ArticlePDF Available

Carbon emission responsive building control: A case study with an all-electric residential community in a cold climate

Authors:

Abstract

In the United States, buildings account for 35% of total energy-related carbon dioxide emissions, making them important contributors to decarbonization. Carbon intensities in the power grid are time-varying and can fluctuate significantly within hours, so shifting building loads in response to the carbon intensities can reduce a building’s operational carbon emissions. This paper presents a rule-based carbon responsive control framework that controls the setpoints of thermostatically controlled loads responding to the grid’s carbon emission signals in real time. Based on this framework, four controllers are proposed with different combinations of carbon accounting methods and control rules. To evaluate their performance, we performed simulation studies using models of a 27-home, all-electric, net zero energy residential community located in Basalt, Colorado, United States. The carbon intensity data of four future years from the Cambium data set are adopted to account for the evolving resource mix in the power grid. Various performance metrics, including energy consumption, carbon emission, energy cost, and thermal discomfort, were used to evaluate the performance of the controllers. Sensitivity analysis was also conducted to determine how the control thresholds and intervals affect the controllers’ performance. Simulation results indicate that the carbon responsive controllers can reduce the homes’ annual carbon emissions by 6.0% to 20.5%. However, the energy consumption increased by 0.9% to 6.7%, except in one scenario where it decreased by 2.2%. Compared to the baseline, the change in energy cost was between −2.9% and 3.4%, and thermal discomfort was also maintained within an acceptable range. The little impact on energy cost and thermal discomfort indicates there are no potential roadblocks for customer acceptance when rolling out the controllers in utility programs.
Abstract
In the United States, buildings account for 35% of total energy-related carbon dioxide emissions, making them important contribu-
tors to decarbonization. Carbon intensities in the power grid are time-varying and can fluctuate significantly within hours, so shift-
ing building loads in response to the carbon intensities can reduce a building’s operational carbon emissions. This paper presents
a rule-based carbon responsive control framework that controls the setpoints of thermostatically controlled loads responding to the
grid’s carbon emission signals in real time. Based on this framework, four controllers are proposed with dierent combinations of
carbon accounting methods and control rules. To evaluate their performance, we performed simulation studies using models of a
27-home, all-electric, net zero energy residential community located in Basalt, Colorado, United States. The carbon intensity data
of four future years from the Cambium data set are adopted to account for the evolving resource mix in the power grid. Various
performance metrics, including energy consumption, carbon emission, energy cost, and thermal discomfort, were used to evaluate
the performance of the controllers. Sensitivity analysis was also conducted to determine how the control thresholds and intervals
aect the controllers’ performance. Simulation results indicate that the carbon responsive controllers can reduce the homes’ annual
carbon emissions by 6.0% to 20.5%. However, the energy consumption increased by 0.9% to 6.7%, except in one scenario where it
decreased by 2.2%. Compared to the baseline, the change in energy cost was between -2.9% and 3.4%, and thermal discomfort was
also maintained within an acceptable range. The little impact on energy cost and thermal discomfort indicates there are no potential
roadblocks for customer acceptance when rolling out the controllers in utility programs.
Keywords: Decarbonization, Building control, Rule-based control, Residential community, Carbon accounting
1. Introduction
Buildings account for 35% of carbon dioxide (CO2) emis-
sions in the United States, which makes buildings important
contributors to decarbonization [1]. With the Biden admin-
istration’s aggressive goal to reduce 50%–52% of greenhouse
gas pollution by 2030 [2], a joint eort between the build-
ings and the electricity sector is emerging to tackle this chal-
lenge. Further, the U.S. Department of Energy has developed a
roadmap with recommendations for how Grid-interactive E-
cient Buildings (GEBs) can provide a clean and flexible energy
resource [3]. On one side, the power generation is adopting
more renewable energy, which is cleaner than traditional coal
and natural gas fired plants. On the other side, buildings can
reduce their carbon emissions through using less energy (e.g.,
adoption of energy eciency measures) or using cleaner energy
(e.g., load shifting to cleaner hours).
Building decarbonization can be achieved during various
phases of the building life cycle, including design, retrofit, and
operation. The design phase often incorporates carbon analy-
sis into early design building energy models [4], and some of
the studies focus on embodied carbon emission reduction [5].
Corresponding author.
Email address: xin.jin@nrel.gov (Xin Jin)
For the retrofit phase, several studies have been reported to op-
timally adopt energy eciency measures, building system elec-
trification, and high renewable penetration in existing commu-
nities for the purpose of enhancing energy performance while
attaining carbon neutrality [6, 7, 8, 9, 10].
The above two phases are generally static and concern the
long-term carbon emission performance of the buildings. Dur-
ing the building operation phase, carbon responsive build-
ing control is more flexible and its deployment requires less
capital investment. Adopting carbon emission reductions as
one of the objectives or control inputs in optimization-based
building control has received increased attention in the last
decade [11, 12, 13, 14, 15]. Jin et al. [16] proposed a user-
centric home energy management system that is based on a
multi-objective model predictive control (MPC) framework.
Carbon emission reduction serves as one of the objectives along
with the minimization of energy cost, thermal discomfort, and
user inconvenience. Leerbeck et al. [17] developed an opti-
mal heat pump controller for building space heating. Using
weather and CO2emission forecasts as inputs to an MPC, ap-
proximately 16% of CO2emissions were saved compared to
typical thermostatic control.
Despite the rapid development of optimization-based control
methods, rule-based control is still the dominant control method
in building automation systems due to ease of implementa-
Manuscript submitted to Applied Energy February 25, 2022
J. Wang, P. Munankarmi, J. Maguire, C. Shi, W. Zuo, D. Roberts, X. Jin 2022.“Carbon Emission
Responsive Building Control : A Case Study With an All-Electric Residential Community in a Cold
Climate,” Applied Energy, 314, pp. 118910, https://doi.org/10.1016/j.apenergy.2022.118910
Carbon Emission Responsive Building Control: A Case Study With an All-Electric
Residential Community in a Cold Climate
Jing Wanga,b, Prateek Munankarmib, Je Maguireb, Chengnan Shia, Wangda Zuoa,b, David Robertsb, Xin Jinb,
aDepartment of Civil, Environmental and Architectural Engineering, University of Colorado Boulder, 1111 Engineering Dr, Boulder, 80309, CO, United States
bNational Renewable Energy Laboratory, 15013 Denver West Parkway, Golden, 80401, CO, United States
tion. For example, Clauß [18] investigated predictive rule-
based control for reducing the annual CO2equivalent green-
house gas emissions (CO2eq.) for a Norwegian single-family de-
tached house. The controlled object was the building heat pump
system. Historical weather and CO2eq.emission data from 2015
were used for simulations. The results showed that the carbon
responsive control cannot reduce the yearly CO2eq.emission
due to the limited daily fluctuations in the average CO2eq.in-
tensity of the Norwegian electricity generation mix.
Carbon accounting methods play an important role in carbon
responsive building control. Although electricity net-metering
has been broadly taken into account in utility bill calculations,
carbon accounting methods have seldom been discussed in the
literature. The emissions reduction eect of clean electricity
backfeeding to the grid has generally been neglected in the de-
sign and evaluation of carbon responsive controllers. Addition-
ally, historical CO2emission data are typically adopted in exist-
ing studies [18]. With the increasing penetration of renewable
energy in the power grid, it is worth exploring evolving emis-
sion forecasts and their impact on control strategies. The table
below compares this study with existing relevant carbon emis-
sion driven control studies.
Table 1: Comparison of existing carbon emission driven control studies.
Ref. Predictive control C O2forecast Carbon net-metering
[11] 7 7 7
[12, 16, 18] 3 7 7
[13, 14, 15, 17] 3 3 7
This paper 3 3 3
In this paper, we propose a rule-based carbon responsive con-
trol framework to reduce the carbon emissions of residential
buildings. Real-time carbon emission data are utilized to inform
the setpoint control of thermostatically controlled loads such
as heating, ventilation, and air-conditioning (HVAC) systems
and water heating (WH) systems. More specifically, the impact
of carbon accounting methods such as carbon net-metering is
studied through the design of various control algorithms. Both
momentary and predictive rule-based controllers are evaluated
to study the value of prediction in real-time carbon responsive
control. In the case study, models for an all-electric residen-
tial community of 27 homes located in Basalt, Colorado, are
adopted for control implementation and evaluation. Addition-
ally, the evolving resource mix in the power grid is considered
in the simulation scenarios. The major contributions of this pa-
per include:
Development of a carbon emission responsive control
framework for decarbonizing residential buildings;
Demonstration of the control framework in a simulated all-
electric residential community in a cold climate;
Evaluation of the impact of carbon accounting methods
and control rules on emission, energy, cost, and discom-
fort;
Forecasting of the community’s future carbon emissions
with the evolving resource mix in the power grid.
The remainder of this paper is organized as follows: Sec-
tion 2 presents the research methodology of the proposed rule-
based carbon responsive control framework. The co-simulation
platform is also described in that section. Section 3 describes
the community energy model and the simulation inputs for the
case study. Section 4 discusses the simulation results with var-
ious metrics such as energy, carbon, cost, and discomfort along
with a sensitivity analysis. Section 5 concludes the work and
recommends future topics for further study.
2. Methodology
2.1. Overview of the research methodology
Rule-based carbon emission responsive control, by defini-
tion, is to make control decisions in response to real-time car-
bon emission signals following certain predetermined rules.
The control objective is to reduce the carbon emission induced
by the power usage of the controlled objects, which are thermo-
statically controlled loads in this work. The thermostat setpoint
is controlled to increase or decrease the load depending on the
carbon intensities, predetermined rules, and operation modes.
During the design of the carbon responsive control rule, a
fundamental step is to determine how the setpoints change with
the emission signal. A common method is to divide the car-
bon emission data range into several regions, where each region
correlates with one setpoint [18]. Evolved from this logic, the
ultimate rule form is to establish a function (most likely linear)
that maps emissions to a set of setpoints. In this paper, we adopt
a three-region design of the control rule, which means divid-
ing the carbon emission data range into three regions with two
thresholds. Between the lower threshold (LT) and the higher
threshold (HT) is the default zone where the default setpoints
Tset,de f ault (Table 2) will be implemented. Below the LT will
be the clean zone and the Tset,clean will be implemented, and
vice versa. We designed the Tset,clean and Tset,unclean to be 0.8C
(1.5F) around Tset,de f ault for space heating and cooling, and
8.3C (15F) for the water heater. Here we note that the default
setpoints should be dependent on the region and user prefer-
ences.
Table 2: Control setpoints for heating, cooling, and water heating.
Controlled object Tset,de f ault (C) Tset,clean (C) Tset,unclean (C)
Space heating 21.1 21.9 20.3
Space cooling 23.9 23.1 24.7
Water heater 51.7 60.0 43.4
2.1.1. Carbon accounting methods
Unlike energy net-metering, which has been extensively
studied in building-to-grid related control studies, the inclusion
of carbon net-metering is relatively rare. To address this re-
search gap, we consider dierent controller design with and
without carbon net-metering to investigate its impact on the
control performance.
For electricity prosumers who both consume and produce
electricity, carbon net-metering means metering the net carbon
2
Start
End
Implement Implement
Yes No
Implement
Yes No
Figure 1: Flow chart of momentary control with carbon net-metering.
Start
YesNo
End
Implement Implement Implement
Yes No
Implement
Yes No
Figure 2: Flow chart of momentary control without carbon net-metering.
emission caused by their electricity consumption and produc-
tion [19]. In other words, exporting electricity back to the grid
will oset carbon emissions from their total emissions. On
the contrary, when carbon net-metering is not included, it is
more advantageous to use the locally generated electricity from
photovoltaic (PV) systems instead of exporting due to the lack
of emission benefits. For controllers considering carbon net-
metering, it makes no dierence whether to use the clean PV
energy locally or to export it to the grid. Both options would
bring in the same amount of carbon emission reduction.
2.1.2. Control rules
According to the information used for making control de-
cisions, rule-based control can be categorized into momen-
tary and predictive control. The momentary rule-based con-
trol adopts the current value of the boundary condition (e.g.,
Start
End
Implement Implement
Yes No
Implement
Yes No
Yes No
Implement
Yes No
Figure 3: Flow chart of predictive control with carbon net-metering.
Start
End
Implement Implement
Yes No
Implement
Yes No
Yes No
Implement
Yes No
Yes No
Figure 4: Flow chart of predictive control without carbon net-metering.
carbon emission signal) and decides the setpoints for the cur-
rent control interval [20]. Nevertheless, the predictive rule-
based control makes decisions based on both current and fu-
ture predictions of the boundary condition [18]. Compared to
optimization-based controllers such as MPC, predictive rule-
based control is simpler but still eective. It is therefore a
promising alternative to MPC given that the rules are well de-
signed [21].
In this work, we propose both momentary and predictive
carbon responsive controllers. For the predictive controller,
the carbon emission information of one future timestep is
adopted to facilitate the determination of setpoints for the cur-
rent timestep. The detailed control algorithms of the four pro-
posed controllers are discussed in the following subsection.
2.2. Implementation of control algorithms
Figures 1–4 present the flow charts of the four proposed car-
bon responsive controllers. Here, the LT and HT are predeter-
3
Federate N Federate 1 Federate 2
Cambium
Emission
Dataset
Carbon
Controller
Agent 1
Carbon
Controller
Agent 2
Carbon
Controller
Agent N
Emission
Emission
Emission
House Agent 1 House Agent 2 House Agent N
Net load Setpoints Net load Setpoints SetpointsNet load
Co-simulation Platform
Figure 5: Architecture of the co-simulation platform.
mined thresholds that are held constant throughout the whole
simulation period. In the momentary controller with carbon
net-metering (Figure 1), the current carbon emission value COt
2
is compared with the LT and HT sequentially to determine
which range it belongs to. If COt
2is below the LT, the current
timestep will be categorized as clean and the setpoints Tset,clean
in Table 2 will thus be implemented. As a result, the loads Pt
hvac
and Pt
wh will be increased and vice versa. If COt
2is between the
LT and HT, the default setpoints Tset,de f ault will be implemented,
which is the same as the baseline setpoints.
The momentary controller without net-metering (Figure 2) is
developed based on the one with net-metering with one more
step to increase the PV self-consumption rate. Prior to com-
paring COt
2with the thresholds, the current house net load Pt
net
is first evaluated. If it is negative, which means the house has
surplus PV generation, the setpoints Tset,clean will be adopted to
increase the loads Pt
hvac and Pt
wh. In this way, the clean energy
can be consumed locally instead of being exported to the grid
without any carbon osetting benefit.
The predictive controllers (Figures 3 and 4) with and without
carbon net-metering are similar to the corresponding momen-
tary controllers in the initial steps. However, the predictive con-
trollers utilize the emission data from the next timestep to fur-
ther facilitate the control decisions. Essentially, when the emis-
sion level of the current timestep falls into the default range,
the carbon emission of the next timestep COt+1
2is then com-
pared with the LT and HT. If the next timestep is categorized
as clean, the loads for the current timestep will be decreased to
save for later, and vice versa. This algorithm enables more fre-
quent setpoint adjustments and therefore leads to more shifting
of loads to cleaner hours.
2.3. Co-simulation platform
A co-simulation platform is used for testing and validating
the performance of the proposed control algorithms. The co-
simulation platform is built on hierarchical engine for large-
scale co-simulation (HELICS) tool [22]. Key capabilities such
as high scalability, cross-platform operability, and modularity
make HELICS suitable for developing co-simulation platforms.
The co-simulation platform manages the data communication
between dierent components of the closed-loop co-simulation.
It is scalable and creates multiple agents depending on the num-
ber of houses in the simulated community.
Figure 5 depicts the components of the co-simulation plat-
form and the data exchange flow between dierent components.
Two types of agents are developed in this platform: (1) a build-
ing energy simulator (i.e., house agent) and (2) a carbon re-
sponsive controller agent. At the beginning of the simulation,
an agent for the building simulation and an agent for the con-
troller are instantiated simultaneously for each house.
Inside each house agent, we used the Operational,
Controllable, High-resolution Residential Energy (OCHRE)
model [23] for modeling the buildings in this study. OCHRE
is capable of implementing the control signals for HVAC, WH,
PV, and battery, and it is designed to be easily integrated into
the co-simulation platform. More details about the community
modeling are discussed in Section 3.
The data exchange between the house agent and the con-
troller agent happens in the following order. At the beginning
of timestep t, OCHRE computes the net house load (Pt
net), and
the house agent sends Pt
net to the controller. At the same time,
the controller agent receives the Pt
net together with the current
carbon emission data from the Cambium database [24]. For
predictive controllers, the future carbon emission data are also
4
received for computing the control actions with the assumption
of perfect forecasts. The controller then computes the control
setpoints based on the predetermined rules and sends it back to
the house agent. After receiving the control action for each de-
vice from the controller agent, OCHRE implements the control
action and proceeds to the next timestep t+1.
3. Case study settings
3.1. Overview of the community
The Basalt Vista community, located in Basalt, Colorado, is
modeled for the case study. This community is intended to pro-
vide aordable housing to schoolteachers in town while also be-
ing highly ecient all-electric homes with enough PV to make
the community approximately net zero, as well as batteries for
load shifting and resilience. Located at an elevation of 2,015
meters, it is in climate zone 7B, which is very cold and dry [25].
The community consists of 12 multifamily buildings, either du-
plexes or triplexes, with a total 27 homes. The homes consist
of 2-, 3-, and 4-bedroom units, ranging from 107 to 156 square
meters [26].
The HVAC systems of all the units are minisplit heat pump
(MSHP) systems. The domestic hot water systems use heat
pump water heaters (HPWH). Each unit is installed with
rooftop PV systems, except for three units that do not have
enough roof space for PV panels. The PV system sizes range
from 7.6 to 11.85 kW. Among all the households, six homes are
assumed to own electric vehicles and are equipped with Level
1 or Level 2 chargers. Four homes are equipped with home
battery systems with capacities of 12 kWh.
3.2. Community modeling
The community is modeled in OCHRE [23], a control-
oriented residential building modeling tool. Each building type
(i.e., 2-, 3-, and 4-bedroom) share the same model formation
with variant configuration parameters. The building envelope
and the controllable loads (i.e., MSHP and HPWH) are de-
scribed in this section. Models of PV, batteries, electric ve-
hicles, and other electric loads can be found in reference [23].
The resistance-capacitance (RC) model is adopted for mod-
eling the building envelope. R represents the thermal resistance
between thermal masses, which involves both conduction and
convection eects. C represents the thermal mass of dierent
building components such as the air inside a room, exterior and
interior walls, furniture, etc. The equation of a node iin the RC
network is:
Ci
dT t
i
dt =
M
X
j=1
Tt
jTt
i
Ri j
+Ht
i,(1)
where Ciis the thermal mass of the node i,Tt
iis the temperature
of node i,Tt
jis the temperature of node jadjacent to node i,Ri j
is the thermal resistance between nodes iand j,Ht
iis the sensi-
ble heat injected to node i, and Mis the total number of nodes
in a house. For each house, approximately 13 Rs and 4 Cs were
used to simulate the thermal dynamics. The detailed structure
of the RC network model can be found in reference [23].
The MSHP in this work has a variable frequency drive that
can modulate to maintain the setpoint of the conditioned space.
A local proportional integral derivative controller adjusts the
variable frequency drive’s speed ratio according to the mea-
sured room temperature and the setpoint. For this community,
the MSHPs are sized to be larger than might typically be sized
according to standard sizing guidance such as the ACCA Man-
ual J [27]. This is to ensure that the heat pump can fully meet
the load even during the coldest hours of the year. With the
oversized heat pumps, backup heaters are not installed, which
makes these homes more energy ecient at the expense of a
higher capital cost.
Further, each home has a highly ecient HPWH with backup
electric resistance element installed. The heat pump itself is
able to heat the tank more eciently by removing heat from the
ambient air and adding it to the tank. In contrast, the backup
electric element is less ecient, but can heat the tank more
quickly. When there is a demand for water heating, the heat
pump turns on first. If there is a suciently large draw of hot
water that the heat pump is unable to keep up, the backup elec-
tric element will then turn on. This will partially recover the
tank temperature before switching back to the heat pump mode
to maintain a high level of energy eciency. The exact con-
trol sequence was derived from the detailed laboratory testing
of this HPWH unit in response to dierent typical residential
hot water draw profiles [28].
3.3. Inputs
The carbon emission data in this work are adopted from
the Cambium data set [24]. Based on modeled futures of the
U.S. electricity sector, Cambium assembles structured data sets
of energy-related metrics (e.g., carbon emissions) to facilitate
long-term decision-making. Specifically, the hourly short-run
marginal carbon emission data from the Standard Scenarios
2020 Mid-case were adopted in the control development of
this work. Though Cambium provides various scenario set-
tings such as high vs. low renewable energy cost, we note
that the simulated data are based on certain assumptions about
the future projected U.S. electric sector. These assumptions are
subject to many uncertainties, such as climate change and pol-
icy impacts, which could aect the results and analysis of this
work.
Figure 6 visualizes the emissions data for the four selected
simulation years from Cambium. We selected these four years
because only even years are available in Cambium and we
needed to avoid leap years as they are not yet supported in
OCHRE. From the figure, we notice that as time goes by, more
hours with zero or low carbon emissions emerge. This aligns
with the increasing adoption rate of renewable energy and de-
carbonization measures of the electric sector. Regardless, we
still observe some hours of high emission rate in year 2046,
which represent the operation of coal-fired power plants. Be-
cause the emission data for the four years are dierent, we con-
ducted a sensitivity analysis to determine the control thresholds
(Section 4.2). As a result, we adopted the absolute values of the
30% and 70% of the emission range in 2022 as the LT and HT
for all scenarios.
5
Figure 6: Emission data comparison across simulated years.
Similarly, the same weather data were used across all four
simulation years to eliminate the impact of weather on build-
ing loads so that the results can be directly compared. Ac-
tual Meteorological Year (AMY) data of year 2012 for Pitkin
County, Colorado, where the community is located, was used
to be consistent with the Cambium weather data file. Here we
note that forecasting future weather is an active research topic
and is beyond the scope of this work. The occupancy, lighting,
and appliance usage schedules for each house were generated
from ResStockTM [29]. The time-of-use rate for the local elec-
tric utility Holy Cross Energy [30] was used for the energy cost
calculation, where the on-peak (4–9 PM) price is $0.24/kWh
and the o-peak (rest of the hours) price is $0.06/kWh.
3.4. Scenarios and evaluation metrics
Based on the case study settings proposed above, we de-
signed five simulation scenarios to evaluate the controller per-
formances. As seen in Table 3, the baseline scenario involves no
carbon responsive building controller. The remaining four sce-
narios each adopts one carbon responsive controller proposed
in Section 2, namely momentary or predictive rule-based con-
troller with or without carbon net-metering. Annual energy
simulations of the four selected years (2022, 2030, 2038, and
2046) were run with the default setpoints listed in Table 2.
The house models were run with a timestep of 1 minute, and
the control setpoints were updated with a 15-minute interval to
avoid excessive cycling of the appliances.
To account for dierent carbon accounting methods, two
ways to calculate the annual carbon emission are adopted. The
first method is carbon net-metered, where the power exported
to the grid can oset the total carbon emission. The second
method does not take account of the exported power back to the
Table 3: Description of simulation scenarios in the case study.
Scenario Carbon accounting method Rule type Year
Baseline N/A N/A 2022–2046
MO-0 Non-net-metering Momentary 2022–2046
MO-1 Net-metering
PR-0 Non-net-metering Predictive 2022–2046
PR-1 Net-metering
grid. The equations for the two methods are as follows:
Enet =
N
X
t=1
etPt
net t,(2)
Enonnet =
N
X
t=1
etPt
net t,Pt
net >0,(3)
where Enet and Enonnet are the annual emission with and with-
out carbon net-metering. tis the timestep, and Nis the total
number of timesteps in a year. etrepresents the real-time car-
bon emissions of the grid power. Pt
net is the net load of the house
(i.e., total load subtracted by PV and battery power). tis the
interval length of each timestep.
Thermal discomfort is quantified by the discomfort degree
hours for the HVAC system. For HVAC heating scenarios,
when the room temperature is below the heating setpoint, it is
considered an uncomfortable time period, and vice versa. The
switching between heating and cooling is dependent on the out-
door temperature. Note that in this paper we consider a set-
point not met as uncomfortable as the actual room temperature
deviates from the predetermined setpoint. We did not adopt
the ASHRAE recommended comfort range (i.e., 20–28C) [31]
6
Figure 7: Community average annual energy consumption of controlled loads for dierent controllers across simulated years.
because it is relatively wide compared to our setpoints, which
leads to almost the same level of discomfort in all scenarios. A
list of the discomfort values according to the ASHRAE Stan-
dard is provided in Table A.2 in the appendix. The discomfort
degree hours in this work can be defined by the following equa-
tion:
Uhvac =
N
X
t=1
|Tt
indoor Tt
set|t,
Tt
indoor <Tt
set (heating)and T t
indoor >Tt
set (cooling),
(4)
where Tindoor and Tset are the actual indoor temperature and set-
point, and tis the simulation interval of the house model (i.e.,
1 minute).
The thermal discomfort for WH is only considered for the
shower in this work [32]. It is measured by the unmet thermal
energy for any shower draws below 43.3C (110F), which is
calculated by:
Uwh =
N
X
t=1
mt
water cp|Tt
water 43.3|t,
Tt
water <43.3,
(5)
where mt
water represents the hot water mass flow rate and Tt
water
is the hot water temperature.
4. Results and discussions
4.1. Results
This subsection presents the annual simulation results with
four performance metrics: energy consumption, carbon emis-
sion, energy cost, and thermal discomfort. Through the com-
parison of the four proposed controllers, we discuss the impact
of carbon net-metering, as well as the eect of prediction in
rule-based control. Additionally, we also investigate the evolu-
tion of results over the years.
4.1.1. Energy consumption
Figure 7 visualizes the community average annual energy
consumption of the controlled loads (HVAC and WH) for dif-
ferent controllers. Compared to the baseline, the annual space
Table 4: Community average annual energy consumption per household for dif-
ferent controllers across simulated years (percentage values are changes relative
to the corresponding baseline).
Year Baseline (kWh/yr) MO-0 MO-1 PR-0 PR-1
2022 4,124 3.8% 1.3% 1.1% -2.2%
2030 4,120 6.6% 3.7% 4.5% 0.9%
2038 4,108 6.7% 4.2% 4.7% 1.6%
2046 4,103 5.8% 3.7% 4.2% 1.6%
heating energy consumption slightly decreases while both the
space cooling and water heating energy increases. The diver-
gent heating and cooling energy changes can be attributed to
the higher emission levels in the heating season than those in
the cooling season (see Figure 6), which leads to lower heating
energy consumption. In general, carbon responsive building
control increases the total household annual energy consump-
tion.
Table 4 lists the community average annual energy per house-
hold. In the table, the baseline energy consumption is shown
in absolute values, whereas the other scenarios are in relative
changes compared to the baseline. From the table, the aver-
age annual energy increase for the studied community is within
6.7%. For controllers accounting for carbon net-metering, the
energy increase is about 2.5% to 3.0% lower compared to
scenarios without carbon net-metering. This indicates the in-
centivizing impact of adopting carbon net-metering, as it en-
courages the exporting of power back to the grid. For con-
trollers without carbon net-metering, the energy usage is in-
creased when PV generation is excessive to increase the self-
consumption rate, which has led to a higher annual energy con-
sumption.
Another observation from Table 4 is that all predictive con-
trollers perform better than the momentary controllers in terms
of energy consumption. This is because the predictive con-
trollers adjust the setpoints based on both current and future
emission values. When the current carbon emission falls in the
default zone but the future emission falls in the clean zone, it
will reduce the current power, and vice versa. Because there are
7
Figure 8: Histogram of carbon emission data for the simulated years.
more time instances classified as clean than carbon-intensive
over the whole year (as shown in Figure 8), the predictive con-
trollers therefore reduce energy consumption more often than
increase it and save more energy compared to the momentary
controllers.
4.1.2. Carbon emissions
Figure 9 illustrates the community average annual carbon
emission of the controlled loads (HVAC and WH). From the
figure, it can be seen that in all scenarios, the emissions caused
by space heating are reduced through the carbon responsive
control. In some scenarios, the emissions produced by cooling
and water heating increased due to the corresponding energy
increase. This is attributed to the fact that when there are con-
secutive clean hours, the controllers might pre-cool or pre-heat
the space/water tank more than necessary, which leads to an in-
crease of the total operational emissions. Overall, the annual
household carbon emissions decreased compared to the base-
line.
Table 5 lists the community average emission per household
for dierent scenarios. In order to eliminate the impact of
dierent carbon accounting methods and focus on the perfor-
mance variance of the controllers, the carbon emission values
listed in Table 5 are all net-metered (Equation 2). The emissions
results calculated without carbon net-metering are presented in
Table A.1 of the Appendix.
Based on Table 5, we can see that the controllers that con-
sider carbon net-metering perform better than those that did not.
This can be attributed to the logic of the controllers, where the
non-net-metered controller increases the loads to do more pre-
cooling/pre-heating when the house net-load is negative. This
has led to a rise in energy as discussed in former sections, which
hinders them from decreasing more carbon emissions.
Similar to the annual energy performance, the emission per-
formance of the predictive controllers is better than the mo-
mentary ones. The reason is that predictive controllers make
decisions informed by the emissions at the next timestep. This
Table 5: Community average annual carbon emission per household for dif-
ferent controllers across simulated years (net-metered; percentage values are
changes relative to the corresponding baseline).
Year Baseline (kg/yr) MO-0 MO-1 PR-0 PR-1
2022 3,854 -6.0% -11.0% -6.8% -12.1%
2030 2,158 -6.3% -18.7% -7.8% -20.5%
2038 2,650 -6.5% -15.2% -7.7% -16.5%
2046 3,453 -6.3% -11.6% -7.1% -12.6%
enables a better shifting eect of loads to cleaner time peri-
ods than the momentary controllers. Cumulatively over a year,
more emission is thus reduced by predictive controllers.
Overall, 6% to 20.5% annual carbon emissions reduction is
seen by the proposed rule-based controllers compared to the
baseline. In terms of the yearly change of emissions, it is inter-
esting that year 2046 is not the one with the lowest emissions.
According to Figure 8, though year 2046 has the most hours of
zero emission, depending on the power profile of the houses, the
zero emission hours might not align with the high-load hours.
Figure 10 plots the heat map of a typical house’s net load in
year 2046. We see that the high-load hours are mostly in winter
and during the nighttime, which align with the carbon-intensive
hours in Figure 6. Therefore, it is safe to infer that the more the
high-load hours align with the low-carbon hours, the lower the
annual emissions will be.
4.1.3. Energy cost
Figure 11 plots the community average annual energy on-
and o-peak costs of the controlled loads in 2022. Similar plots
for the other simulation years can be found in Appendix A.
From Figure 11, we can see that the on-peak energy costs all
decreased and the o-peak costs all increased compared to the
baseline. Figure 12 plots the house net load in response to the
real-time emission signal on a sample winter day of a sample
house. The figure shows that when the carbon emission of the
on-peak hours exceeds the higher threshold at around 6 PM, the
house net load drops below the baseline as all controllers lower
the heating and WH setpoints to reduce the loads. However, a
certain level of the rebound eect is seen later, where the con-
trollers consume slightly more energy than the baseline due to
the lower setpoints earlier. Overall, during the on-peak hours,
more energy is saved due to the high emissions level. Over the
whole year of simulation time, the on-peak hours are relatively
more carbon-intensive and o-peak hours cleaner, which has
led to the reduction of on-peak cost and increase of o-peak
cost in Figure 11.
Table 6 summarizes the community average annual energy
cost per household for each simulation year. Based on the ta-
ble, all controllers that do not consider carbon net-metering per-
form better in terms of total cost. This can be attributed to the
fact that controllers without net-metering tend to use more en-
ergy around noon (e.g., 11 AM to 3 PM in Figure 12) when the
house net load is negative. Because of the rebound eect, they
8
Figure 9: Community average annual carbon emission of controlled loads for dierent controllers across simulated years.
Figure 10: Annual house net load heat map (year 2046, house b5).
Figure 11: Community average annual energy on- and o-peak costs of con-
trolled loads for dierent controllers in 2022.
will consume less energy later when PV generation decreases
and the net load is positive. This enables shifting the load from
on-peak hours to o-peak hours, which leads to more energy
cost savings by the non-net-metered controllers. Additionally,
all predictive controllers have more cost savings than the mo-
mentary ones. The reason is similar to the energy savings dis-
cussion in Section 4.1.1.
Generally, we see an annual energy cost change of -4.1% to
3.4% on top of the baseline. In terms of the yearly trend of
cost savings, we notice that the cleaner the year, the less cost
reduction potential. Here, clean means more hours of carbon
emission under the lower threshold in Figure 8. More specif-
ically, years 2030 and 2038 have more hours under the lower
threshold, and the energy cost increased in all scenarios except
Figure 12: House net load in response to the emission signal (winter 2022,
house b5).
Table 6: Community average annual energy cost per household for dierent
controllers across simulated years (percentage values are changes relative to
the corresponding baseline).
Year Baseline ($/yr) MO-0 MO-1 PR-0 PR-1
2022 718 -2.7% -1.1% -4.1% -2.9%
2030 719 1.3% 3.4% 0.1% 1.7%
2038 717 0.5% 2.9% -0.6% 1.3%
2046 715 -1.0% 1.2% -1.8% 0.0%
one.
4.1.4. Thermal discomfort
Table 7 lists the heating, cooling, and water heating dis-
comfort values for all scenarios across the simulation years.
From the table, we see that carbon responsive control brings
in a higher level of discomfort in space heating and cooling
compared to the baseline, where constant setpoints were imple-
mented annually. However, the hot water discomfort has been
9
Table 7: Community average heating, cooling, and water heating discomfort metric values for each simulation scenario.
Year Heating (C-hrs/yr) Cooling (C-hrs/yr) Hot water (kWh/yr)
Baseline MO-0 MO-1 PR-0 PR-1 Baseline MO-0 MO-1 PR-0 PR-1 Baseline MO-0 MO-1 PR-0 PR-1
2022 8 24 27 30 33 0 2.2 4.1 1.9 3.4 1.3 0.15 0.23 0.19 0.25
2030 8 34 40 31 36 0 1.0 4.4 1.0 3.5 1.3 0.26 0.40 0.26 0.37
2038 8 69 79 64 73 0 1.7 7.5 1.5 6.0 1.3 0.37 0.57 0.39 0.58
2046 8 55 66 51 59 0 2.8 6.8 2.4 5.6 1.3 0.37 0.64 0.31 0.56
slightly improved, which is validated through a higher annual
average hot water temperature. Generally, the annual discom-
fort levels in all the carbon responsive scenarios are maintained
within an acceptable range.
When carbon net-metering is considered in the control, the
discomfort level increases compared to scenarios without net-
metering. This is because when power exporting does not
bring emission benefits, the controller tries to increase the self-
consumption rate of the surplus PV generation by consuming
more energy. This has led to a relatively higher indoor temper-
ature in the heating season and a relatively lower temperature
in the cooling season, which means a more comfortable indoor
environment for the occupants.
Comparing scenarios with and without predictive control, we
notice that predictive control can lead to either an increase or
decrease of the discomfort level. There is no direct correlation
between the two. Though predictive controllers look ahead for
one timestep, the decision to increase or decrease the setpoints
is dependent on the emission level at the next timestep, which
is rather stochastic in the time frame of one year of simulation
time.
4.2. Sensitivity analysis
Control thresholds. The choice of the lower and upper
thresholds in rule-based control is essential. To investigate how
the thresholds aect the control performance, we conducted a
sensitivity analysis. The original lower and upper thresholds
were adjusted based on the simulation scenario MO-1 2022. A
list of the control performance metrics can be found in Table 8.
The percentages (e.g., 10&90) here represent the LT and HT,
which are based on the emission data range of 2022. The same
absolute values of the LT and HT were used across dierent
years.
From the table, carbon emissions decrease while the control
threshold range gets narrower. Specifically, the emission drops
drastically from thresholds 20&80 to 30&70 because the lat-
ter is more sensitive to emission changes, which leads to more
frequent carbon responsive setpoint adjustments.
The relationship between energy consumption and the
thresholds is less explicit. Given the same emission signal in-
put, the closer the thresholds, the more hours that fall out of the
default zone. Hence, the energy consumption variation depends
on the distribution of the emission data. For instance, when
the thresholds change from 20&80 to 30&70, more hours are
Table 8: Comparison of average performance metric values per household with
various control thresholds.
Lower & upper threshold percentages
10&90 20&80 30&70 40&60
Emission (kg/yr) 3,852 3,819 3,429 3,088
Energy (kWh/yr) 4,123 4,106 4,177 4,046
Total energy cost ($/yr) 718 715 711 706
On-peak cost ($/yr) 627 625 613 617
O-peak cost ($/yr) 91 90 97 88
Discomfort
Heating (C-hrs/yr) 8 11 27 238
Cooling (C-hrs/yr) 0 0.1 4 21
Hot water (kWh/yr) 1.3 1.2 0.2 0.9
becoming clean compared to hours that are becoming carbon-
intensive. This leads to more energy consumed over the whole
year.
The annual energy cost sinks with the narrowing threshold
range, making the 40&60 thresholds the most cost-eective
range. However, the zone thermal discomfort level also reaches
the highest in the 40&60 range. This is because the frequent
thermostat changes cause the room temperature to swing so that
it is more likely to fall outside the comfort zone. The hot water
temperature depends not only on the setpoint but also the water
draw profile. There is therefore no explicit correlation between
the hot water discomfort level and the thresholds.
To summarize, the 30&70 lower and upper control thresh-
olds chosen in this work best balance the benefits of emissions
reduction and energy cost, as well as thermal comfort for the
homeowners.
Control interval. The impact of control interval is studied
through varied intervals based on scenario MO-1 2022. From
Table 9, we observe that the performance of the 15-minute
and 30-minute control intervals is very similar to each other.
The 15-minute interval performs better at energy and comfort,
whereas the 30-minute is better at emissions and cost. When
the interval becomes larger than 30 minutes (e.g., 60 minutes
and 120 minutes), then the larger the interval, the worse the
control performance. This is probably because in these cases,
the controller changes the setpoints too infrequently, which hin-
ders the benefits of carbon responsive control. Considering the
10
Table 9: Comparison of average performance metric values per household with
various control intervals.
Control interval (minute)
15 30 60 120
Emission (kg/yr) 3,429 3,387 3,480 3,740
Energy (kWh/yr) 4,177 4,184 4,240 4,329
Total energy cost ($/yr) 711 710 714 724
On-peak cost ($/yr) 613 612 613 619
O-peak cost ($/yr) 97 98 101 105
Discomfort
Heating (C-hrs/yr) 27 31 38 49
Cooling (C-hrs/yr) 4 5 6 12
Hot water (kWh/yr) 0.2 0.2 0.2 0.3
balance between control performance and the building thermal
dynamics, we chose the control interval to be 15 minutes.
5. Conclusion
In this paper, we propose a carbon emission responsive con-
trol framework for thermostatically controlled loads. Within
this framework, the four various controllers adjust thermostat
setpoints according to projected carbon emission signals. The
impact of carbon net-metering in both momentary and predic-
tive rule-based controllers is investigated through the controller
design and a case study. Sensitivity analysis is conducted to
evaluate the role of control thresholds and control interval in
the controller design.
Based on the simulation results, the average annual house-
hold carbon emissions are decreased by 6.0% to 20.5% com-
pared to the baseline. The average annual energy consumption
is increased by less than 6.7% due to more clean hours over
the year. The annual energy cost change lies between -4.1%
and 3.4% on top of the baseline. All on-peak energy costs de-
creased while all o-peak costs increased, indicating that the
carbon intensities during on-peak hours are higher than those
during o-peak hours. Generally, the annual discomfort levels
in all the carbon responsive scenarios are maintained within an
acceptable range.
Evaluating the impact of carbon net-metering, we found that
controllers with carbon net-metering show 2.5% to 3.0% less
energy consumption and 5% to 12.7% less emission than con-
trollers without carbon net-metering. This indicates the incen-
tivizing impact of adopting carbon net-metering, as it encour-
ages the exporting of power back to the grid. For controllers
without carbon net-metering, higher annual energy consump-
tion and carbon emissions result from attempting to increase
the PV self-consumption rate. However, all controllers that do
not consider carbon net-metering perform better in terms of the
total cost. Due to the rebound eect, they tend to be shifting
loads from on-peak hours to o-peak hours, causing the total
cost to sink. Further, because more energy is consumed, non-
net-metering controllers tend to create a more comfortable in-
door environment for the occupants.
All predictive controllers perform better than the momentary
controllers in terms of energy consumption, carbon emission,
and energy cost. This is attributed to the enhanced load shifting
eect by the predictive controller design. Also, this finding ver-
ifies the claim in reference [21] that predictive rule-based con-
trollers are promising alternatives to optimization-based con-
trollers because they are simpler and still eective.
We notice in some scenarios the emissions produced by space
cooling and water heating are higher compared to the base-
line due to the increased energy consumption from load shift-
ing. This indicates rule-based control solely informed by car-
bon emission signals may end up with higher emissions, which
could be overcome by using optimization-based control meth-
ods such as MPC.
Future work includes:
Investigating better designs of the control rules to achieve
synergetic emission, energy, and cost reductions.
Incorporating other types of controllable loads such as
schedulable loads into the carbon emission responsive
control framework.
Comparing the performance of the developed rule-based
control to optimization-based control.
Acknowledgements
This work was authored by the National Renewable Energy
Laboratory, operated by Alliance for Sustainable Energy, LLC,
for the U.S. Department of Energy (DOE) under Contract No.
DE-AC36-08GO28308. Funding provided by the U.S. Depart-
ment of Energy Oce of Energy Eciency and Renewable En-
ergy Building Technologies Oce. The views expressed in the
article do not necessarily represent the views of the DOE or the
U.S. Government. The U.S. Government retains and the pub-
lisher, by accepting the article for publication, acknowledges
that the U.S. Government retains a nonexclusive, paid-up, ir-
revocable, worldwide license to publish or reproduce the pub-
lished form of this work, or allow others to do so, for U.S. Gov-
ernment purposes. This research was also partially supported
by the National Science Foundation under Awards No. CBET-
2217410. The authors would like to gratefully acknowledge
Holy Cross Energy and Habitat for Humanity Roaring Fork Val-
ley for providing the building floor plans and the site plan of the
community, the AMI data, and collaborative discussions.
11
Appendix A.
Table A.1: Community average annual carbon emission per household for dif-
ferent controllers across simulated years (non-net-metered; percentage values
are changes relative to the corresponding baseline).
Year Baseline (kg/yr) MO-0 MO-1 PR-0 PR-1
2022 8,318 -5.1% -4.0% -5.4% -4.4%
2030 7,201 -4.6% -3.4% -4.9% -3.8%
2038 7,476 -5.0% -3.6% -5.3% -4.0%
2046 7,793 -5.4% -4.0% -5.8% -4.3%
Figure A.1: Community average annual energy on- and o-peak costs of con-
trolled loads for dierent controllers in 2030.
Figure A.2: Community average annual energy on- and o-peak costs of con-
trolled loads for dierent controllers in 2038.
References
[1] U.S. Energy Information Administration, U.S. energy-related carbon
dioxide emissions, 2019, Tech. rep. (2020).
[2] The White House, Fact sheet: President Biden sets 2030 greenhouse gas
pollution reduction target aimed at creating good-paying union jobs and
securing U.S. leadership on clean energy technologies, accessed 9 July
2021 (2021).
URL https://www.whitehouse.gov/briefing-room/
statements-releases/2021/04/22/fact- sheet-president-
biden-sets- 2030-greenhouse- gas-pollution- reduction-
target-aimed- at-creating- good-paying- union-jobs-and-
securing-u- s-leadership- on-clean- energy-technologies/
Figure A.3: Community average annual energy on- and o-peak costs of con-
trolled loads for dierent controllers in 2046.
Table A.2: Community average indoor discomfort values for each simulation
scenario based on ASHRAE Standard 55.
Year ASHRAE discomfort (C-hrs/yr)
Baseline MO-0 MO-1 PR-0 PR-1
2022 3.8 3.5 3.5 3.9 3.9
2030 3.8 2.1 2.1 2.1 2.1
2038 3.8 3.2 3.3 3.3 3.3
2046 3.8 3.1 3.2 3.7 3.8
[3] A. Satchwell, M. A. Piette, A. Khandekar, J. Granderson, N. M. Frick,
R. Hledik, A. Faruqui, L. Lam, S. Ross, J. Cohen, K. Wang, D. Urigwe,
D. Delurey, M. Neukomm, D. Nemtzow, A national roadmap for grid-
interactive ecient buildings, Tech. rep. (2021). doi:https://doi.
org/10.2172/1784302.
[4] G. Legorburu, A. D. Smith, Incorporating observed data into early design
energy models for life cycle cost and carbon emissions analysis of campus
buildings, Energy and Buildings 224 (2020) 110279. doi:https://
doi.org/10.1016/j.enbuild.2020.110279.
[5] Y. Cang, Z. Luo, L. Yang, B. Han, A new method for calculating
the embodied carbon emissions from buildings in schematic design:
Taking “building element” as basic unit, Building and Environment
185 (2020) 107306. doi:https://doi.org/10.1016/j.buildenv.
2020.107306.
[6] M. Almeida, M. Ferreira, Ten questions concerning cost-eective energy
and carbon emissions optimization in building renovation, Building and
Environment 143 (2018) 15–23. doi:https://doi.org/10.1016/j.
buildenv.2018.06.036.
[7] S. M. Garriga, M. Dabbagh, M. Krarti, Optimal carbon-neutral retrofit of
residential communities in Barcelona, Spain, Energy and Buildings 208
(2020) 109651. doi:https://doi.org/10.1016/j.enbuild.2019.
109651.
[8] J. Langevin, C. B. Harris, J. L. Reyna, Assessing the potential to reduce
US building CO2emissions 80% by 2050, Joule 3 (10) (2019) 2403–
2424. doi:https://doi.org/10.1016/j.joule.2019.07.013.
[9] A. Fathy, A. Salib, M. Krarti, Transitioning from net-zero energy homes
to carbon-neutral grid-connected communities, ASME Journal of Engi-
neering for Sustainable Buildings and Cities 1 (4) (2020). doi:https:
//doi.org/10.1115/1.4049074.
[10] Y. Lou, Y. Yang, Y. Ye, W. Zuo, J. Wang, The eect of building retrofit
measures on co2emission reduction–a case study with US medium oce
buildings, Energy and Buildings 253 (2021) 111514.
[11] T. Renugadevi, K. Geetha, K. Muthukumar, Z. W. Geem, Optimized en-
ergy cost and carbon emission-aware virtual machine allocation in sus-
tainable data centers, Sustainability 12 (16) (2020) 6383. doi:https:
//doi.org/10.3390/su12166383.
[12] H. Ghazzai, A. Kadri, Joint demand-side management in smart grid for
green collaborative mobile operators under dynamic pricing and fairness
12
setup, IEEE Transactions on Green Communications and Networking
1 (1) (2016) 74–88. doi:https://doi.org/10.1109/tgcn.2016.
2646818.
[13] Z. T. Olivieri, K. McConky, Optimization of residential battery energy
storage system scheduling for cost and emissions reductions, Energy and
Buildings 210 (2020) 109787. doi:https://doi.org/10.1016/j.
enbuild.2020.109787.
[14] T. P´
ean, R. Costa-Castell´
o, J. Salom, Price and carbon-based energy flex-
ibility of residential heating and cooling loads using model predictive
control, Sustainable Cities and Society 50 (2019) 101579. doi:https:
//doi.org/10.1016/j.scs.2019.101579.
[15] J. Gasser, H. Cai, S. Karagiannopoulos, P. Heer, G. Hug, Predictive en-
ergy management of residential buildings while self-reporting flexibility
envelope, Applied Energy 288 (2021) 116653. doi:https://doi.org/
10.1016/j.apenergy.2021.116653.
[16] X. Jin, K. Baker, D. Christensen, S. Isley, Foresee: A user-centric home
energy management system for energy eciency and demand response,
Applied Energy 205 (2017) 1583–1595. doi:https://doi.org/10.
1016/j.apenergy.2017.08.166.
[17] K. Leerbeck, P. Bacher, R. G. Junker, A. Tveit, O. Corradi, H. Mad-
sen, R. Ebrahimy, Control of heat pumps with CO2emission intensity
forecasts, Energies 13 (11) (2020) 2851. doi:https://doi.org/10.
3390/en13112851.
[18] J. Clauß, S. Stinner, I. Sartori, L. Georges, Predictive rule-based con-
trol to activate the energy flexibility of Norwegian residential buildings:
Case of an air-source heat pump and direct electric heating, Applied
Energy 237 (2019) 500–518. doi:https://doi.org/10.1016/j.
apenergy.2018.12.074.
[19] X. Zhang, Y. Wang, How to reduce household carbon emissions: A review
of experience and policy design considerations, Energy Policy 102 (2017)
116–124. doi:https://doi.org/10.1016/j.enpol.2016.12.010.
[20] B. Alimohammadisagvand, J. Jokisalo, K. Sir ´
en, Comparison of four
rule-based demand response control algorithms in an electrically and heat
pump-heated residential building, Applied Energy 209 (2018) 167–179.
doi:https://doi.org/10.1016/j.apenergy.2017.10.088.
[21] D. Fischer, H. Madani, On heat pumps in smart grids: A review, Renew-
able and Sustainable Energy Reviews 70 (2017) 342–357. doi:https:
//doi.org/10.1016/j.rser.2016.11.182.
[22] B. Palmintier, D. Krishnamurthy, P. Top, S. Smith, J. Daily, J. Fuller,
Design of the HELICS high-performance transmission-distribution-
communication-market co-simulation framework, in: 2017 Workshop on
Modeling and Simulation of Cyber-Physical Energy Systems (MSCPES),
IEEE, 2017, pp. 1–6. doi:https://doi.org/10.1109/mscpes.
2017.8064542.
[23] M. Blonsky, J. Maguire, K. McKenna, D. Cutler, S. P. Balamurugan,
X. Jin, OCHRE: The object-oriented, controllable, high-resolution res-
idential energy model for dynamic integration studies, Applied Energy
290 (2021) 116732. doi:https://doi.org/10.1016/j.apenergy.
2021.116732.
[24] P. Gagnon, W. Frazier, E. Hale, W. Cole, Cambium documentation: Ver-
sion 2020, Tech. rep., National Renewable Energy Laboratory (2020).
[25] The American Society of Heating, Refrigerating and Air-Conditioning
Engineers, ASHRAE Standard 169-2006: Weather data for building de-
sign standards (2006).
[26] Aspen/Pitkin County Housing Authority, Basalt Vista housing, accessed
30 June 2021 (2021).
URL https://www.apcha.org/374/Basalt-Vista- Housing/
[27] Air Conditioning Contractors of America (ACCA), Manual J: Residential
load calculation, Tech. report, ACCA (2008).
[28] B. Sparn, K. Hudon, D. Christensen, Laboratory performance evaluation
of residential integrated heat pump water heaters, Tech. rep., National
Renewable Energy Laboratory (2014).
[29] E. Wilson, C. Christensen, S. Horowitz, J. Robertson, J. Maguire, Energy
eciency potential in the U.S. single-family housing stock, Tech. rep.,
National Renewable Energy Laboratory (2017).
[30] Holy Cross Energy, Holy Cross Energy electric service taris, rules and
regulations, accessed 8 July 2021 (2021).
URL https://www.holycross.com/wp-content/uploads/2018/
12/Electric-Service- Tariffs-Rules- and-Regulations-
12.17.2018.pdf
[31] The American Society of Heating, Refrigerating and Air-Conditioning
Engineers, ASHRAEstandard 55 – thermal environmental conditions for
human occupancy (2020).
[32] J. Maguire, How hot is hot enough? Quantifying unmet loads in simula-
tion, in: 2020 ACEEE Hot Water Forum, ACEEE, 2020.
13
... They concluded that community-scale energy storage performs better than home energy storage both economically and environmentally. Wang et al. [12] compared different carbon-emission responsive controllers for residential thermostatically controllable loads in a cold climate. They found that the carbon responsive controllers can reduce the homes' annual carbon emissions by 6.0% to 20.5% with limited impact on thermal comfort and energy costs. ...
Article
Full-text available
To combat climate change and meet decarbonization goals, the building sector is improving energy efficiency and electrifying end uses to reduce carbon emissions from fossil fuels. All-electric buildings are becoming a trend among new constructions, introducing opportunities for decarbonization but also technical challenges and research gaps. For instance, further investigation is needed to understand how the adoption of energy efficiency measures (EEMs) and distributed energy resources (DERs) in all-electric communities would affect energy consumption, carbon emissions, and grid planning. This paper presents a case study of a mixed-use, all-electric community located in Denver, Colorado. We use URBANopt™, a physics-based urban energy modeling platform to model the community and then evaluate the impact of EEMs and DERs (i.e., photovoltaics [PV], electric vehicles [EVs], and batteries) on the community’s energy usage, carbon emissions, and peak demand. The results show that adding EEMs and PV led to both energy consumption and carbon emissions reductions across all building types. However, we saw fairly limited impact of EEMs and PV on buildings’ peak demand in our case. Additionally, due to overnight EV charging activities and higher grid carbon intensity at night, the carbon emissions in multifamily buildings have a noticeable increase compared to scenarios without vehicles. Finally, the addition of batteries helped reduce peak demand by 11%–29%. The modeling workflow and evaluation methods can be applied to similar communities to evaluate their performance and the effect of integrating EEMs and DERs.
Article
Full-text available
Building retrofits have great potential to reduce CO2 emissions since buildings are responsible for 36% of emissions in the United States. Several existing studies have examined the effect of building retrofit measures on CO2 emission reduction. However, these studies oversimplified emission factors of electricity by adopting constant annual emission factors. This study uses hourly emission factors of electricity to analyze the effect of building retrofit measures on emission reduction using U.S. medium office buildings as an example. We analyzed the CO2 emission reduction effects of eight building retrofit measures that related to envelope and mechanical systems in five locations: Tampa, San Diego, Denver, Great Falls, and International Falls. The main findings are: (1) estimating CO2 emission reduction with constant emission factors overestimates the emission reduction for most measures in San Diego, while it underestimates the emission reduction for most measures in Denver and International Falls; (2) The same retrofit measure may have different effects on CO2 emission reduction depending on the climate. For instance, improving lighting efficiency and improving equipment efficiency have less impact in emission reduction in cold climates than hot climates; and (3) The most energy efficient measure may not be the most efficient emission measure. For example, in Great Falls, the most energy efficient measure is improving equipment efficiency, but the most efficient emission measure is improving heating efficiency.
Article
Full-text available
In recent years, renewable energy resources have been increasingly embedded in the distribution grids, raising new issues such as reverse power flows, and challenging the traditional distribution system operation. In order to mitigate these issues, it has been proposed to operate the distribution system more flexibly. For instance, residential buildings are ideal candidates to offer energy flexibility locally and defer avoidable and expensive system expansions. Due to advances in smart meter technologies and trends towards digitalization, it becomes more and more common that electrical appliances in residential buildings are equipped with remote communication and control capabilities. This paper aims at quantifying a flexibility envelope of actively controlled flexible buildings and at analyzing the sensitivity of flexibility levels with respect to system configuration, control strategy and objective function settings. We consider rooftop photovoltaic units, air-sourced heat pumps for space heating and domestic hot water, thermal energy storage and electric vehicles. The results show that an optimal control aiming at minimizing energy costs while limiting peak power can lead to savings of up to 25% compared to the existing rule-based control. When carbon emissions are considered in the cost function, the optimized controller leads to an emission reduction of up to 21%. The time-dependent quantification of the flexibility envelope further reveals that high and low power levels can only be sustained for a limited period, whereas medium power levels can be sustained the longest.
Article
Full-text available
Cloud data center’s total operating cost is conquered by electricity cost and carbon tax incurred due to energy consumption from the grid and its associated carbon emission. In this work, we consider geo-distributed sustainable datacenter’s with varying on-site green energy generation, electricity prices, carbon intensity and carbon tax. The objective function is devised to reduce the operating cost including electricity cost and carbon cost incurred on the power consumption of servers and cooling devices. We propose renewable-aware algorithms to schedule the workload to the data centers with an aim to maximize the green energy usage. Due to the uncertainty and time variant nature of renewable energy availability, an investigation is performed to identify the impact of carbon footprint, carbon tax and electricity costin data center selection on total operating cost reduction. In addition, on-demand dynamic optimal frequency-based load distribution within the cluster nodes is performed to eliminate hot spots due to high processor utilization. The work suggests optimal virtual machine placement decision to maximize green energy usage with reduced operating cost and carbon emission.
Article
Full-text available
As energy prices increase along with a societal desire for reducing the impact of global warming, it is common for campus decision-makers to consider the life cycle and carbon emissions of future and existing buildings. This work proposes a technique for leveraging observed energy and weather data to represent internal loads for heating, cooling, and occupancy of different building types to simplify the analysis of multiple buildings. Template energy profiles and modular energy models simulate the operation and energy usage of buildings with different HVAC equipment types. This approach makes it possible to evaluate multiple buildings from a broader campus-level perspective. This work proposes a discrete multi-objective optimization framework to select the optimal HVAC system for each building to minimize life cycle cost and emissions. A case study for the University of Utah campus demonstrates the application of the proposed framework to consider a group of new buildings with different principal building activity and size with five different HVAC systems. Results show it is possible to reduce overall carbon emissions of the buildings by 15% while only increasing the life cycle cost by 2.4% by considering a combination of systems rather than a single HVAC system type.
Article
Full-text available
An optimized heat pump control for building heating was developed for minimizing CO 2 emissions from related electrical power generation. The control is using weather and CO 2 emission forecasts as inputs to a Model Predictive Control (MPC)—a multivariate control algorithm using a dynamic process model, constraints and a cost function to be minimized. In a simulation study, the control was applied using weather and power grid conditions during a full-year period in 2017–2018 for the power bidding zone DK2 (East, Denmark). Two scenarios were studied; one with a family house and one with an office building. The buildings were dimensioned based on standards and building codes/regulations. The main results are measured as the CO 2 emission savings relative to a classical thermostatic control. Note that this only measures the gain achieved using the MPC control, that is, the energy flexibility, not the absolute savings. The results show that around 16% of savings could have been achieved during the period in well-insulated new buildings with floor heating. Further, a sensitivity analysis was carried out to evaluate the effect of various building properties, for example, level of insulation and thermal capacity. Danish building codes from 1977 and forward were used as benchmarks for insulation levels. It was shown that both insulation and thermal mass influence the achievable flexibility savings, especially for floor heating. Buildings that comply with building codes later than 1979 could provide flexibility emission savings of around 10%, while buildings that comply with earlier codes provided savings in the range of 0–5% depending on the heating system and thermal mass.
Article
Electrification and the growth of distributed energy resources (DERs), including flexible loads, are changing the energy landscape of electric distribution systems and creating new challenges and opportunities for electric utilities. Changes in demand profiles require improvements in distribution system load models, which have not historically accounted for device controllability or impacts on customer comfort. Although building modeling research has focused on these features, there is a need to incorporate them into distribution load models that include DERs and can be used to study grid-interactive buildings. In this paper, we present the Object-oriented, Controllable, High-resolution Residential Energy (OCHRE) model. OCHRE is a controllable thermal-electric residential energy model that captures building thermal dynamics, integrates grid-dependent electrical behavior, contains models for common DERs and end-use loads, and simulates at a time resolution down to 1 minute. It includes models for space heaters, air conditioners, water heaters, electric vehicles, photovoltaics, and batteries that are externally controllable and integrated in a co-simulation framework. Using a proposed zero energy ready community in Colorado, we co-simulate a distribution grid and 498 all-electric homes with a diverse set of efficiency levels and equipment properties. We show that controllable devices can reduce peak demand within a neighborhood by up to 73% during a critical peak period without sacrificing occupant comfort. We also demonstrate the importance of modeling load diversity at a high time resolution when quantifying power and voltage fluctuations across a distribution system.
Article
In this study, optimal carbon-neutral designs are evaluated for grid-connected communities that include net-zero energy (NZE) homes in Boulder, CO. Specifically, the economic and environmental benefits are assessed for residential communities with various mixtures of both NZE and non-NZE homes. Optimization techniques based on life cycle costs including both capital costs and operating costs are used to design NZE homes as well as carbon-neutral communities. Using both energy efficiency measures and rooftop PV systems, the analysis estimates the capital costs required for NZE homes. Moreover, optimally sized distributed renewable systems are determined to achieve carbon-neutral operation for various types of residential communities. First, the impact of occupancy behavior in designing NZE homes as well as carbon-neutral communities is evaluated using three use patterns of appliances (i.e., refrigerator, television, dishwasher, and clothes washer) as well as domestic hot water. Then, different fractions of NZE homes are considered when designing carbon neural communities. The analysis indicates that occupant behavior can significantly affect the design of NZE homes as well as the capital costs to achieve this design. For instance, good behavior can result in 21.28% capital cost savings while bad behavior can result in a 10.42% increase in capital cost. Moreover, the analysis indicates while the communities made up partially or fully of NZE homes can attain carbon-neutral target with lower distributed generated (DG) capacities than non-NZE communities, they require high total capital costs at least based on current costs for distributed renewable technologies and utility electricity prices. Finally, it is found to be more cost-effective to share distributed power systems for communities rather than individual homes with their own rooftop PV system to attain carbon-neutral design.
Article
In the schematic design stage, accounting for the embodied carbon emissions (ECEs) from buildings during design optimization is indispensable for saving energy and reducing emissions from the source. Since the traditional inventory analysis method is more widely applied in the mature construction drawing design stage or with a detailed bill of quantities, developing a method suitable to provide design feedback during the schematic design stage is urgently needed. The key issues of the method that need to be solved are the lack of design information and the difficulty in obtaining quantities. To address the limitations, this paper proposes a BIM-based calculation method that takes the "building element" (BE) as the basic unit, and the accuracy is verified by a frame-structure case study. For this method, it is imperative to create carbon emission factors (CEFs) of the BEs as the basis of calculation combined with the quantities extracted from the model schedules by encoding, which can significantly contribute to the calculation. The results show that this method has good accuracy and high efficiency and can identify potential hotspots of ECEs, which conforms to design process and habits and provides design feedback function for carbon emission reduction (CER). At the schematic design stage, the application of this method can quantitatively evaluate the ECEs and optimize the design from the BE level so that the concept of low carbon design can be truly put into practice and the energy savings and emission reductions of buildings can be controlled as early as possible.
Article
The introduction of dynamic electricity pricing in residential markets has created the possibility for residential electricity consumers to reduce their electric bills using battery energy storage systems (BESSs) for load shifting and/or peak load reduction. While there are numerous system designs and model formulations for minimizing electric bills under dynamic prices the use of these systems has the potential to cause an increase in emissions from electricity generation. The increase in emissions is linked to the difference in fuel mix of marginal generators throughout the day as well as inefficiencies associated with energy storage. In this work a multi-objective optimization model is designed to balance the competing goals of minimizing electricity costs for the home owner as well as minimizing carbon dioxide (CO2) emissions from the operation of a BESS under dynamic prices. A total of 22 different regions in the US are analyzed. Optimizing only for energy cost resulted in an annual increase of CO2 emissions in all but two regions ranging from 70 to 2200 kg per household. The multi-objective model when using a social cost of carbon of 42 $/ton can be used to economically reduce these additional emissions in most regions by anywhere from 49 to 1450 kg of CO2 per year.
Article
This paper investigates the optimal retrofit solutions for residential communities to be carbon neutral in Barcelona, Spain. In particular, three different sizes for the residential communities are considered depending on the number of housing units: small, medium, and large. First, the best cost-effective energy efficiency measures (EEMs) have been determined and applied to all the buildings of the communities, in order to reduce their overall energy demand. The retrofit optimization results demonstrate that combinations of passive and active strategies would achieve source energy use saving and electrical peak demand savings up to 57% and 29%, respectively. Then, a variety of renewable energy generation systems has been explored to determine the optimal grid-connected technologies for meeting the electricity demands while achieving annual carbon neutral conditions for the residential buildings. The analysis shows that wind turbines can reduce the cost of electricity generation up to 21.5% compared to the current utility prices (LCOE value as 0.247 €/kWh) and electricity sell back policies. The results indicate that for PV systems to compete with wind turbines, their capital costs need to be lower by at least 40% compared to the current level costs.