and near zero at end loads. The low delta-T problem identified in this report almost certainly originates on the load side (three-way valve bypass, oversized pumps, fouled coils), but without distribution and end-load sensing, the root cause cannot be diagnosed and the team is left guessing. Further plant-side optimization has reached its ceiling; meaningful efficiency gains now require visibility into where the chilled water is going and how it is being used. See Section 10: Data Accountability & Strategic Energy Planning.

The table below summarizes annual runtime, estimated start/stop cycles, and loading statistics for each chiller. Chiller_6 logged the most runtime at 4,517 hours (51.6% of the year), while Chiller_4 had the least at 812 hours (9.3%). Chiller_6 recorded the highest number of start cycles (516), which may indicate aggressive staging transitions or frequent load swings.

High cycle counts relative to peers can accelerate wear on compressor contactors and increase the risk of nuisance trips. Machines with disproportionately high cycles compared to their runtime may benefit from optimization improvements.

When a chiller starts, the initial load it receives is a key reliability indicator. A hard start is defined as a chiller receiving >80% load within its first 5-minute sample after startup. Hard starts stress compressor bearings, guide vanes, and motor windings, increasing the risk of mechanical failure and reducing equipment life. Ideally, chillers should ramp gradually through part load before reaching high capacity.

The box plot above shows the distribution of startup loads for each machine. Most starts across the fleet land in the 30–50% range, which is reasonable. Hard starts (>80%) are infrequent but worth monitoring.

Short cycling accelerates mechanical wear, wastes startup energy, and indicates the staging logic is hunting around a load threshold. The table above categorizes cycles into three severity tiers:

• Extreme (<15 min): Compressor oil hasn't circulated fully, severe bearing wear risk, refrigerant migration issues. Fleet total: 48 cycles.
• Critical (<30 min): Chiller hasn't reached thermal steady-state, stress on evaporator/condenser tubes. Fleet total: 133 cycles.
• Short (<60 min): Startup energy penalty not recovered, added mechanical wear on compressor bearings and seals, staging deadband too narrow. Fleet total: 999 cycles.

Chiller_7 has the highest short-cycle rate at 50% of all cycles.

The heatmap below reveals when short cycling is most prevalent. Red cells indicate months where a high percentage of a chiller's cycles were shorter than 60 minutes. Shoulder seasons (spring and fall) typically show the worst short cycling because cooling load oscillates around staging thresholds. Summer months generally have near-zero short cycling since load is consistently high enough to keep machines running.

Staging analysis shows how many chillers operate simultaneously throughout the year. The plant spent 7,985 hours (94.2% of the year) running 1 to 4 chillers. Five or more chillers were required for only 5.4% of the year, confirming that the plant rarely approaches its full design capacity.

The heatmap below reveals seasonal and diurnal patterns in chiller staging. Darker cells indicate periods with more chillers running on average. Summer afternoons show the highest staging levels, while winter nights represent minimum cooling demand.

The load duration curve sorts the total plant tonnage from highest to lowest across all 101,664 data points. It reveals the percentage of time the plant operates at or above any given load level. The 95th percentile load is 5,040 tons, meaning the plant exceeds this level only 5% of the time. The median load is 2,318 tons (23.5% of design).

This section examines how each chiller is loaded when running. The ideal operating range for centrifugal chillers is typically 30–90% of rated capacity. Operating below 30% risks surge conditions and poor efficiency, while sustained operation above 90% leaves no margin for transient loads.

The heatmap shows the percentage of each chiller's running hours spent in each 10% load band. Concentration in the center bands (40–80%) indicates healthy loading; heavy presence in the extreme bands warrants further investigation.

The overlaid histogram provides a higher-resolution view of where each chiller spends its running time. The shaded zones mark the non-ideal regions below 30% and above 90%.

Key observation: Chiller_2 spends 28.3% of its running hours above 90% capacity, meaning that out of every 100 hours it runs, roughly 28 of those hours are near full load. The fleet average is only 9.2% of running hours above 90%. When Chiller_2 is loaded above 90%, there are typically 3 other chillers running simultaneously, meaning available plant capacity exists to redistribute load. This pattern suggests the chiller sequencing logic may be favoring Chiller_2 rather than balancing load across the fleet. Sustained high-load operation reduces compressor life, increases energy consumption, and leaves no margin for transient load spikes.

At the other end, Chiller_7 and Chiller_3 concentrate the majority of their running hours in the 40–50% range, indicating they serve primarily as base-load or lead machines at lower plant demand levels.

The faceted panels below show each chiller's hourly running capacity over the full year. This view makes it easy to compare seasonal utilization patterns across the fleet and identify machines that are consistently loaded higher or lower than peers. Gaps indicate days when a chiller did not run. Persistently high daily averages (near 80–100%) suggest the machine is being relied on disproportionately, while long idle periods may point to maintenance outages or sequencing logic that deprioritizes the machine.

The time series below shows total plant tonnage (blue) and the number of chillers running (red) over the full year at hourly resolution. This view highlights seasonal ramp-up/ramp-down patterns and any anomalous events such as unexpected full-plant shutdowns or sudden load spikes.

The stacked bar chart shows each chiller's contribution to total monthly runtime hours. This reveals seasonal patterns in chiller utilization and whether specific machines are favored or idled during certain months. Uneven distribution across similarly-sized machines may indicate manual overrides, maintenance windows, or sequencing preferences.

The treemap below sizes each chiller by its total annual ton-hour production: the product of running capacity and design tonnage integrated over time. Larger blocks represent machines that contributed the most cooling to the plant. This accounts for both how long a chiller ran and how hard it was loaded.

Plant cooling load is strongly correlated with outdoor air temperature (OAT). The scatter plot below shows total plant tonnage vs. OAT at hourly resolution, with points colored by the number of chillers running. A well-staged plant will show clean transitions in color (chiller count) as OAT increases. Scattered colors at a given OAT level may indicate inconsistent staging logic.

Overstaging at low loads: There are approximately 37 hours where 3 chillers are running to produce less than 1,600 total tons, roughly 505 tons per machine, or just 35–40% load each. This occurs primarily in cold-weather months (October through December account for the majority) when OAT averages around 47°F. At these load levels, one or two chillers could comfortably handle the demand at much better part-load efficiency. Running three machines at 35–40% each wastes compressor energy, increases auxiliary power (condenser pumps, tower fans), and contributes to the short-cycling pattern observed in the shoulder-season analysis. The staging-down logic should be reviewed to ensure chillers are shed sooner as load drops.

The chilled water supply (CHWS) and return (CHWR) temperatures are key indicators of plant performance. The CHWS should track the active setpoint. A widening gap between CHWR and CHWS (delta-T) at higher loads indicates effective heat transfer across the evaporator. Persistent setpoint deviations may indicate insufficient capacity or control issues.

The system delta-T represents the temperature difference between the chilled water return and supply. A healthy plant typically maintains 8–14°F delta-T under load. The red-shaded zone below 5°F highlights periods of critically low delta-T. Low delta-T is a compounding problem: it forces the plant to circulate more water to meet the same cooling load, driving up pump energy consumption. It also degrades chiller performance by reducing the evaporator approach temperature, which can push chillers into surge or force additional machines online prematurely. Common root causes include three-way valve bypass at AHUs, oversized or improperly controlled secondary pumps, and coils that are fouled or undersized for current loads.

The duration curve sorts delta-T values from highest to lowest across all running hours, showing the percentage of operating time at or above each delta-T level.

Effective chilled water optimization requires visibility at every level of the distribution system, not just the plant. The pyramid below illustrates the current state of data sensing and metering at Redstone Technology Campus, from the central plant through to end-use loads. True plant optimization is only possible when flow, temperature, and power data are available at each tier, enabling accurate tonnage calculations, efficiency benchmarking, and load attribution. Load-side data is what connects plant operations to long-term capital planning. It determines whether the right investment is another chiller or a $50k valve retrofit on a secondary loop.

Current State: The plant itself is mostly complete. The distribution tier has partial visibility. Differential pressure, temperatures, and pump status/speed are trended, but flow metering and BTU calculations are absent, preventing true load quantification across secondary loops. At the tertiary level, two small BTU meters (laundry and security) provide minimal coverage but leave the vast majority of subloops unmetered. End-load visibility remains incomplete. This gap prevents the team from answering fundamental questions: Where is the chilled water going? Which loads are consuming the most energy? Where is the low delta-T originating? Without answers, optimization is limited to the plant in isolation, a ceiling that has largely been reached.

Plant Level (Mostly Complete → Complete)

• Confirm true kW metering on all 7 Trane CVHF chillers (dedicated CTs vs. derived from running capacity)
• Verify or install evaporator flow sensors (mag meters) for measured tonnage (GPM × ΔT × 500)

Distribution (Partial → Mostly Complete) | DP, temps, and pump speed already trended; flow metering is the critical gap

• Add flow meters to major secondary loop headers to unlock per-loop BTU/tonnage calculations
• Correlate existing pump speed and DP data with loop delta-T to identify bypass and oversized pumps
• Map secondary distribution architecture (which loops serve which building zones)
• Prioritize the 2–3 highest-flow / lowest-delta-T loops first

Tertiary Loads (Minimal → Partial) | Laundry and security BTU meters installed; expand to major subloops

• Deploy BTU meters on primary load zones and process subloops
• Use existing laundry and security meter data to characterize process vs. comfort load profiles

End Loads (Incomplete → Minimal)

• Trend CHW coil valve position on the 5–10 largest AHUs (valve at 100% + low coil delta-T = fouling or bypass)
• Monitor discharge and mixed air temps to identify units overcooling and reheating
• Spot-check fan coil units for two-way valves acting as three-way (common retrofit failure that kills delta-T)

Phase 1 (Immediate): Finalize plant-level sensing. Confirm true kW metering on all 7 chillers, verify or install evaporator flow sensors, and ensure all data flows to BDX at 5-minute intervals. This closes the plant tier to complete and unlocks accurate kW/ton benchmarking.

Phase 2 (Near-Term): Add flow meters to the 3–5 largest secondary distribution loops, pairing with the DP, temperature, and pump data already being trended. Install flow and temperature sensing on the primary-secondary bridge (decoupler) pipe. Bridge flow direction and magnitude are the single best indicator of whether the plant is producing more or less than the distribution system is consuming. Map distribution architecture. This moves the distribution tier from partial to mostly complete and reveals the root cause of the low delta-T problem.

Phase 3 (Near-Term, parallel with Phase 2): Extend BTU metering to tertiary subloops serving major load centers (primary load zones and process loads). Begin trending coil valve position on the top 10 AHUs. This enables load attribution, demand analysis, and targeted coil maintenance.

Phase 4 (Ongoing): Expand end-load monitoring as budget allows. Use the data accumulated in Phases 1–3 to build predictive load models, optimize chiller sequencing based on actual (not estimated) tonnage, and implement supply temperature reset strategies informed by real-time distribution conditions.

The table below lists the BDX trends referenced by this analysis.

Data Source: BuildingLogiX Data Exchange (BDX) at https://demo-site.bdxcloud.com
Collection Period: January 09, 2025 through December 28, 2025
Sample Interval: ~5 minutes (POINT-level aggregation)
Total Observations: 101,664

Trends Used:

• Trend 138 — Running Capacity (%) for all 7 chillers
• Trend 85 — Plant totals: tonnage, power, efficiency, OAT, wet bulb
• Trend 150 — CHW system temperatures: supply, return, setpoint
• Trend 115 — Outdoor air temperature (weather station + facility sensor)

Key Assumptions:

• A chiller is considered "on" when its runningCapacity > 0%.
• Start cycles are estimated by counting transitions from off (0%) to on (>0%) in the 5-minute data. Brief comm losses could inflate cycle counts.
• Tonnage is calculated as design tons × running capacity % for each chiller. Design tonnages are from nameplate data.
• The ideal chiller operating band is defined as 30–90% of rated capacity. This is a general guideline; manufacturer-specific curves may differ.
• Hourly resampling is applied to time-series and scatter charts for visual clarity. All tabular statistics use the full 5-minute dataset.

Equipment (from nameplate tags):

• Chillers 1–3: York YKEP3A3H, 1,400–1,500 tons, R-134a
• Chillers 4–5: Trane CVHG1500, 1,425–1,500 tons, R-514A
• Chillers 6–7: Carrier 19XRV1250, 1,250–1,275 tons, R-134a (2011)
• Total Plant Design Capacity: 9,850.0 tons