Short-term curtailment of HVAC loads in buildings.

INTRODUCTION
With the rapidly increasing demands placed on utilities, exacerbated by the increasing cost of
generating energy, reducing peak loads and minimizing energy use assume greater importance than
ever before. It is well documented that building sub-systems (e.g. HVAC, lighting, equipment) can
use as much as 40% (PNNL, 2008) of the nation's total energy resources. To combat these
increasing demands, utilities offer incentives to customers that can shift peak demand to non-peak
times or reduce peak loads when notified by the utility that the demand on the grid is close to
capacity. Utilities have developed the concept of critical peak pricing (CPP) (PG&E 2007, SCE
2007a, and SCE 2007b) and real time pricing (RTP) (Gabel et al. 1998) wherein the utility provides
customers with a pricing structure typically one day in advance. Then the utility offers a monetary
reward to customers who can shift or reduce their electric demand from the grid during peak
hours.
Building owners and operators have responded to time-of-day pricing by resorting to traditional
methods, such as load-shedding (i.e. by turning off non-essential components that use electricity
during peak hours), or by using non-traditional methods, such as Demand Response (DR) as
documented by various publications including Motegi et al (2007). Demand response consists of
demand control algorithms that, along with a host of other strategies (e.g. lighting load
control) reduce peak load or shift demand and use the thermal mass of the building (e.g., HVAC
load control) to achieve these demand reductions during peak hours (Xu, 2009). The building can be
pre-cooled during non-peak hours using electricity at lower than normal prices, which allows for an
increase of thermal set-points (i.e. zones and/or HVAC components) to reduce cooling demands
during peak hours. The building's normal operation can be restored when the grid demand
is sufficiently reduced. Lee and Braun (2006a and 2006b) have conducted research to assess what
they term 'demand limiting' by using the thermal mass for small commercial buildings. In an
experimental setting and in some parametric simulations of zone set-point manipulation before and
during peak hours, the use of three thermostat control profiles in conjunction with pre-conditioning
the building thermal mass lead to a reduction in peak cooling loads of approximately 20% to 40%.
Research by Xing (2004) and Reddy and Norford (2002) in the area of building load control and
optimization for demand limiting and peak load shedding present some preliminary results of
simulation modeling in which thermostat and economizer setpoints are varied for a simple VAV
system and a three-zone building model. This review of the literature reveals that little work has
been done to develop, test and implement effective load-reduction strategies when the
utility notification is not provided a day ahead but rather on the same day. Computational
simulations are needed to provide a direction for further experiments since computer simulation
enables a more comprehensive consideration of various control strategies and the influence
of building type, climate, etc.
In order to develop effective control strategies to reduce peakdemand in response to short
notification (only a few hours advancednotice) of critical peak rates and time-of-use charges from
utilities,computational simulations were performed to examine the effectiveness ofvarious control
strategies under the ASHRAE research project, 1390-RP.The objective of 1390-RP was to identify
and evaluate peak demandreduction strategies formulated in response to short notification ofcritical
peak rates and time-of-use charges from utilities, for use incommercial buildings. The term "demand
reduction" is usedthroughout this paper to represent the aver-age electrical energysavings during
the peak demand window. The short notification time maybe as short as 15 to 30 minutes, and as

long as a few hours. Everincreasing pressure on utilities has resulted in pricing forecasts thatare on
the order of hours instead of days, and previously researchedstrategies like pre-cooling of building
thermal mass will not be asuseful in reducing peak load when the demand on the grid is at
itshighest. To overcome this limitation, strategies specific to short-termelectrical price forecasts
have been investigated and identified. Thepeak demand window was chosen to be 5-9 A.M. in winter
and 2-5 P.M. insummer.
BUILDING AND HVAC SYSTEM SPECIFICATIONS
The buildings selected for this study were obtained from the computer simulation reference
buildings provided by the United States Department of Energy (DOE) (Deru et al. 2009). This study
included small, medium, and large representations of the office and retail building types. The
building characteristics were selected to meet the minimum requirements for those building types
according to ASHRAE Standards 90.1-2004 and 62.1-2004. The buildings have cooling setpoint at
24[degrees]C (75.2[degrees]F) and heating setpoint at 21[degrees]C (69.8[degrees]F) during
occupied hours. Table 1 summarizes the selected building types and associated HVAC systems. The
table also provides building floor areas, cooling and heating types and efficiencies, and fan control
modes for each HVAC system.
Table 1. Selected HVAC System Types and Associated Building Types Office Retail Building
Type Small Medium Large hvac training sacramento ca Small Medium Area 511 4,982 46,320
46,320 174 ([m.sup.2]) ([ft.sup.2]) 5,500 53,626 498,588 498,588 1,873 HVAC PSZ MZ-VAV MZVAV Dual Duct PSZ Cooling Type DX DX Chilled Chilled DX water water COP 3 3 4.45 4.45
3 Heating Type Gas Gas Hot water Hot water Gas furnace furnace furnace Eff. (%) 80 80 80 80
80 Fan control Constant Variable Variable Constant Constant Building Type Large Area 348
2294 2294 ([m.sup.2]) ([ft.sup.2]) 3,745 24,693 24,693 HVAC PSZ MZ-VAV Dual Duct Cooling
Type DX Chilled Chilled water water COP 3 4.45 4.45 Heating Type Gas Hot water Hot water
furnace Eff. (%) 80 80 80 Fan control Constant Variable Constant Note: PSZ - packaged single
zone system MZ-VAV - multizone variable air volume fan system Dual Duct - constant volume dual
duct system DX - direct expansion refrigeration system
Two types of building constructions were chosen to represent lightand heavy thermal mass buildings
to examine the impact that thermal masswould have on a building's peak demand reduction
potential. Fivegeographical regions were chosen to study climate specific variations inthe results.
The cities selected and corresponding climate types are:Miami - hot and humid, Baltimore - mixed
humid, Albuquerque - mixed dry,Phoenix - hot dry, and Minneapolis - cold.
SIMULATION PROTOCOL
The development of simulation protocol consists of selecting control strategies and developing the
simulation approach, including selection of prototype days.
Selected Control Strategies
Many control strategies were identified to reduce building peak demand. After careful selections
and suggestions from the Project Monitoring Subcommittee (PMS) (Gu et al. 2011), four control
strategies were selected and investigated thoroughly in the ASHRAE project: lighting power
density reduction, global thermostat setpoint setback control, chilled water temperature reset, and
supply air temperature adjustment. The last two control strategies are applied only to central plant
systems used for large office and large retail buildings and are more practical for use with building
automation systems.

Simulation Approach
Over 30,000 EnergyPlus computer simulations were performed. These simulations reported building
average energy savings potential over the peak demand window by building type, geographical
location, and day type (e.g., summer peak, winter peak, etc.) for individual control strategies as
well as combinations of strategies where two or more individual strategies were used. To reduce the
time required for the simulations, prototype days were selected to represent typical working days
in different seasons. Choosing these prototype days and simulating only a 10 consecutive day
period for each prototype day, instead of a full annual simulation, dramatically reduced the time
required to compute the results. For each simulation, the last day in the 10 consecutive day period
provides the results for the specific prototype day. The selected prototype days are Summer Peak,
Summer Mid, Summer Low, Fall Cool High, Winter Peak, Winter Mid, Winter Low, and Fall Heat
High (Table 2).
Table 2. Prototype Day Description Summer Peak A day to present a summer day with peak
cooling energy use during peak demand period in a season Summer Mid A day to present a
summer day with average cooling energy use during peak demand period in a season Summer
Low A day to present a summer day with low cooling energy use during peak demand period in a
season Fall Cool High A day to present a fall day with peak cooling energy use during peak
demand period in a season Winter Peak A day to present a winter day with peak cooling energy
use during peak demand period in a season Winter Mid A day to present a winter day with average
heating energy use during peak demand period in a season Winter Low A day to present a winter
day with low heating energy use during peak demand period in a season Fall Heat High A day to
present a fall day with peak heating energy use during peak demand period in a season
Eleven different thermostat temperature profiles were used torepresent different approaches, or
control profiles, for controlling theHVAC system. Linear, exponential, and negative exponential
equationswere used to calculate the eleven different thermostat temperatureprofiles. Each profile
represents a thermostat schedule based on inputsto the mathematical model. In the same manner,
we also developedsetpoint profiles of chilled water temperature reset, and supply airtemperature
adjustment. The detailed descriptions are provided in thefinal report (Gu et al. 2011).
When the demand window closed, the thermostat schedule was gradually restored to the original
thermostat setpoint temperature in a linear manner over a 1-hour time period for the global
thermostat setpoint setback control. This recovery strategy was used to minimize the spike in
building demand that would occur if the thermostat setpoint temperature was immediately returned
to the setpoint that was in use prior to the implementation of the demand reduction strategy. For
the chilled water temperature reset and supply air temperature adjustment control, both
temperatures were gradually restored to the original setpoint temperatures in the same manner as
the global thermostat control. For the lighting power reduction control, the lighting power density
was set to the original value used in the base case.
SIMULATION RESULTS ANALYSIS
This section analyzes the impact of thermal mass, four individual control strategies and
combinations of control strategies using simulation results. The relationship between individual
control strategies and combined control strategies are discussed. Thermal comfort was also
simulated to investigate the impact of the control strategies on occupant comfort. The detailed
analysis may be found in the project final report (Gu et al. 2011).
Thermal Mass Impact

Simulation results show that the impact of the building thermal mass on the building peak demand
is relatively small for the building types selected for this study. In this study it was assumed that
utility notification of a demand reduction event would occur only a few hours in advance and
insufficient time was available to precondition the building. The relatively small average energy
savings over the peak demand window associated with thermal mass were found to be insignificant
(<2%) compared to the average energy savings obtained from other control strategies investigated
during the course of this project. Therefore, the impacts of thermal mass on peak demand
reduction need not be considered a significant contributor to savings found for control strategies
investigated during this project.
Lighting Power Density Reduction
The lighting power density (LPD) reduction strategy is an effective way to meet a peak demand
reduction requirement. This strategy is easy to implement by simply turning off nonessential lights.
Figure 1 shows the percent peak reductions of facility electricity averaged over all locations and
prototypes days. Lighting power density varied between 70% and 90% of the base case value during
the peak demand windows. The line drawings are based on 3 discrete points at 70%, 80% and 90%.
Since the percent peak demand reductions vary linearly with the percent changes of lighting power
density, it is very convenient for building operators or control engineers to decide how much
lighting power density should be reduced given a specific peak demand reduction target. It should
be pointed out hvac sacramento parts that any LPD reduction will increase heating energy use
in winter and reduce cooling energy use in summer.
[FIGURE 1 OMITTED]
The LPD reduction may be achieved by a building automation system or manual switches. Multiple
stages of LPD reductions are recommended to smooth out the changes in building energy use
during the peak demand window.
Global Thermostat Temperature Setpoint Setback
The use of a global thermostat temperature (TST) setpoint setback schedule achieved average
energy savings of 40% for small buildings, 30% for medium buildings and 20% for large buildings
over the peak demand windows in this study using a maximum thermostat setback
temperature differential. Office buildings generally showed 5% - 10% less savings than found for
retail building types. These savings are dependent on geographical location and HVAC system
types. For this project, the maximum thermostat setback temperature differential is
3.3[drgrees]C (6[drgrees]F). Using this control strategy it is possible to achieve these savings even
for the simplest of HVAC systems (e.g., PSZ and packaged MZ VAV). This strategy applies to those
buildings using a zone thermostat (analog or digital) and to those using more complex
building automation systems. This study also found that the global thermostat strategy had less of
an impact on buildings served by a central plant using dual duct systems since this system type
would respond to a zone setpoint temperature increase by reducing the cold deck air flow rate. This
in turn would increase the hot deck air flow rate for this constant volume system and increase
heating energy required to maintain the hot deck supply air temperature setpoint.
The selection of the maximum setback differential will impact the degree to which occupant comfort
is sacrificed during the peak demand window. Since the shape of the setback curve can impact the
average energy savings over the peak demand window, it is a good practice to test different shapes
to maximize savings.

Chilled Water Temperature Reset

Resetting the chilled water control temperature (CWT) for chilled water plant loops without
addressing pump speed control can reduce building peak demand by up to 7%. For this project, the
chilled water temperature (6.7[drgrees]C [44.1[drgrees]F] for the baseline) was increased by a
maximum of 5[drgrees]C (9[drgrees]F). Buildings with dual duct systems achieved higher percent
reductions than buildings with VAV systems. Although an increase in chilled water temperature will
improve the efficiency of chillers, the increase in pump and fan energy could offset these savings in
certain instances. For this reason, additional simulations were performed to limit the chilled water
pump flow rate to be no greater than the flow rate at the beginning of the peak demand window.
When including pump speed control, savings up to 15% can be achieved. Therefore, it is important
to prevent the chilled water flow rate from increasing at the same time when chilled water
temperature is reset.
Selection of the maximum setback differential should take into consideration the impact it will have
on dehumidification capacity during the peak demand window. Pump speed control can be
extremely important for VAV systems. The pump speed was limited to the pump flow rate simulated
just prior to the start of the peak demand window.
Supply Air Temperature Adjustment
Supply air temperature (SAT) adjustment does have the potential to reduce building peak demand,
but only for specific HVAC system types. For this project, the maximum reset supply air
temperature differential is 5oC (9oF). For dual duct systems, results showed up to a 16% savings in
building demand for cooling operation. For VAV systems, results generally showed up to a 5%
savings, however, building peak demand could actually increase up to 10% for the Summer-Peak or
SummerMid day types for certain climate locations. VAV systems will increase fan speed
in response to an increase in supply air temperature which can increase energy use. For this
reason, additional simulations were performed to limit the VAV fan flow rate to be no greater than
the flow rate at the beginning of the peak demand window. When implementing a fan speed control
strategy for VAV systems, building peak demand reductions up to 12% were possible.
Dual duct systems mix the cool and hot air upstream of the individual zones before entering the
spaces. The sum of the cold and hot deck air stream flow rates is constant. When the cold deck
supply air temperature is increased, the cold deck air flow rate also increases to meet the same
load. As a result, the hot deck air flow rate decreases and energy savings is primarily due to a
reduction in the hot deck heating requirement. Since this is a constant volume system, fan
speed control is not applicable.
The selection of the maximum setback differential will impact the dehumidification capacity of the
system during the peak demand window. Fan speed control can be extremely important for VAV
systems. The fan speed was limited to the fan flow rate simulated just prior to the start of the peak
demand window. In practice, fan speed regulation is strongly recommended when resetting the
supply air temperature setpoint.
Combined Control Strategies
As previously described, demand savings due to each individual strategy provided a reasonable
savings in building peak demand for specific building and HVAC system types. In general, a single
control strategy did not provide the maximum possible savings and various combinations of these
strategies were investigated to determine how these control strategies worked in combination.

Computer simulations also showed that combining thermostat reset strategies with chilled water or
supply air temperature reset strategies did not provide savings equal to the sum of the savings for
the individual strategies. In other words, relationships of peak demand reductions between
individual strategies and the corresponding combined strategy are not linear. This result applies
whether or not a speed control strategy is used since a reset in thermostat temperature reduces
zone loads, and therefore the required supply air or chilled water flow rate, which eliminated
savings due to the water or air reset strategies.

The combined control strategies of lighting power density

reduction and global thermostat setpoint setback control
can be implemented in small office, small retail, medium
office and medium retail buildings. Control strategies
combining two or more of the individual
control strategies of lighting power density reduction,
global thermostat setpoint setback control, chilled water
temperature reset, and supply air temperature
adjustment can be implemented in large office and
retail buildings.
The percent reductions of electrical demand savings
during peak demand periods in a summer peak day using
combined control strategies (i.e., utilizing all available
strategies mentioned above) are shown in Figure 2. The
savings with light and heavy mass constructions are
also provided in the figure. The peak demand reductions
vary between 10% and 40%. Figure 2 also indicates the insignificant impact of thermal mass
by comparing demand reductions between light and heavy construction types.
[FIGURE 2 OMITTED]
Figure 3 presents averaged percent reductions using all available control strategies with maximum
savings for each individual control strategy for different building and HVAC system types. The
values are averaged over all prototype days and geographical locations. The difference between
Figures 2 and 3 is that Fig. 2 presents average results from a summer peak day, while Fig. 3
presents average results from all prototype days. Although the same combined control
strategies are applied in both Figures 2 and 3, different setpoint profiles may have been selected
for different building and HVAC system types. The selection criteria chose specific setpoint profiles
based on the maximum savings. These specific setpoint profiles were then combined to simulate the
overall savings.
[FIGURE 3 OMITTED]
To compare the impact different control strategies had on peak electricity use, the electricity use
for individual and combined control strategies in a large office building for a summer peak cooling
day in Miami are presented in Figure 4. The selected cases are mentioned in the legend using the
previously defined abbreviations. The "+" sign indicates combined control strategies. It is observed
that the maximum saving occurs in the case with combined strategies (LPD+TST+SAT+CWT,
lowest line in the demand window). For each individual control strategy, the performance order
with the best first is CWT, SAT, LPD and TST (upper most four lines below the base case). For the

cases with LPD plus one of the other 3 control strategies, the performance order with the best first
is LPD+CWT, LPD+SAT and LPD+TST.
[FIGURE 4 OMITTED]
Relationship Between Individual Control Strategies and Combined Control Strategies
In order to reach the maximum peak demand reductions, it was essential to combine different
control strategy types. Since there are many possible combinations, only a select few variations
were simulated. The selected combined control strategies were based on the best performance of
each individual control strategy. If savings from individual strategies were found to be additive,
percent reductions from combined control strategies could be derived from the reductions of single
measures. After performing a statistical analysis for individual control strategies and combinations
of these strategies, the results for combined control strategies may be calculated from the
individual savings found for small and medium office building types where only lighting and
thermostat control strategies were used. However, for large office and retail building types, the
savings for combinations of control strategies should not be calculated using the results
for individual strategies (i.e., individual results are not additive when thermostat reset is combined
with air or water reset strategies).
Thermal Comfort Check
Understanding that the use of thermostat setback can adversely affect occupant comfort, a
summary of simulated occupant comfort for the same summer demand simulations presented in
Figure 2 was also compiled and is shown in Figure 5. For this study, a setback
temperature difference of 3.3 [degrees]C (6 [degrees]F) was used to limit the maximum possible
offset from the original thermostat temperature schedule. The simulation results from the summer
peak day are selected to present thermal comfort check, because the summer peak day with
the setback control strategies may represent possible severe scenarios.
[FIGURE 5 OMITTED]
The Fanger comfort model is used to describe occupant thermal comfort as a people-weighted
average for the entire building. ASHRAE Standard 55-2010 specifies the acceptable predicted mean
vote (PMV) range is between -0.5 and +0.5. Figure 5 shows that the comfort levels for the Summer
Peak light construction baseline simulations are in the range of -0.5 to 0.1 for all building types
(each "Baseline" result in the figure). This range is between a slightly cool (slightly cool = -1)
perception to just barely warmer than a neutral perception (neutral = 0) of how occupants perceive
the indoor environment. Comparing the baseline thermal comfort to that simulated when all control
strategies were combined shows an increase in the PMV value towards the warm comfort region
(i.e., comparing the "Baseline" results to the "All" results for each building type). The Fanger PMV
values range from -0.3 to 0.6 for the combined simulations (each "* - All" result in the figure) and is
primarily due to higher indoor temperatures. Since these values do not exceed the slightly warm
criteria (slightly warm = 1), and only a few are slightly higher than the maximum value specified by
ASHRAE Standard 55-2010, it is assumed that the control strategies selected for this study are
considered feasible for use when implemented as building demand reductions strategies. However,
note that these summary values are averages over the entire building (i.e., multiple zones)
and specific zones may have lower or higher values than reflected in these average data.
CONCLUSIONS AND RECOMMENDATIONS

The results of this study provided the following conclusions:

* Higher percent reductions were achieved for the

smaller buildings and medium retail building which
used the PSZ HVAC system type, as shown in
Figure 3.
* The average values of percent reductions shown in
Figure 3 may be generally applied to small office,
small retail, medium office, medium retail and large
office with dual duct systems. The values for each
of these building and HVAC system types varied in
a narrow range for all prototype days and locations
simulated. Results for large retail VAV and dual
duct systems were more dependent on prototype
days and geographical location.
* Linear superposition of combined control
strategies may be applied to small and medium
building types with DX cooling coils (i.e., the
savings of individual strategies may be added together to find the combined savings). However,
linear superposition of combined control strategies may not be applied to large office and retail
building types served by a central plant.
The following recommendations for future work are provided below:
* More building types are needed to develop comprehensive control algorithms for real building
controllers. It is recommended to select all 16 reference buildings provided by the US DOE.
* The recommendations from the present study provide a direction to implement peak control
demand strategies in real buildings. Formal testing of existing buildings will verify whether the
recommendations are valid or not. Feedback from building testing is expected to pro-vide more
requirements for modeling. Several cycles from building testing and modeling are necessary to
create more effective and realistic control strategies.
ACKNOWLEDGMENTS
The authors thank the American Society of Heating, Refrigerating and Air-conditioning Engineers,
Inc. (ASHRAE) and the research manager, Mike Vaughn, for funding this work. Sincere appreciation
is expressed to members of the ASHRAE Project Monitoring Subcommittee: John House,
Steve Blanc, Carlos Haiad, Srinivas Katipamula, and Peng Xu.
REFERENCES
Deru, M., K. Field, D. Studer, K. Benne, B. Griffith, P. Torcellini, M. Halverson, D. Winiarski, B. Liu,
M. Rosen-berg, J. Huang and M. Yazdanian & D. Crawley. 2009 DOE Commercial
Building Benchmark Models for Energy Simulation, National Renewable Energy Laboratory,
Golden, Colorado

Gabel, S.D., L. Carmichael and G. Shavit. 1998. Automated control in response to real-time pricing
of electricity. ASHRAE Journal, November, 26-29.
Gu, L., R. Raustad & M. Basarkar, 2011, "Short-Term Curtailment of HVAC Loads in Buildings,"
Final report for the ASHRAE research project 1390-RP, FSEC-CR-1881-11, Florida Solar
Energy Center, Cocoa, FL 32922
Lee, K.-h. and J.E. Broun. 2006a. Evaluation of methods for determining demand-limiting set point
trajectories in commercial buildings using short-term data analysis. Proc. SimBuild 2006.
2nd National Conference of IBPSA-USA, Cambridge, MA
Lee, K.-h. and J.E. Broun. 2006b. Development of methods for determining demand-limiting set
point trajectories in commercial buildings using short-term data analysis. Proc. SimBuild 2006.
2nd National Conference of IBPSA-USA, Cambridge, MA
Motegi, N., M.A. Piettie, D.S. Watson, S. Kiliccote, P. Xu. 2007. Introduction to commercial building
control strategies and techniques for demand response. Lawrence Berkeley National Laboratory
report LBNL-59975. May. Available from http://drrc.lbl.gov/pubs/59975.pdf
PG&E. 2007. PG&E commercial building critical-peak-pricing tariff.
http://www.pge.com/tariffs/pdf/E-CPP.pdf
PNNL, 2008, http://eere.pnl.gov/building-technologies
Reddy, T.A. and L.K. Norford. 2002. Building operation and dynamics within an aggregated load.
Final Report for ASHRAE Research Project 1146-RP. ASHRAE, Atlanta
SCE. 2007a. Southern California Edison critical peak pricing rate schedules
http://www.sce.com/NR/rdonlyres/ B73F4175-162B-4C4F-B953-4E0A94863390/0/
CPPFactSheet0407.pdf
SCE. 2007b. Southern California Edison large-commercial-building interruptible-service
tarriff. http://www.sce.com/NR/sc3/tm2/pdf/ce76-12.pdf
Xing, H.-Y. 2004. Building load control and optimization. Dept. of Architecture Ph.D. theses,
Massachusetts Institute of Technology, Cambridge MA
Xu, P. 2009, "Case Study of Demand Shifting with Thermal Mass in Two Large Commercial
Buildings," ASHRAE Transactions, vol. 115, pt. 2, Louisville 2009
This paper is based on findings resulting from ASHRAE Research Project RP-1390
Lixing Gu, PhD, PE
Member ASHRAE
Lixing Gu is a principal research engineer and Richard Raustad is a senior research engineer in the
Building Research Division at the Florida Solar Energy Center in Cocoa, FL, which is associated
with the University of Central Florida.