Abstract

Installed central air conditioning system capacities in 368 recently-built Florida homes were compared with Manual J load values calculated from house audit data. Over 50% of the homes had installed systems with a cooling capacity greater than 120% of the calculated Manual J value. Thirty-one percent of the 78 heat pump homes in this study had installed heating capacity greater than 120% of the calculated Manual J value. Using submetered data from the air conditioning system, a definite increase in peak energy use was observed for systems sized over 120% of Manual J calculations. The effect of system sizing on the system run-time fractions of units is shown. The homes that had units sized smaller than the calculated Manual J value tended to have higher percentages of maximum hourly run-time fractions. There was little difference in binned percentages of large runtime fractions between homes with installed-to-calculated ratios of 1.0 to 1.2 and those with installed-to-calculated ratios greater than 1.2.

Keywords: Air conditioning, sizing, load calculation, residential, monitoring, energy, heat pump

Introduction

Accurate sizing of residential HVAC equipment is an important goal that benefits both homeowners and utilities. Proper sizing will result in equipment large enough to meet loads during peak periods, but small enough to be energy efficient.

An oversized system will require an increased initial investment by the homeowner. Also, occupant comfort may be reduced if air conditioner run-times are shortened and less moisture is removed from the air (Khattar et. al., 1987). Short-cycling systems will also often run at lower efficiency than correctly sized systems and may increase utility peak load, especially on hot summer afternoons (Henderson, 1992). Recent studies have provided increasing evidence of increased peak demand due to oversizing (Reddy and Claridge, 1993, and Neal and O'Neal, 1994). Submetered air conditioner data from ten Florida homes also suggests increased machine size will result in greater utility coincident demand (Parker, et. al., 1996). Oversized heat pumps can also lead to inefficiencies in the heating season due to poor operating efficiency at start-up. Proper equipment sizing is also complicated by the proliferation of simplified sizing methods which may lead to inaccurate results (Proctor et. al., 1995).

In an effort to determine how closely HVAC equipment installation is adhering to recommended sizing practices, a study was conducted of installed HVAC equipment size in Florida homes built in 1990 or later.

Data Collection

House Selection

Five Florida counties were pre-selected for the study that also encompassed many more factors than HVAC equipment sizing. These counties were chosen in part because of their projected growth of new residential construction. There was some attempt to obtain a larger than typical fraction of energy-efficient homes within the sample for each county. This selection strategy was implemented to evaluate the accuracy of Florida's energy code (State of Florida, 1993) and efficient technology upgrades.

On-site Audits

A house audit was performed including a homeowner questionnaire, an on-site indoor inspection, and an outdoor inspection. A plan of the house was drawn to scale. In instances where construction drawings were available, the drawings were verified against the "as-built" house. The questionnaire consisted of approximately 35 customer lifestyle questions that the auditor asked each homeowner. Questions concerned occupant behavior on air conditioning, heating, and ventilation control habits, as well as homeowner modifications since occupancy. The indoor inspection was used to document indoor equipment and appliances in the homes. The outdoor inspection section included information about outdoor air conditioning equipment and ductwork, house envelope components such as walls, windows, doors, ceilings and roofs. An attic inspection was made to check the insulation installation quality. Duct work and home air tightness were pressure tested. Extensive window data were recorded that included window areas, orientations, exterior shading, window coverings, tinting and frame types.

House drawings included a floor plan, wall and ceiling dimensions, room descriptions, major equipment locations and house orientation. Conditioned floor area was carefully documented and checked. Each elevation and any anomalies were photographed.

Collected house data was entered into a database for analysis. Quality control included checking the validity of original field data and verifying database entries. Complete verifications of key fields and spot verifications of other entries were also made. Selected average house characteristics from the audited database are given in Table 1.

Table 1
Characteristics of Audited Homes

Metered Data

Using a data logger, fifteen minute electrical energy use was recorded at each site. The submetered data consisted of total kWh, space cooling kWh, heating and air handling, and a channel for either electric water heating, a refrigerator, pool pump or interior temperature measurement.

Use of Manual J as the Theoretical Sizing Method

The Air Conditioning Contractors of America's Manual J (ACCA, 1986) load calculation procedure was used as the sizing standard. A survey of residential contractors in Florida previously conducted by the Florida Solar Energy Center found that 33 % of responding contractors use Manual J to accomplish sizing in new residential construction (Vieira, et. al., 1995). Also, an additional 34% of responding contractors use a software application to accomplish sizing calculations. Over half of these contractors used popular software products that are replications of the Manual J procedure.

Development of The Manual J Spreadsheet

A spreadsheet was developed to produce Manual J calculations for each of the research homes. The Manual J procedure was followed as written in the ACCA Manual. Linear interpolation of Manual J Heat Transfer Multipliers was used for all insulation values. The spreadsheet was verified against manual calculations.

Building window, wall, door, ceiling and floor areas were determined from on site measurements and construction drawings. Because actual building leakage is rarely known when an HVAC system is sized, all homes were assumed to experience an infiltration level halfway between Manual J's "Best" and "Average" infiltration specifications.

Selection of Manual J Design Temperature Differences

The results of the Manual J calculations depend heavily on the selected design temperature difference and to a lesser extent on local humidity levels. The design temperature difference for cooling is the result of subtracting the desired inside set temperature from the outside design temperature, obtained from weather data produced by the American Society of Heating, Refrigerating and Air Conditioning Engineers or from other reliable sources. Similarly, the design temperature difference for heating is the result of subtracting the outside design temperature from the inside temperature. Manual J recommends an indoor temperature of 70 ^oF for heat loss calculations and a temperature of 75 ^oF for heat gain calculations "unless the owner, builder or codes specify otherwise" (ACCA, 1986). The Manual J calculations require a specified design temperature difference for heating and for cooling for each location. A geographically representative range was selected for each Florida region.

Table 2
Chosen Design Temperature Differences

Design temperature differences shown in Table 2 were chosen for this study. These temperatures were obtained using geographically specific design temperatures obtained from the ACCA Manual J and interior temperatures of 75 ^oF for cooling and 70 ^oF for heating. The resulting temperature difference for each location was then rounded up to the nearest 5 ^oF. This simplifies the Manual J calculation by eliminating the interpolation of temperature values as many contractors may do; however, rounding up the temperature values also adds some error that will usually result in slightly higher Manual J values.

In order to assess the sensitivity of rounding the design temperature difference in the Manual J calculation, 24 houses were randomly selected from the larger study and evaluated at different design temperature differences. An increase of the design temperature difference by 5 ^oF from 15 ^oF to 20 ^oF results in a typical increase of the Manual J total summer load calculation by 11.8% (standard deviation of 2.1%). This increase will vary somewhat based on location and house characteristics. For winter load calculations, an increase of the design temperature difference by 5 ^oF from 30 ^oF to 35 ^oF results in a typical increase in the Manual J heating load of 16.4% (standard deviation of 0.34%). When the 5 ^oF increase of the design temperature difference is from 20 ^oF to 25 ^oF, the Manual J heating load results in a typical increase in of 23.9% (standard deviation of 0.62%).

The basic Manual J calculations in this study reflect a measure of oversizing due to rounding. Winter design temperature differences were rounded up by 0 ^oF to 3 ^oF depending on the location, resulting in an estimated increase in the basic load calculation of 0% to 9.8%. Summer design temperature differences were rounded up by 3 ^oF to 4 ^oF depending on the location, resulting in estimated increase in cooling load calculations of 7.1% to 9.4%. Therefore, figures presented in this study for actual equipment oversizing are conservative.

Sensitivity of Manual J To Interior Window Shading

Interior window shading will produce marked changes in heat gain through windows. In new construction, however, cooling equipment must often be selected before the type and extent of interior window shading is known. Manual J calculations were done on 389 of the 423 audited homes in this study using the actual interior window shading existing in the house (the majority of the homes consisted of interior blinds).⁽²⁾

In order to assess the sensitivity to interior window shading, the calculations were then repeated using no interior shading on the windows to produce maximum heat gain. Total Manual J heat gain loads for these homes increased by an average of 12% when interior window shading was eliminated and ranged from 0 to 28%. The distribution is shown in Figure 1. Obviously, for homes where window interior shading is unknown or assumed, heat gain predicted by Manual J calculations may differ from the heat gain of the completed house.⁽³⁾

Figure 1. Distribution of Manual J load increases when interior
window shading is ignored.

Results

Comparison Results for Cooling Equipment

Manual J equipment sizing calculations were compared to the size of the equipment actually installed in each house. Homes having more than one air conditioning system were omitted from this analysis because the sum of the system capacities may be oversized without each individual system being oversized. Manual J recommends that the "cooling only" equipment should be sized 0% to 15% greater than the Manual J sensible heat gain value, and that latent equipment capacity should at least be as great as the Manual J latent load. Based on these recommendations, houses with installed systems greater than 120% of the calculated Manual J cooling loads will be referred to as "oversized".

Figure 2 shows the comparison of Manual J summer loads to the installed equipment size for 368 Florida homes.

Figure 2. Florida installed cooling equipment size vs. calculated Manual J load.

The dotted line diagonally intersecting the graph indicates where the data would fall if every installed cooling system was sized exactly according to Manual J specifications; the solid line indicates 120% of Manual J. A large majority of these homes have cooling systems that are greater than the Manual J calculation. The cooling systems in this study average 23% larger capacity than Manual J calculations as depicted in Table 3. Table 3 also shows that 53% of the cooling systems in the sample are greater than 120% of Manual J total cooling load.

Table 3
Cooling System Specifications

The sizing analysis for homes conditioned by heat pumps is unique because the size of the single heat pump unit affects both the cooling and heating capacity. In homes in Central Florida, where heating requirements exceed cooling requirements, it is possible that oversized cooling capacity may result if the heat pump is being sized to meet the entire heating load. Figure 2 compares the 78 heat pump houses in this study to the 290 houses without heat pumps. The heat pump houses averaged a cooling capacity 26 % larger than the Manual J load calculation, while non-heat pump houses averaged 22% larger than Manual J. This difference of 4% (90% confidence interval of 0.2% to 8.1%) indicates that the presence of heat pumps may have contributed slightly to cooling system oversizing.

Additional reasons for contractors choosing to oversize have been obtained in a state-funded research study (Vieira, et. al., 1995). Contractors consider it less time consuming in terms of calculations and potential call-backs to simply install large systems than to size correctly, or to predict which customers will desire extraordinarily low cooling thermostat settings.

Comparison Results for Heating Equipment

Figure 3 shows the relationship of Manual J winter loads compared to the installed heat pump size for 78 Florida homes. Again, the dotted line diagonally intersecting the graph indicates where the data would fall if every installed system was sized exactly according to Manual J specifications and the solid line shows 120% of Manual J. Average heat pump system size for the 78 homes was 34,386 Btu/hr which is 14% greater than Manual J calculations. Twenty five of the homes, or 32%, are sized over 120% of Manual J.

Figure 3. Central Florida installed heating equipment (heat pumps)
vs. calculated Manual J load.

Homes with electric resistance furnaces are installed in 5 kW increments. No attempt was made to compare their sizing against Manual J calculations due to lack of data and the installation increments used.

Impact of Sizing on Summer Peak Energy Use

Between 4 and 5 PM EDT on June 24th, 1994, the utility experienced their summer system peak. The summer peak coincided with a hot and sunny day during the early summer of 1994. The temperature measured at project weather stations ranged from 90^oF to 94^oF with nearly clear conditions. Although the maximum air temperature was reached around 1 PM, the peak cooling demand occurs five hours later when occupants returned home. Figure 4 shows the effect of oversized units on the hourly cooling load on June 24, 1994.

Figure 4. Effect of equipment oversizing on hourly energy use during utility's peak demand day.
Houses that have oversized systems tend to have increased afternoon cooling demand.

The homes with oversized systems (>120% of Manual J) average about 13% (0.3 kWh) greater electrical load for the cooling system than homes without oversized systems. T-tests of the data indicated there was a difference in this peak hour energy use as well as peak hour energy use per square foot at the 99% significance level between homes with and without oversized systems.

Summer Air Conditioning Use

Multivariate analysis of summer air conditioning use consistently showed an energy penalty associated with oversizing. For the cooling season, a coefficient of 872 kWh and a t-value of 2.0 was associated with the sizing ratio (installed/Manual J) using a regression model with an adjusted R-square of 0.37 on 308 of the 423 homes. The other homes were eliminated due to small portions of missing data or due to non-occupancy during the summer. If the 872 kWh estimate is used, on average a home with a system sized 20% greater than Manual J would have 174 kWh more cooling energy use, and one sized 50% greater would have 435 kWh more energy use than a unit installed at the size calculated using Manual J. These values correlate to 3.7%, and 9.3% of the summer cooling load for the sample of homes used in the regression. These results are consistent with results from other researchers (Henderson,1992)(Lucas,1993). T-tests of data separated into two groups, one with installed to calculated cooling capacity ratios greater than 1.2, and one less than or equal to 1.2, indicated there was a significant difference in summer energy use and summer energy use per square foot at the 99% significance level.

Impact on AC Runtime Fractions

The on and off times of individual air conditioners was not measured in this study. The 15-minute cooling system electricity consumption was metered. To estimate run times, the peak 15 minute cooling electricity consumption was determined for each house for each month of the study. This peak was then compared to the next highest nine values to assure that it was not an anomaly or error in the data. This 15-minute value was multiplied by 4 to determine an estimated peak hourly load when the air conditioner was running for the full hour. All metered data was then summed hourly, compared to the peak hourly load to determine run time fractions, and placed in bins.

Calculated air conditioner run time percentages for June, July and August are shown in Figure 5. The plots show percent of total monthly hours the air conditioners ran at given percentages of the full load for three different ranges of installed-to-Manual J calculated sizes. Note that there is little difference shown for 90% to 100% runtime fractions between units sized more than 20% greater than the calculated Manual J value and those sized 1.0 to 1.2 greater than the Manual J value. For the months of June, July and August, 1994, homes sized at or below Manual J had run-times in the 90% to 100% range 8.45%, 14.19% and 10.04% of the time, respectively, on average. These run-time percentages are 30% to 50% greater than for homes with systems sized at or above Manual J. Interior temperatures were only measured on a few homes so it is not possible to correlate homeowner thermostat behavior with this data, or to accurately determine how much of the time the thermostat setting was not obtained. Nevertheless, it appears likely that the homes with undersized systems had a higher percentage of hours where the load was not met than the other homes.

Figure 5a. Average binned air conditioner run-times by ratio of installed air conditioner
capacity to calculated Manual J load for June 1994.

Figure 5b. Average binned air conditioner run-times by ratio of installed air conditioner capacity to
calculated Manual J load for July 1994.

Discussion

The study is significant in that there are a large number of houses that have pertinent information on the subject of equipment sizing: a Manual J calculation, installed equipment capacity, metered electrical consumption of the system, and numerous fields of data on the occupants and home. Based on the results presented here, as well as in the references, the ideal cooling system capacity for use in homes, at least on average, appears to be close to Manual J. Our recommended sizing procedure for homes in humid climates that will use a single one-speed system for cooling is:

There are many circumstances that are more complicated. Homes with heat pumps may have heating loads that exceed cooling loads (many in this study had higher heating loads despite Florida's mild climate), so a compromise needs to be reached based in part on the length of the cooling season versus the heating season. Homes that have two or more systems serving one area add a degree of complication also. Finally, systems that are multi-speed have the potential to run efficiently at low speed most of the time even if the higher speed capacity chosen is larger than a calculated sizing procedure would indicate.

Conclusions

A number of conclusions can be made from the research presented:

The authors would like to thank Kris Bradley of Quantum Consulting who managed the collection of metered data, and Kashif Hannani, Safvat Kalaghchy, Maria Mazzara and Danny Parker of the Florida Solar Energy Center who helped develop the database and conducted the statistical analysis.

References

Air Conditioning Contractors of America: Manual J - Load Calculation for Residential Winter and Summer Air Conditioning, Seventh Addition, 1986.

Henderson, H.I. Jr., "Simulating Combined Thermostat, Air Conditioner and Building Performance in a House," ASHRAE Transactions, Volume 98, Part 1, 1992.

Khattar, M.K., M.V. Swami and N. Ramanan, "Another Aspect of Duty Cycling: Effects on Indoor Humidity," FSEC-PF-118-87, Florida Solar Energy Center, January, 1987.

Lucas, R.G.,"Analysis of Historical Residential Air-Conditioning Equipment Sizing Using monitored Data,"Report to DOE by Pacific Northwest Laboratory, PNL-8542,1993.

Neal, Leon and Dennis O'Neal, "The Impact of Residential Air Conditioner Charging and Sizing on Peak Demand," ACEEE Summer Study on Energy Efficiency in Buildings, Vol. 2, pp. 189-200, 1994.

Parker, D.S, S.F. Barkaszi, Jr., J.R. Sherwin, and C.S. Richardson, "Central Air Conditioner Usage Patterns in Low-Income Housing in a Hot and Humid Climate: Influences on Energy Use and Peak Demand," ACEEE Summer Study on Energy Efficiency in Buildings, Vol. 8, pp. 147, 1996.

Proctor, J., Z. Katsnelson and B. Wilson, "Bigger Is Not Better - Sizing Air Conditioners Properly", Home Energy, Volume 12, Number 3, May/June 1995

Reddy, T.A., and D.E. Claridge, "Effect of Air-Conditioner Oversizing and Control on Electric-Peak Loads in Residences," Energy, Vol. 11, pp. 1139-1152, 1993.

State of Florida Department of Community Affairs, 1993 Energy Efficiency Code For Building Construction, Tallahassee, Florida, 1993.

Vieira, R.K., J. Klongerbo, J.K. Sonne and J.E. Cummings, "Florida Residential Air Conditioning Sizing Survey Results," ASHRAE Transactions 1995, V.101, Part 2.

1. Patrick James is residential research program manager at Florida Power and Light, 9250 W. Flagler St., Miami, FL 33174. Jo Ellen Cummings and Jon Klongerbo are research assistants, Jeffrey K. Sonne is a research engineer, and Robin K. Vieira is a principal research analyst at the Florida Solar Energy Center, 1679 Clearlake Road, Cocoa, FL 32922.

2. The remaining 34 houses lacked window-covering data.

3. All Manual J heat gain calculations referred to in the remainder of this study were done using actual installed interior window shading.