Systems, Man & Cybernetics - October 2017 - 29

The IoT and Indoor
Climate Control
Connected and distributed sensing provides ample
oppor tunities to optimize the way we control the
indoor climate in our buildings. Indoor climate control
in buildings entails the control of HVAC equipment and
the electric lighting (and rarely the motorized blinds),
and it accounts for more than 70% of the energy use in
buildings [19], [20]. With connected and distributed
sensing, we can continuously tune the operating schedules and set points to enhance occupant comfort and
energy efficiency.
Model-based predictive control (MPC) is one of the
commonly studied methods to optimize the operation of
HVAC equipment [21], [22]. In MPC, a data-driven model of
a building inputs near real-time short-term forecasts for
disturbances (e.g., weather forecasts, plug-in equipment
loads). The predictions of the model are then employed

Annual HVAC WOI per 1,000 m2

to under take optima l control
assist them in making preventive
decisions. For instance, Gunay [2]
maintenance decisions.
Beyond fault
Automated fault detection and
deployed an MPC algorithm in--
diagnoses, the data
diagnostics is an emerging re--
side commercial building controllers to ca lculate the optima l
sea rch f ield seek ing sca lable
from control and
warm-up duration of each thermethods to optimize the mainteautomation networks
mal zone in the morning. As a
nance of HVAC equipment and
result of this algorithm, on severecontrol infrastructure [9]-[13].
may have the portents
ly cold winter days, the heating
Fault de--tection and diagnostics
of the failure of critical
equipment was set to start earlier
methods utilize sensor and actuequipment such as airthan on moderately cold days. As
ator data collected from building
another example, Huchuk et al.
automation and control networks
handling unit supply
[23] employed an ensemble Kal(BACnets) to detect and isolate
fans and pumps.
man filter-driven thermal network
HVAC faults automatically [7], [8].
model of a perimeter office space
Be--yond fault diagnoses, the data
to determine the control sequence
from control and automation netfor automated window blinds,
works may have the portents of
with the objective to optimize the passive solar gains
the failure of critical equipment such as air-handling unit
through windows. The literature on MPC in buildings is
(AHU) supply fans and pumps. Thus, the faults that can
vast (e.g., [24]-[28]). Variations of MPC applications in
affect the comfort and productivity of many occupants
buildings include the model complexity (e.g., number of
can be predicted in advance (e.g., prognosis).
resistance capacitances in the thermal network models),
In addition, the data from submeters for steam (or
the parameter estimation approach (e.g., ensemble Kalhot water), chilled water, and electricity, when normalman filter, unscented Kalman filter, and extended Kalized with local weather data, can be used to establish a
man filter), and the building subsystem to be controlled
performance benchmark of a building cluster [14]-[16].
(e.g., the blinds, zone, and system- or plant-level HVAC
A building cluster is a group of buildings typically served
equipment). In all cases, near real-time weather foreby the same central heating and cooling plant (e.g., unicasts from the Internet and other sensor inputs from
versity campuses). Innovative data visualization tools for
control networks were blended to execute better conoperators can be developed to compare the energy pertrol decisions.
formance of different buildings and identify energy perAdaptive control is a method to continuously tune
formance deviations in time [17], [18]. In a similar
the operating setpoints and schedules based on recurfashion, graphical user interface designs can be tailored
ring occupancy and adaptive occupant behavior patterns.
such that occupants are informed about the energy
impact of their behaviors (e.g., thermostat setpoint
changes). For example, sever a l com mercia l home
thermostats display insights about the relative energy
50
performance (e.g., the top 20% in the community).
Line-Fit for Academic and
Administrative Buildings Only
WOI = 0.037 EUI
R 2 0.85

40
30

Line-Fit for All Buildings
WOI = 0.023 EUI
R 2 0.59

20
10
0
0

200 400 600 800 1,000 1,200
EUI (kWh/m2)

Scale for the Bubble Plot
Construction Year 1960
Construction Year 2010

All Buildings
Administrative and
Academic Buildings

Figure 2. The relationship between the WOI and the

annual EUI of 44 buildings in a university campus in
Ottawa, Canada.

O c tob e r 2017

IEEE SYSTEMS, MAN, & CYBERNETICS MAGAZINE

Table of Contents for the Digital Edition of Systems, Man & Cybernetics - October 2017