Calculate data center free-cooling hours from an hourly weather file

How do you calculate free-cooling hours?

Calculate data center free-cooling hours by testing each valid weather-file hour against the system's temperature limit, approach temperatures and operating restrictions, then summing the qualifying hours. With Anvilfield inputs of 15°C supply water and 5°C combined tower/exchanger approach, the full-cooling weather threshold is 10°C wet-bulb. Anvilfield estimates 4,880 qualifying hours, or 55.71% of the Chicago O'Hare TMY3 year, versus 4,280 hours at a 10°C dry-bulb threshold for an equal-approach dry-cooler comparison. These are weather-eligible hours; plant capacity and operating permissions are checked separately.

Sources: Chicago TMY3 weather, ASHRAE 2020 tower approach, PsychroLib 2.5.0.

Weather-eligible hours pass the ambient screen; capacity-qualified hours also meet coincident load; recorded operating hours come from plant trends. Choose the cooling path, validate the weather, derive wet-bulb and classify intervals before aggregating. Retain provenance, equations, inclusion rules, sensitivity and monthly availability together.

The free cooling calculation starts after the system boundary is chosen. Use the data center economizer guide for the mode arrangement, then retain the inputs and row-level results described here.

Does the system use dry-bulb or wet-bulb?

Use outdoor dry-bulb for a dry-fluid cooler and entering-air wet-bulb for an evaporative tower. Baltimore Aircoil's technical explainer distinguishes those heat-rejection limits; a water circuit connected to a dry cooler still follows dry-bulb. Trace the heat path before selecting a weather column. BAC cooling methods.

ASHRAE's 2020 HVAC Systems and Equipment Handbook, Chapter 40, defines cooling tower approach as leaving cold-water temperature minus entering-air wet-bulb; SPX uses the same positive difference. Tower range is entering water minus leaving water: range follows heat load and flow, while achievable approach follows tower capability at the stated load, flow and entering-air condition. ASHRAE 2020; SPX approach definition.

For a closed-circuit evaporative fluid cooler, identify what the OEM leaving-fluid rating already includes. Add only downstream exchanger approaches; adding the internal coil again understates available hours. The tower operation guide explains the process, and the cooling-system-types overview compares the architectures.

Which hourly weather file should you use?

Use a typical meteorological year for a repeatable representative-year screen and actual-year observations for variability or comparison with plant trends. NREL's TMY3 manual, revised May 2008, describes a year assembled from selected typical months and cautions against using it to size for extremes. Keep design-weather sizing separate from the annual duration calculation. TMY3 manual.

Anvilfield's Chicago O'Hare TMY3 calculation uses 8,760 hourly records from station 725300, with the original weather file, its SHA-256 hash and the psychrometric calculation retained for reproduction. The provenance table records the station, time basis, source period and model assumptions together. Retain source/uncertainty flags: NREL's May 2008 manual describes filled meteorological inputs, so zero missing fields does not mean every value was measured.

NOAA's Integrated Surface Database is an actual-weather alternative containing hourly and synoptic observations with quality information. Normalize observations into defined intervals before counting; several reports within an hour do not create several cooling hours. Match station elevation and exposure to the facility, and record any supported site correction separately. Tower plume recirculation changes entering-air conditions at the equipment. NOAA ISD; ASHRAE tower conditions.

For recent climate conditions, rerun a declared span of complete actual years and report each count plus the minimum, median and maximum beside the TMY benchmark. A TMY percentage describes its constructed year; it is not a future-demand probability.

How do you validate timestamps and missing weather?

Validate calendar coverage before calculating availability. A complete ordinary year has 8,760 one-hour intervals and a complete leap year has 8,784; subhourly records contribute their actual duration. Build the expected interval grid for the declared schedule, then join observations onto it: absent rows become unknown hours. Never stretch the preceding observation across a gap without a declared filling method.

EnergyPlus 24.1 specifies missing cutoffs of 99.9°C for dry-bulb and dew point, 999 for relative humidity, and 999999 Pa for station pressure. Reject blanks, nonfinite entries and values at or above the required field's cutoff before physical checks. The table distinguishes the EPW format range from the psychrometric route's accepted inputs. EPW Data Dictionary, v24.1.0.

The Chicago file has one record per hour, hours 1 through 24 and minute 0. EnergyPlus 24.1 documents minutes 1 through 60, so preserve this file's raw labels and normalize from its hour-ending field: hour h represents [h-1,h) on the nominal day, with hour 24 ending at next-day midnight. Do not interpret its minute 0 as an hour-start timestamp or apply daylight-saving changes to its UTC-6 local-standard-time series. This screen assigns each reported weather state to its one-hour interval; subhourly crossings require finer data.

Retain TMY source-year labels but preserve the nominal month/day sequence; sorting by source year destroys the constructed year. Check nominal month/day/hour/minute coverage and monthly lengths. For actual multi-year observations, identify duplicates by normalized timestamp and duration across the declared years, rather than by month and day alone.

What denominator belongs next to the percentage?

Report qualifying, valid modeled, scheduled and unknown hours for the same schedule. Exclude off-schedule intervals from every total; an unknown hour is scheduled but lacks a required input. For missing required input, classify the interval as unknown: qualifying/scheduled is the lower availability bound, and (qualifying + unknown)/scheduled is the upper bound. Label qualifying/valid separately as conditional on valid data, and never silently expand a partial year to 8,760 hours.

Align civil-time operating schedules and tariffs to the file's fixed offset before joining them. This Chicago screen has 8,760 valid and scheduled hours, zero missing required-weather records and zero duplicate intervals.

How do you derive wet-bulb from an EPW file?

Calculate wet-bulb from dry-bulb, a humidity input and station pressure. Standard EPW supplies dry-bulb, dew point, relative humidity and pressure, rather than a dedicated wet-bulb field. Skip eight header lines; year through minute occupy columns 1 through 5, source/uncertainty flags occupy column 6, and dry-bulb, dew point, RH and pressure occupy columns 7 through 10. Preserve the flags. EnergyPlus 24.1.

Select SI units in PsychroLib 2.5.0 and call GetTWetBulbFromTDewPoint(Tdb, Tdp, P) with temperatures in °C and station pressure in Pa. An RH-based route uses a fraction, so 50% becomes 0.50; sea-level-corrected pressure is not the required station pressure. PsychroLib identifies its basis as the 2017 ASHRAE Handbook, Fundamentals, Chapter 1, including below-freezing branches. Pinned PsychroLib source.

Validate the selected humidity route. Flag an invalid unused RH field without rejecting wet-bulb from valid dry-bulb, dew point and pressure. The Chicago verifier also checks RH; general imports need route-specific flags. If dew-point and RH routes disagree, retain the chosen route and quantify changed classifications rather than switching silently.

The first weather row, source year 1986, January 1, hour 1, has dry-bulb -12.2°C, dew point -16.1°C and station pressure 99,500 Pa. Anvilfield calculates wet-bulb -13.05°C, which passes the 10°C screen; winter permission is tested separately. Keep the unrounded calculated wet-bulb for classification, using the psychrometric chart guide to interpret the variables.

How do approach temperatures set the changeover?

Subtract tower approach and heat-exchanger approach from required supply-water temperature to obtain the maximum qualifying outdoor wet-bulb. Tower leaving water is warmer than entering-air wet-bulb by the tower approach; the load-side heat-exchanger outlet is warmer again by the cold-end exchanger approach. Add pump heat or another exchanger explicitly when it falls inside the delivery boundary. ASHRAE 2020 approach definition.

Define supply Ts and return Tr on the exchanger's load side, with Tr greater than Ts. The worked inputs are 15°C supply, 21°C return, 3°C tower approach and 2°C HX approach; the latter is load-side outlet minus tower-side inlet.

Support the cold-end screen with an exchanger selection checking both terminal temperature differences, flow, fluid and capacity. Glycol changes heat capacity, viscosity and equipment performance; use selected curves instead of a universal temperature penalty. The waterside economizer guide provides loop context.

Estimate thermal fraction by dividing the temperature drop supplied by the economizer by the required return-to-supply drop, bounded between zero and one. The fraction assumes constant load-side flow. Limit its thermal output to selected economizer capacity at that interval's temperatures, flows and fluid.

Where does partial free cooling start and stop?

In this fixed-approach model, full cooling covers wet-bulb at or below 10°C, partial cooling covers above 10°C but below 16°C, and no useful precooling remains at or above 16°C. The upper boundary is 21°C return minus 5°C approach; equality leaves no modeled temperature difference to remove return-water heat.

At 13°C wet-bulb, the model delivers 18°C water and removes half the 21°C-to-15°C drop. Anvilfield estimates 500 kW economizer cooling and 500 kW remaining chiller thermal load for a declared 1,000 kW coincident thermal load with sufficient selected capacity.

DOE's revised July 2024 design guide, §5.4.2, describes an upstream series exchanger precooling water before the chiller trims to setpoint. An integrated parallel plant can also economize during simultaneous chiller operation when its piping and controls permit, but calculate branch flows and mixed supply temperature explicitly. The series fraction cannot be applied unchanged to a split-flow arrangement. An exclusive economizer-or-mechanical sequence uses only its qualifying full-mode hours; use selected performance maps for a plant estimate. DOE design guide.

Which hours count, and how should weather bins be drawn?

Classify raw hourly values first, then aggregate bins. Anvilfield's full-weather rule includes equality. In the Chicago O'Hare TMY3 file, including dry-bulb values exactly equal to 10°C gives 4,280 eligible hours, while a strict below-10°C rule gives 4,143 hours. The difference is 137 records; two calculated wet-bulb values also equal 10°C, giving 4,880 inclusive versus 4,878 strict hours. Weather inputs; Wet-bulb method.

Use nonoverlapping, right-closed bins as shown below. A bin that crosses a changeover threshold cannot be classified by its midpoint: return to the hourly records. Each valid wet-path interval belongs to exactly one state, so full plus partial plus no contribution equals valid modeled duration. Dry-path eligibility is an alternative design comparison and is never added to wet-path hours.

How do you reproduce the count in Excel?

Import the downloadable hourly-results.csv into an Excel table named Weather. Name three cells Ts, Tr and Atotal and enter 15, 21 and 5, all in °C; add dt_h = 1 and valid_wet = TRUE and valid_dry = TRUE to every validated Chicago row. For another dataset, first build its scheduled interval grid, including missing rows, then derive duration and required-field validity flags. Apply these formulas only to scheduled intervals.

Sum the duration columns annually, then pivot by month or use SUMIFS to reproduce the monthly table. Keep wb unrounded and obtain it from the pinned calculation instead of a display column. Changing the three named cells changes the screen; retain Tr greater than Ts and reconcile the three wet states plus unknown duration against the schedule.

Which operating restrictions reduce weather availability?

A usable full-cooling interval must pass every gate: weather, positive coincident load, schedule, selected capacity, operating permission and control state. Compare available economizer capacity with that hour's thermal load at the required delivery temperature and the permitted equipment configuration. Apply changing supply and return temperatures and coincident loads hour by hour. Keep every failed-gate reason, but exclude an interval only once when reasons overlap.

The baseline weather screen includes freezing weather and applies no outages. A plant model needs the approved winter envelope, minimum water temperature, fluid specification, cell and flow constraints, and freeze-protection sequence; there is no universal 0°C outdoor cutoff. Use tower commissioning and water-treatment records to establish these permissions. ASHRAE 2020 cold-weather operation.

Direct airside calculations must resolve mixed and supply air, fan heat, moisture limits and available airflow. Outdoor RH alone does not establish supply RH after mixing or heating; smoke and contamination lockouts require their own dated records. EPA's airside guidance describes the mixing, humidity and filtration considerations. EPA ENERGY STAR.

For direct-to-chip liquid cooling, include CDU approach and model the residual air load separately. Whole-facility compressor-off hours are the intersection of the liquid-loop and residual-air full-mode flags, not the average of their percentages. Use the permitted facility-water temperature at the modeled boundary; an IT water-class maximum is not an outdoor threshold.

How should deadband and restart delay be modeled?

Run the sequence chronologically with documented initial state, enable and disable thresholds, and delays. A histogram discards the order required to calculate switching; monthly TMY joins also need a stated control-state convention. Use weather and BMS trends fine enough to resolve control delays; this hourly screen applies no hysteresis or delay.

Worked example: Chicago O'Hare hourly availability

At a 10°C inclusive weather threshold, Anvilfield estimates 4,880 wet-bulb-eligible hours and 4,280 dry-bulb-eligible hours in the same Chicago O'Hare TMY3 file, with no operating lockouts applied. Their annual shares are 55.71% and 48.86%; the equal-approach wet screen gains 600 hours. The weather and input manifest supplies the complete calculation boundary.

The wet-path classification also gives 1,683 partial hours, or 19.21%, and 2,197 no-contribution hours, or 25.08%. The three states reconcile to 8,760. The monthly table shows where partial operation matters: an annual full-hour total alone hides the shoulder-season opportunity.

Method note. Anvilfield estimates use the Chicago O'Hare TMY3 EPW downloaded on November 5, 2026, with 8,760 valid hourly records. Wet-bulb is calculated from each record's dry-bulb, dew point and station pressure using PsychroLib 2.5.0 in SI units. The model holds supply and return water at 15°C and 21°C, uses a combined 5°C approach, counts full cooling at wet-bulb ≤10°C and partial cooling above 10°C but below 16°C, and sums one hour per qualifying record. Approach sensitivity changes only the combined approach; displayed temperatures are rounded after classification, and the weather screen applies no operating lockouts.

How many hours does one degree of approach cost?

Anvilfield estimates that increasing combined approach from 5°C to 6°C at 15°C supply removes 260 Chicago O'Hare TMY3 wet-bulb-eligible hours. Reducing approach from 5°C to 4°C adds 264 hours. The increments differ because the weather distribution near each threshold differs.

Hold supply, schedule, weather and the inclusive test fixed when comparing approaches. Increasing supply temperature or reducing approach shifts the screening threshold equally, while equipment capacity and power require separate evaluation. A 1°C approach difference equals 1.8°F; do not add 32 when converting a temperature difference.

Record a local entering-air correction or sensor bias separately from equipment approach. Before attributing lost hours to fouling, compare calibrated weather and water measurements at matched load, flow and fluid, and check plume recirculation. A measured approach change becomes a useful maintenance signal only after those conditions match.

Table G changes only combined approach, using the same Chicago TMY3 records and PsychroLib 2.5.0 conversion.

How do eligible hours become energy and cost savings?

Calculate interval energy savings by multiplying duration by the difference in total cooling-system electrical input, including compressors, pumps, fans and auxiliaries. Apply the interval's energy rate afterward. Partial-mode savings need their own power model because a thermal fraction is not an electrical-input fraction.

In the declared Anvilfield energy case, all 4,880 full-weather hours are usable at 250 kW baseline cooling input and 50 kW economizer input. The 200 kW reduction at $0.12/kWh yields an estimated 976,000 kWh and $117,120 reduction in energy charges. These power and rate values are calculation inputs.

Demand-charge savings require a coincident reduction at the applicable billing peak; annual kWh does not determine them. Retain intervals with negative savings where added pump or fan input exceeds avoided compressor input. Apply water-use restrictions as operating gates when they prohibit wet operation.

Add project water, sewer, treatment, maintenance and capital quantities before calculating net operating savings or payback. Compute annual PUE from annual facility and IT energy, as discussed in the PUE guide; averaging instantaneous PUE readings or treating eligible hours as a PUE reduction gives a different quantity.

Why do calculators and published limits disagree?

Match the sink, delivery boundary, load, approach, humidity limits, weather vintage and full-versus-partial definition before comparing counts. The AKCP-published Free Cooling Tool credits Open-Meteo ERA5 and provides environmental-class and baseline/free-cooling PUE inputs. Reconcile a matched-input run rather than treating unlike model boundaries as contradictory results. Free Cooling Tool.

ASHRAE TC 9.9's July 8, 2024 letter, discussing the 2021 fifth-edition Thermal Guidelines, gives H1 a 22°C recommended maximum and 27°C allowable maximum while discussing 27°C as the broader recommended maximum. These are IT environmental limits, not tower wet-bulb thresholds; trace outdoor conditions through the cooling system to the equipment environment. ASHRAE TC 9.9 letter.

Trane's October 2021 PKG-PRC024E-EN catalog contains two different enable descriptions: page 20 gives a factory 7°F entering-water versus mixed-air offset, while page 36 gives 4°F plus approach, with a default 4°F approach, yielding 8°F. Use the selected unit's approved sequence and as-left settings to resolve the conflicting defaults; record enable and disable settings independently. Trane catalog.

Vertiv's Liebert DSE calculator concerns pumped-refrigerant economization and uses location, load and operating inputs. Its product model and Trane's coil sequence serve different equipment from the tower/HX screen. Follow the selected product's performance and controls for its operating calculation. Vertiv calculator description.

What belongs in the calculation and procurement scope?

Specify three scope levels: a weather screen with fixed inputs; a plant model adding selections, load and control sequences; and a savings model adding interval electrical input and tariffs. Each should deliver source weather, a quality report, the delivery boundary, hourly classifications, monthly summaries and approach sensitivity. Price the engineering effort from the requested inputs and deliverables.

Estimators scoping economizer retrofits for public facilities find early project signals in budgets and capital plans through PorchRocket Bid Control.

Assign capacity and approach to supplier selections, changeover to the approved sequence, exclusions to the owner's operating records, energy rates to the utility tariff and verification to the BMS export. Rerun the calculation when supply setpoints, installed equipment, load, fluid or control logic changes. Keep the station and source period visible in every option comparison.

Calculation and field-verification checklist

Accept the calculation when another reader can reproduce its hours from the saved inputs and trace every operational exclusion to a record. Compare operating trends with the same year's local weather and load/supply schedule. Keep predicted full status, economizer enable, compressor status and delivered water temperature as separate signals.

DOE's revised July 2024 design guide, §8.9, recommends continuously trended measurements archived for at least one year to establish annual energy totals. Choose a trend interval that resolves the installed changeover sequence, with weather, thermal delivery and total cooling electrical input on the same time basis. DOE monitoring guidance.

When weather qualifies but a compressor remains on, check selected capacity, supply setpoint, valve position, sensor calibration, approach, active lockout and minimum-run timer. When weather and delivery agree but savings are low, inspect fan/pump input and baseline power assumptions. The BMS commissioning guide explains the control evidence to retain.

FAQs, reproducible inputs and how to cite this page

Read the twelve calculation FAQs, or reproduce the tables from the saved package. Download calculation inputs, read the parameter manifest, and download hourly classifications. The package contains the original weather bytes, pinned PsychroLib dependency with its MIT notice, calculation script, row output and summary tables.

Extract the package into one directory and run python3 calculate.py. The script reads the adjacent weather file and local psychrolib.py; the manifest identifies dependency commit, checksums, units, humidity route, timestamps, interval duration, approaches, inequalities and denominator. Compare regenerated monthly, annual and sensitivity results with the retained results.json.

The verifier asserts this Chicago dataset's complete non-leap coverage and valid inputs. To adapt it, implement path-specific missing-data handling and unknown duration, preserve actual-year interval lengths, and emit the selected denominator. Keep original weather bytes unchanged and document every transformation.

Sources checked

1. S1. DOE EnergyPlus: Chicago O’Hare TMY3, NREL dataset 2008
2. S2. NREL: Users Manual for TMY3 Data Sets, revised May 2008
3. S3. NOAA NCEI: Global Hourly Integrated Surface Database, undated
4. S4. ASHRAE: HVAC Systems and Equipment, 2020, Chapter 40
5. S5. PsychroLib 2.5.0: pinned source, based on ASHRAE Fundamentals 2017
6. S6. EnergyPlus v24.1.0: EPW Data Dictionary
7. S7. SPX: Cooling Tower Approach, undated
8. S8. Baltimore Aircoil: dry, evaporative, adiabatic and hybrid cooling, undated
9. S9. DOE/FEMP and NREL: Best Practices Guide, revised July 2024
10. S10. EPA ENERGY STAR: Use an Air-Side Economizer, undated
11. S11. ASHRAE TC 9.9: technical input letter, July 8, 2024
12. S12. Trane: PKG-PRC024E-EN, October 2021, pages 20 and 36
13. S13. Vertiv: Liebert DSE economizer calculator description, undated
14. S14. AKCP: Free Cooling Tool interface, undated, comparison source

Reviewed by Anvilfield on 2026-11-05. Sources checked: 2026-11-05. Update log: 2026-11-05, final article review; all fourteen listed sources reopened, calculation tables reproduced, missing-interval and partial-operation rules clarified.

How to cite this page: Anvilfield, Calculate data center free-cooling hours from an hourly weather file, 2026, https://anvilfield.com/field-guides/datacenter/data-center-free-cooling-weather-bin-model/. For a specific table, append its fragment and identify the Chicago O'Hare TMY3 dataset, PsychroLib 2.5.0 and review date.

Related tools

FAQ

People also ask

• What is BMS commissioning?
• How are data centers cooled?
• What is direct-to-chip liquid cooling?
• What is free cooling in a data center?
• What is PUE?