Printer Friendly

Predicting condensate collection from HVAC air handling units.


Recent water shortages in the southeastern US and elsewhere have increased the perceived value of fresh water beyond its financial cost. Local, interstate, and international disputes over water rights are likely to further this trend. The increasing concern for water resources within ASHRAE is reflected in two new Standards that address water use in buildings: Standards 189.1-2010 (Standard for the Design of High Performance Green Buildings Except Low-Rise Residential Buildings) and 191P (Standard for the Efficient Use of Water in Building, Site and Mechanical Systems). Both standards provide requirements for water using systems, addressing efficient usage as well as water reclamation and reuse.

Condensate collection from air handling units (AHUs) is one method of water reuse that has been successfully incorporated in new buildings and is even required in new construction in some locations. (1) Condensate can most easily be routed directly to a cooling tower sump, but with storage (and usually in conceit with a rainwater collection system) can be used for irrigation or ornamental purposes; with further processing it can be used for indoor applications such as toilet flushing or even potable water.

Tom Lawrence is a Public Service Associate in the Faculty of Engineering at the University of Georgia, Jason Perry is a Research Engineer in the Faculty of Engineering Outreach Service at the University of Georgia and Peter Dempsey is an engineering undergraduate student at the University of Georgia

While incorporating condensate collection into new buildings can be relatively straightforward, retrofitting existing buildings can be more complicated. Since existing buildings comprise approximately 98% of the building stock (the other 2% being new construction), they represent a significantly greater immediate benefit to society in terms of energy or water consumption savings. It is thus worthwhile to study and facilitate condensate collection retrofits in existing buildings.

It is easy to estimate the cost to install a condensate collection system, but it is more difficult to calculate the financial payback due to water savings. Dire water shortages in some areas may make the financial question moot, but for now in the majority of cases it is likely that some financial justification will be necessary.

Currently, prediction methods and tools are not widely available for evaluating whether condensate collection is worthwhile. Guz (2005) suggests a rule of thumb of 0.1 to 0.3 gallons (0.4--1.1 L) of condensate per ton of air conditioning, per hour of operation, but this only applies to San Antonio. A simple online calculator (no longer available) created by Wilcut and Fry determined the steady-state condensate production rate for a given set of conditions, but was not useful for predicting condensate over a season of varying weather. Painter (2009) developed a prediction model for dedicated outdoor air handling units with enthalpy wheel energy recovery, in which he used the expected difference in humidity ratio on the entering and leaving sides of a cooling coil. He developed the model to predict condensate production in three locations in Texas using annual daily average temperature and humidity data.

With this paper we present our methodology for predicting the amount of water collected from an air handling unit, and we describe an attempt to validate and refine the model with empirical measurements taken throughout the 2009 cooling season in Athens, Georgia. The model is designed to be adapted to any location for which hourly data are available.


The system used in this study was the second condensate collection retrofit installed at the University of Georgia (UGA), and has been in operation since February 2009. (see Lawrence, et al., 2010). It is comprised of a stainless steel collection basin measuring about 2 ft (600 mm) square and 9 in (230 mm) deep, a 1/6 HP (125 W) sump pump with an external diaphragm switch, and an analog totalizing water meter with 1/10th gallon (0.38 L) resolution. The system is equipped with a check valve above the pump to prevent backflow into the basin.

The basin was installed so that it could intercept the original path of the condensate drain pipe without changing the slope or the dimensions of the U-trap at the drain outlet from the AHU. An emergency overflow pipe is connected near the top of the basin and leads to the existing floor drain, so that the original drain path would be completed in the case of a pump failure.

The meter was installed at eye level in the vertical section of pipe above the pump so that it would always be measuring full pipe flow. This was a lesson learned from the first installation at UGA in which the meter was installed horizontally in a section of pipe with essentially open-channel flow, raising concerns that the meter might be fooled into reporting more water than is really flowing through it.

After the meter, the pipe runs up and over the AHU, through the penthouse wall, across the roof at a slope of 1/4 inch per foot (21 mm per m), and joins a pipe from another condensate collection system before dropping down an exterior wall to the sump of the building's cooling tower.

Measurement Equipment and Methods

Dataloggers were used measure the air temperature and relative humidity at the outdoor air intake and in the fan section of the AHU. We employed a combination temperature and relative humidity datalogger, which has a RH measurement accuracy of [+ or -] 3.5% from 25% to 85% over the range of 59[degrees] to 113[degrees]F (15[degrees] to 45[degrees]C) and [+ or -] 5% from 25% to 95% over the range of 41[degrees] to 131[degrees]F (5[degrees] to 55[degrees]C). The logger temperature accuracy is [+ or -] 0.72[degrees]F from 32[degrees] to 104[degrees]F ([+ or -] 0.4[degrees]C from 0[degrees] to 40[degrees]C).

We recorded fan speed during the study using fan motor current as proxy data for fan speed and hence airflow. A current transformer (CT) rated for 50 amperes was placed on one leg of the three phase motor circuit and connected it to a datalogger. The CT and datalogger combination is rated as being accurate to [+ or -] 2.25 A.

We synchronized and programmed the three dataloggers to record every five minutes, and we downloaded data from them about once a month.

Airflow Baseline

From our initial observations of the supple air fan operation, the fan current draw (as indicated on the variable frequency drive display) tended to be within a fairly narrow range, but did vary some. We obtained the baseline airflow of the AHU by conducting pitot tube traverses across the supply air ductwork leading from the AHU, and recorded the current to the fan. The baseline airflow measurements were made when the fan was in normal operation and the current draw near the 'nominal' level. We next also conducted pitot tube traverses with the fan speed manually set near the upper end and then again near the lower end of what our data loggers had recorded as being the fan normal operating range. From the pitot tube traverses we then calculated the supply airflow at the 'baseline' current, which was used as the primary reference point for determining the airflow for all data points during the cooling season.

Water Meter Validation

To ensure that the water meter was not affected by turbulence due to its proximity to the pump, a calibration procedure was performed using a graduated one gallon (3.8 L) container. One gallon at a time was added to the system's collection basin until the pump was triggered and the water pumped out. This process was repeated approximately 15 times, with a recording taken from the meter each time. There was no significant error between the measured input of water and the cumulative meter reading at the end of the process.


General method for computing condensation

For simplicity, consider the process of a unit conditioning 100% outdoor air (such as with a dedicated outdoor air system or DOAS). The psychrometric chart shown in Figure 1 represents a path of outdoor air as it passes across the cooling coil for the 0.4% cooling design condition in Athens, Georgia. Assuming a supply air condition of 55[degrees] F (12.8[degrees] C) and 85% relative humidity (wet bulb T=52.5[degrees] F or 11.4[degrees] C), the humidity ratio changes across the coil from 0.0141 to 0.0078 lb/[lb.sub.air] (kg/[kg.sub.air]). The difference in absolute humidity ([omega]) between the incoming outdoor air and supply air leaving the unit represents the amount of condensation that occurs. Thus, for every pound (or kg) of air supplied by the unit, 0.0141 - 0.0078 or 0.0063 pounds (kg) of water are condensed.


The total amount of condensate expected is determined by the equation below:

Condensate collected = Airflow x density x 60 [min/hr] x [DELTA][omega] (1)

Assuming for example 1,000 cfm (472 l/s) of outdoor air is being conditioned, the total amount of condensate expected would be.

Condensate = 1,000 [[ft.sup.3]/min] x [lb/13.133[ft.sup.3]] x 60 [min/hr] x (0.0141 - 0.0078) [[lb.sub.water]/[lb.sub.dry air]] = 28.8 [lb/hr] (13.1 kg/hr)

This is approximately 3.5 gallons (13.1 liters) per hour at the cooling design condition.

For the condensate prediction study described in this paper, a spreadsheet model was developed which computed an estimate of the condensate collection rate expected during each five-minute data logging time period through the course of the cooling season. The model uses the following data inputs:

* Outdoor air temperature and relative humidity (from data logger)

* Supply air temperature and relative humidity (from data logger)

* Outdoor air and supply air humidity ratio (computed from recorded data set)

* Supply fan input current (from data logger)

* Air handling unit supply airflow rate at 'baseline' flow and specific current input values (measured using pitot tube traverses at 'normal operating' fan speed and other points by manually adjusting fan speed at the VFD controller)

Spreadsheet model logic

The following computational steps are performed by the spreadsheet model for each of the five-minute data recording periods.

1. Compute the differential between outdoor and supply air humidity ratio. ([DELTA][omega])

2. Estimate the supply airflow rate for this period, assumed to be a function of the cube root of the current.

Supply Flow This Period = [([Current this period/Current at baseline flow]).sup.[1/3]] x Baseline flow (2)

Baseline current = 33.9 Amps (one leg of 3 phase circuit);

Baseline flow = 19,128 cfm

3. Multiply the supply volumetric airflow by density and 60 min/hr to get supply air mass flow rate ([m.sub.air]) in lbm/hr.

4. Compute the predicted condensate collection rate. Condensate flow = [m.sub.air] x [DELTA][omega] The result is the predicted condensate collected in lb/hr.

5. Convert the predicted condensate mass flow rate into a volume flow rate, gallons/hr and gallons/min (gpm).

6. Determine the predicted total condensate produced for this data logging time period Condensate produced = Volume flow (gpm) x 5 min

The total condensate produced is summed up between each field meter reading period, which was typically done on a daily basis during each weekday. For each period between field meter readings, the total predicted condensate is compared to the actual measured amount.


Baseline model evaluation

We made a total of 108 water meter readings during the 2009 cooling season. Data were recorded from April 1 through the end of September. By April 1 the installed water meter had already recorded some condensate collected but our data logging equipment was not installed and validated until then. A second AHU in this mechanical room was retrofitted with a condensate collection system in early October, and the output pipe from this new unit connected into the collection basin used for our study AHU, therefore we had to stop data recording on September 29. Since this was getting near the end of the condensate collection and cooling season, it was not felt to be a significant problem.

Readings were made nearly every day during the normal workweek (Monday through Friday; the building is locked to outsiders on weekends and holidays). The predicted condensation results were computed using the supply fan airflow as measured in the spring and using the installed data logger measurements for outdoor air and supply air conditions and these compared to the actual condensate collected using the installed totalizing flow meter.

Using the prediction method described earlier and the recorded measurements from the dataloggers, the condensate quantities predicted by the model were consistently lower than the actual quantities measured throughout the cooling season. The total predicted condensate for this air handling unit during the cooling season was 134,021 gallons (507,325 liters), but the actual amount collected was 171,793 gallons (650,307 liters), for a net under-prediction error of 28%. Figure 2 shows the results for each data reading period in terms of average condensate flow rate (gallons per minute) during the recording period, which makes for an easier comparison than just the total predicted or actually collected, as the time lengths between readings varied.


To check how consistent this error was through the cooling season and if it depended on the weather conditions, Figure 3 is provided which shows a scatter diagram of the ratio of actual to predicted total condensate collected versus the average outdoor air humidity ratio. Although the predicted values seem to be a little closer to the actual during periods of higher ambient humidity, this is not a strong trend and does not appear to provide any insight to the analysis.


Evaluation of potential sources of error - Supply airflow rate

The amount of predicted condensate during any given period is a direct linear function of the supply airflow rate used in the calculation, since the condensate collection predicted is based on the humidity ratio of mass of water per mass of dry air (lb/lb or kg/kg). Thus, if the actual flow were say 10% higher than used in the calculation, there would be an underestimation of condensate collection by 10% even if all other data were perfectly known and the equation was perfectly accurate and applicable to this situation.

This is illustrated by considering a case where the supply airflow used in the condensate prediction calculation is arbitrarily increased by 30%, representing a case where the actual airflow were 30% more than determined by the field measurements. The resulting total condensate predicted for this cooling season is 174,227 gallons (659,522 liters), or an error compared to the actual measure of about 1%. The plot of average condensate flow rate for each data recording period shows a fairly good match as well (Figure 4). While this 30% extra fan flow seems to be a possible explanation for the difference, it is not the only potential contributor to the error measurement. The 30% error in airflow measurement is also considered larger than what is generally considered acceptable in practice (more on this later).


Evaluation of potential sources of error - Relative humidity measurements

The accurate measurement of relative humidity has been an issue in the past within the HVAC industry. For example, earlier versions of humidity sensors used in economizer controllers had a propensity for early failure, leading to bypassing of the economizer control and giving a black eye to this concept for years.

The particular humidity sensors used for this study have a manufacturer's stated accuracy of [+ or -] 3.5% for the majority of the temperature range that they were used to measure. Since the calculation of condensate collection potential in this study involved a differential of humidity ratio between the incoming outdoor air and the supply air, there potentially could be anywhere between a -7% and +7% error (double one sensor) even if the sensors used were within the manufacturer's specifications.

Consider the extreme case of this and with the error in the proper direction to bring the predicted condensate level closer to the actual measured value. This would be if the case were that the actual relative humidity differential between incoming outdoor air and the supply air were 7% larger. A plot of the average condensate flow rates during each data recording period is given in Figure 5. The predicted annual condensate collected under this scenario would be 165,541 gallons (626.643 liters), for a 4% error in the prediction. While this scenario is possible, it relies on a 'best case' assumption of the error in relative humidity measurements.


Other scenarios were also evaluated with an assumed larger relative humidity differential of 1% and 3.5% as well. These results had corresponding levels of improvement in the condensate prediction.

Evaluation of potential sources of error - One realistic possibility

This scenario evaluated one case where the relative humidity and supply fan flow rate error values would be within that expected for a 'typical' case. For this evaluation, we evaluated the predicted condensate flow assuming the following errors in sensor readings or measurements.

* Relative humidity - assumed a 3.5% higher difference between outdoor air and supply air than we measured

* Supply airflow - assumed the supply air fan had a 15% higher flow rate

A 15% error in measurement of the supply fan airflow rate is a very reasonable estimate. For example, the U.S. Green Building Council's LEED-2009 program for IEQ Credit 1 considers a [+ or -] 15% differential in measured incoming outdoor airflow an acceptable value.

Figure 6 gives a comparison plot of the average condensate flow rates during each data recording period. The results compare very similarly to the actual measured values, and the total annual condensate predicted is 170,428 gallons (645,142 liters), for an under-prediction of only 1% compared to the actual measured value.


Evaluation of potential model simplifications - Assume constant supply airflow rate

This final section evaluates the impact of using more simplified model approaches. One of these is to use just an 'average' value for the supply fan flow rate. Since we were not directly measuring the supply flow, only the electrical current input to the fan, this will have to be approximated. Fortunately, the fan speed did not vary considerably during any given day or through the course of the cooling season. In fact, our baseline airflow measurement taken in the early spring was at a fan speed (based on current reading) considered very representative for the entire cooling season. This flow rate was measured at 19,128 cfm (9,027 liters/s), so this scenario evaluated the predicted collection of condensate assuming the supply airflow was a constant 19,128 cfm (9,027 liters/s). One other point to note is that there is an error introduced from measuring the current as well; our current transducers and datalogger together are accurate to [+ or -] 2.25 A and we were measuring current in the range of about 30 to 40 A

The resulting predicted annual condensate was 136,532 gallons (516,830 liters), essentially the same as the baseline case when a fan speed (using measured current) correction was applied. In fact, this number is slightly closer to the actual measured amount of condensate collected.

Evaluation of potential model simplifications - Assume constant supply air humidity ratio

Another possible simplification is to assume that the supply air humidity ratio is constant, and thus there would be no need to measure the supply air temperature and relative humidity. This is an important simplification that makes estimation of any AHU for condensate collection potential (retrofit or new installation) much easier.

For this scenario, we assumed that the supply air humidity ratio was that if the supply air temperature and relative humidity were 56[degrees] F and 85%, respectively. This corresponds to a humidity ratio of 0.008 [lb.sub.water]/[lb.sub.dryair] ([kg.sub.water]/[kg.sub.dryair]).

The resulting predicted annual condensate was 133,288 gallons (504,548 liters), again essentially the same as the baseline case using measured temperature and relative humidity of the supply and the corresponding humidity ratio for those conditions.

Other Sources of Error

Several other potential sources of error might exist, and one of these could include the sensor locations. For example, the outdoor air intake is shielded from direct sun, but there may still be some error from that since the intake is on the south side of the building. The supply air conditions were recorded with the data logger located in an easy to access section of the fan chamber; it was assumed the air is fairly well mixed but this may not have exactly been the case.


This study evaluated a model for predicting condensate collected from an AHU that conditions 100% outdoor air. A number of cases and possible scenarios for the impact of measurement error were studied to see if these alone could account for the under-prediction of condensate compared to the measured value during an entire cooling season. A summary of all the evaluation scenarios is given in Table 1.
Table 1 - Summary of Condensate Prediction Scenario Results

 Case Condensate Error, %
 or Predicted,
 gal (liters)

Actual Measured 171,793 (650,307) -

Baseline 134,021 (507,325) 28%

Airflow sensitivity

30% higher supply airflow 174,227 (659,522) 1%

Humidity measurement sensitivity

1% greater difference in RH 138,335 (523,654) 24%

3% greater difference in RH 149,551 (566,112) 15%

7% greater difference in RH 165,541 (626,643) 4%

Error attributable to measurement
accuracy only?

3.5% greater difference in RH and 170,428 (645,142) 1%
15% higher supply airflow

Potential simplifications (using baseline
case values for other parameters)

No fan speed correction for airflow 136,532 (516,830) 26%

Use 'typical' supply air humidity ratio 133,288 (504,548) 29%

Even though the baseline model using measured values for key parameter inputs such as outdoor and supply air temperature and relative humidity and an estimation of the supply airflow based on fan current draw underpredicted the condensate that would be collected, there are several potential scenarios that could explain this simply by error introduced by sensor (in)accuracy. One very real possibility was discussed with an assumed 3.5% error in relative humidity (the manufacturer's advertised accuracy) and 15% error in airflow.

Even if there were a 30% error in predicted condensate, this may be acceptable if the only answer really desired was if one should install or retrofit a condensation collection system or not. A 30% error in estimated condensate would result in 30% error of the potential cost of water savings or recovery, which may or may not be significant to the decision maker.

We also determined that two simplifications could be made to the prediction if one is only concerned with the total annual condensate collected. For this AHU, a constant supply fan flow could be assumed. This assumption may not apply to all AHUs across the board as it would depend on the variation in fan speed expected and how wide that variation is. The assumption of a constant supply air humidity ratio also can be reasonably assumed, where this should be based on the average supply air conditions expected.

Based on all these results, we conclude that our model for estimating the condensate collection potential for any AHU with 100% outdoor air is a relatively simple and valid approach. But what if the AHU is not 100% outdoor air (as most are not)? We feel the model is applicable there as well, with it being up to the engineer to determine or estimate the incoming outdoor airflow and variation in flow to use.

The purpose of this study was to validate our approach for estimating the potential application of condensate to a new or existing AHU and the amount of water expected. Before this study, we have done this using typical weather data (Marion and Urban 1995). Based on the fairly successful results of this study, we can safely recommend applying this model approach using 'typical' weather data to predict condensate collection during a 'typical' cooling season.


Guz. K. 2005. "Condensate Water Recovery", ASHRAE Journal 47(6):54-56.

Lawrence, T.M., J. Perry and P. Dempsey, 2010, "Making Every Drop Count: Retrofitting Condensate Collection on HVAC Air Handling Units", ASHRAE Journal 52(1):48-54.

Marion, W. and K. Urban. 1995. Users Manual for TMY2s. National Renewable Energy Laboratory, Golden, Colorado

Painter, F. 2009. "Condensate Harvesting from Large Dedicated Outside Air-Handling Units with Heat Recovery", ASHRAE Transactions 2009, 115(2):xxx

Wilson, A. 2008. "Alternative Water Sources: Supply-Side Solutions for Green Buildings", Environmental Building News, May 2008. Available from: [accessed December 2009].

T.M. Lawrence, Ph.D.


Jason Perry

Associate Member ASHRAE

Peter Dempsey

Student Member ASHRAE
COPYRIGHT 2010 American Society of Heating, Refrigerating, and Air-Conditioning Engineers, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2010 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Lawrence, T.M.; Perry, Jason; Dempsey, Peter
Publication:ASHRAE Transactions
Date:Jul 1, 2010
Previous Article:Transient and steady state models for open-loop air-based BIPV/T systems.
Next Article:Carbon neutrality based on native-site carbon storage.

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters