HeatScore: Intelligent Thermostat System

Intelligent Thermostat System
A Report Submitted in Partial Fulfillment
of the Requirements for SYDE 362
Al Amir-khalili, 20198393
Chris Best, 20200501
Kyle Morrison, 20203836
Vladimir Naoumov, 20198926
Faculty of Engineering
Department of Systems Design Engineering
Supervisor: Professor P. Fieguth
April 6th, 2009.
Course Instructor: Professor S. Birkett
When it comes to heating in the home, there is a balance to be struck between
user comfort and energy use. The HeatScore system seeks to reconcile these two
metrics by creating and extrapolating a user-specific usage pattern, and using that
pattern to determine when to heat the rooms, and more importantly, when not to
heat the rooms. By only heating the rooms that the user actually uses, the system
can reduce the energy use of the home, which in turn benefits the environment. The
system is ingenious in the sense that it automatically develops this usage pattern
based solely on selfish inputs at specific times that the users want the temperature
to change. Furthermore, the user interface is designed to be intuitive in order to
encourage inputs, and thus have a more accurate picture of the user’s needs. A
custom algorithm is used to transform these inputs into a heating plan for the future,
and by corollary to recognize when it can reduce the heat usage of the home. The
final design of the system meets the various criteria of usability, functionality and
performance defined by the group, and so the initial prototype is considered to be a
success. Further areas of investigation include adding addition parameters (in other
words, knowledge) to the algorithm, more functionality to the physical system (such
as user feedback), and more universality in the form of cooling or humidity control.
List of Tables
List of Figures
1 Background
Problem Area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Environmental Motivation . . . . . . . . . . . . . . . . . . . . . . . .
Prior Art and Novelty . . . . . . . . . . . . . . . . . . . . . . . . . .
2 Requirements
User Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Functional Requirements . . . . . . . . . . . . . . . . . . . . . . . . .
Performance Requirements . . . . . . . . . . . . . . . . . . . . . . . .
3 Design Analysis
User Interface: Thermostat . . . . . . . . . . . . . . . . . . . . . . . .
Design Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Innovative Aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Further Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Algorithm: The Relevance Scheduler . . . . . . . . . . . . . . . . . .
Design Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Innovative Aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Further Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4 Testing and Evaluation
Functionality Testing . . . . . . . . . . . . . . . . . . . . . . . . . . .
Thermostat Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Central Unit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Prototype Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
User Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Environmental Assessment . . . . . . . . . . . . . . . . . . . . . . . .
Evaluation of System Benefits . . . . . . . . . . . . . . . . . . . . . .
Lifecycle Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5 Summary and Recommendations
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Recommendations for Further Designs . . . . . . . . . . . . . . . . .
List of Tables
User Requirements & Corresponding Performance Requirements . . .
Functional Requirements & Corresponding Performance Requirements
List of Figures
Layout of the System: Cross Section of the House . . . . . . . . . . .
Interface of the Thermostat . . . . . . . . . . . . . . . . . . . . . . .
Aggregated Temperature Schedule Given One Input . . . . . . . . . .
Aggregated Temperature Schedule Given Series of Inputs . . . . . . .
Sample Profile for Programmable Thermostat . . . . . . . . . . . . .
Sample Profile for HeatScore system . . . . . . . . . . . . . . . . . .
Problem Area
The problem of heating control in low to medium density residential housing is well
studied. At the highest level, it involves a tradeoff between the preferences of the
human occupants in terms of temperature, and the drive to minimize energy usage.
With any type of heating the energy used to heat a home has both cost and environmental impact associated with it. Existing systems typically involve a thermostat
control by which a user sets a preferred temperature or a temperature programme.
The thermostat then controls a heating mechanism which aims to achieve that temperature.
The system described in this report aims to improve upon existing home heating
control systems in the comfort/energy trade-off. Only the heating control aspect of
home heating will be considered; the mechanisms used to generate heat is outside the
scope of this project except insofar as the control system needs to interface with them.
By building intelligence into the control portion of existing heating techniques, the
design aims to minimize energy usage while preserving or improving human comfort
levels in a way that is not possible with existing solutions.
This is accomplished by designing a system that uses extremely simple, intuitive
user input to infer a complex heating pattern. The assumption is made that a person
always knows whether the temperature around them is warmer or cooler than they
would prefer at a given instant in time. The system then takes this preference as an
input and corrects the room temperature as quickly as possible, but it also records
multiple inputs over time in order to develop a predicted heating preference pattern.
This allows for a complex, near-optimal heat distribution plan to be created without
the need for sophisticated planning on behalf of the user.
Environmental Motivation
According to Natural Resources Canada, space heating accounts for 58.9% of residential sector energy use in Canada in 2006. This accounted for 39.6 of the greenhouse
gas emissions due to energy use in this sector [1]. This is clearly an area where a
small increase in efficiency could result in large positive environmental impact.
One way to accomplish an increase in efficiency would be to improve the efficiency
of heating mechanisms themselves by using more efficient furnaces, insulating homes,
and similar techniques. To a large extent this has been done. According to natural
Resources Canada, the proportion of medium and high efficiency heating systems
compared to normal efficiency ones in single detached residences increased steadily
between 1999 and 2006 [2].
This is progress, and it shows that people are willing to adopt technology that
can present energy savings, which translates to both environmental and cost savings.
However, thermal efficiency can only go so far. Even with extremely high thermal
efficiency, there is room for further environmental savings by intelligently controlling
the heating system, in particular turning it down or off when it is not needed. That
is the benefit that will be pursued by this project.
Prior Art and Novelty
The relevant patents found while searching for novelty fell into three general categories. The first were patents which relate to normal programmable thermostats, such
as “Programmable time varying control system” [3] and “Programmable Thermostat” [4]. This technology is widely commercially available. In fact, 31% of Canadian
houses have one, although one quarter of these do not use the programmability [5].
The second group of relevant patents involves improved thermostat control, with no
significant modification to the usage pattern, such as “Wireless Programmable Thermostat Unit” [6]. The closest to the design from this group was “Thermostat with
one button programming feature” [7], which upon closer inspection only allows a
temporary modification to a fixed preset. Finally, there are systems which allow detailed monitoring of energy usage, such as US patent #6786421 [8]. The technology
here would complement the design very well, but in no way cover the design.
What is novel and ingenious about the approach presented in this report is the process by which a complex heating plan is developed automatically. Taking extremely
simple user inputs, for which the user does not have to consider factors beyond their
immediate needs, and translating a set of these inputs into an efficient, granular,
predictive heating pattern is the key feature. No patent or commercial product discovered during the group’s research leveraged this approach.
To generate the system requirements, one must examine the problem to determine
what the user would expect from a solution generated from the user requirements.
These, in turn, are explored to see how the designed system should be addressing each
system requirement on a conceptual level, generating a set of functional requirements.
The final step is to see how the system performance can be evaluated, generating a
set of performance requirements. These three sets of requirements are presented in
the following sections.
User Requirements
The requirements below are essential to ensure that the users are satisfied with how
the proposed system addresses the problem of high cost (both monetary and environmental) of heating the home during the cold season.
• Interface must be intuitive to the user.
• The system must minimize the number of required user inputs once the learning
cycle has been completed.
• One has to be able to install the device into the house without any house
modifications or significant modifications to the existing heating system inside
the house.
• The system must improve heating efficiency without introducing user discomfort.
• Must have a central override feature, that would turn the system off and disable
heating for example when the users go on vacation.
• The system must not create unsafe situations by overheating or over-cooling.
Functional Requirements
The following set of functional requirements were developed to ensure that the system
will be able to satisfy the user defined requirements.
• The system must be able to measure the temperatures in each of the controlled
zones of the house.
• The thermostats must be able to transmit user inputs and temperature readings
via wireless to the central controller on several frequencies to ensure it would not
cause interference with the variety of other signals present in different houses.
• The central unit must be able to store and process the input signals to generate
a temperature pattern for each room.
• The system has to be able to be retrofitted onto existing household heating
• The individual thermostats must fit into a typical lighting switch wall slot and
be powered by the existing 110V line.
Performance Requirements
Having analyzed the user and functional requirements, the following set of performance requirements was developed. Tables 2.1 and 2.2 demonstrates how end user
requirements and system requirements translate into measurable performance requirements for the system, requirements that the final product must satisfy.
Table 2.1: User Requirements & Corresponding Performance Requirements
Intuitive interface
The interface must be understandable by 95% of users
who spend at most 5 minutes looking at the user manual.
Minimizing the num- The number of inputs required to maintain a developed
ber of inputs
user pattern is at most 1 per day
Ease of installation
Compatible with 80% of existing wall light switch slots.
Compatible with 80% of existing heating systems.
Both room and central units operate are powered by
Improving efficiency
The system must generate at least 5% energy savings
versus alternative systems available to the users.
Must have a central This requirement can be viewed directly
override feature
Table 2.2: Functional Requirements & Corresponding Performance Requirements
Measuring tempera- Temperature is read in rooms to at least 0.5 degree pretures
cision. If greater precision can be achieved at the same
cost, the better solution should be implemented.
Transmitting signals The system must be able to transmit the signal with at
from thermostats to most 5% data loss within a 25 meter range.
the central unit
The system must be able to operate on 10 different frequencies.
The operation of the The unit has enough memory to store 12 months of input
central unit
and temperature profiles for up to 10 zones.
Must have an uptime of 99% with no data loss in case
of power failure
These requirements were used both during the design of the final product and for
creating the testing sequences described in Chapter 4.
Design Analysis
The design of the proposed system is divided into its two main components to facilitate the process of analysis. The two components are the thermostat and the trend
extrapolation algorithm.
User Interface: Thermostat
The thermostat is the main mode of interaction between the user and the system. As
specified in the requirements, the interface needs to be simple, easy to use, and universal (see section 2.3). These requirements improve the marketability of this system
by increasing the number of environments in which this system may be implemented.
Design Process
A method is desired to create a link between the user-friendliness and user-comfort
requirements. Once this link has been made, another method can be designed to
minimize the environmental impact without any compromises to the user-comfort
and user-friendliness.
Current thermostats set the room temperature as given by the user in degrees.
Studies have shown that comfortable room temperatures for humans are not only
functions of temperature. Comfortable room temperatures are in fact defined to be
a region as a function of both humidity and temperature [9]. Therefore, setting an
absolute temperature is to be avoided. From this fact, the most reasonable way to
determine the desired temperature is to measure it in relation to the current room
temperature. The only similarity between the proposed thermostat and a conventional one is the inclusion of a temperature sensor inside the thermostat. Other than
that, the entire interface is to be changed to allow for relative inputs.
The initial idea behind the layout of the thermostat was to use a simple two button
interface. As described in the approach section of the background, the two buttons
behave as a means to capture the user’s desired temperature relative to the current
room temperature. One button tells the system that the user desires the room to
be warmer and the other tells the system that a lower temperature is desired. The
method of parsing button presses is described in the algorithm section of this chapter
(section 3.2). Now that a link has been made between comfort and user-friendliness,
the design may be refined iteratively according to other requirements.
Existing households with poor insulation are the main targets of this system.
Therefore, the thermostat must be designed in such a way that allows for it be
retrofitted into existing heating systems without extensive modification to the current
heating system and the house. Thermostats are ideally installed in accessible locations and require a means to communicate with the heating system of the house. In
situations where zone heating is used, it is desired for a thermostat to be installed in
each of the divided heating zones. To address these concerns, the thermostat design
allows for it to be just a drop-in replacement for a light switch. The reasoning is
that light switches are normally located in accessible locations and each heat zone
typically has an independent light switch. In order to avoid the need to install additional wiring (connecting the thermostat to the heating system) each thermostat is
equipped with a wireless transmitter. The user inputs are captured and are sent to a
controller, which in turn operates the existing heating system.
The transmission of data from a thermostat via WiFi has already been done
before [6]. Attempts were made to use a new and innovative means of communication.
However, as is the case with new innovations, a lot of research is required to justify
Layout of the System
Living Room
Furnace Controller
Figure 3.1: Layout of the System: Cross Section of the House
the method’s selection. An alternative method of communication is to use Power Line
Communication (PLC). However, forcing data to be sent over power lines has not been
subject to a lot of testing in this domain. For instance, PLC modems are especially
big. Installing an entire modem inside a light switch may conflict with building
codes. As a result, a decision is made in favour of wireless communication. Wireless
transmitters are very small, require very little power, and are easily interfaced with
using commercial micro-controllers.
At this point, a prototype of the thermostat is created incorporating the features
discussed so far. Its functionality is subjected to testing in tandem with a crude
prototype of the Relevance Scheduler (section 3.2). It is observed that the design, in
its current form, has failed to incorporate a feedback mechanism. Therefore, the next
iteration of the design process is to implement a feature that enables the proposed
thermostat to provide the user with feedback regarding the behaviour of the heating
system. This feature is extremely important since it eliminates situations where inputs
are duplicated. Input duplication mostly occurs when users are unaware of how the
system is responding to the previous input. This issue is magnified in the case of
house heating. House heating is a special case because it takes quite some time for
the system to heat the room to the desired temperature. Within the time it take for
the system to heat up the room, other users who are unaware of the previous input
may inadvertently increase the input temperature beyond what is actually desired.
The proposed solution to this problem involves the inclusion of an array of feedback
LEDs on the thermostat.
Final Interface Design
Recessed push buttons used to avoid
false positives. The red button increases
the target temperature.
LED Array representing relative difference
between current room temperature and
target temperature.
The blue push button decreases the
target temperature.
Figure 3.2: Interface of the Thermostat
The LED array, in effect, displays the difference between the current room temperature and the desired temperature to which the system is heating the room. As
the temperature in the room moves towards the desired, the number of LEDs turned
on decreases. The LEDs only activate when a button is pressed, and remain on for a
period of ten seconds before turning back off. This avoids unwanted illumination at
This concludes the main iterative design process. Some of the other final design decisions (including button shape and placement) are addressed in the Further
Analysis section.
Innovative Aspects
There are three key features that make this design unique and innovative. These
aspects, as summarized from the previous section, are as follows:
• The system heats the house using relative desired temperatures
• The thermostat captures inputs with an intuitive two button input
• The system eliminates the need for house modifications by using wireless communication.
Further Analysis
Additional analysis was carried out to make a decision regarding what type of buttons
would be ideal for the thermostat. After discussing the design with an expert [10],
the following critical design criteria were developed:
• The buttons must be recessed to avoid the accidental button presses while trying
to toggle the light switch.
• Buttons must be colored in contrast to aid users’ intuitive distinction of the
functionality of the two.
• The buttons must be shaped like triangles and positioned above and below the
light switch to further juxtapose the two buttons for people who are visually
These criteria are integral in insuring that the design is user-friendly. Therefore,
they are taken into consideration in finalizing the design (See Figure 3.2).
Algorithm: The Relevance Scheduler
The Relevance Scheduler behaves as an interface between the thermostat and the
heating unit. It parses the binary inputs passed on to it from the two buttoned thermostat and generates a schedule based on pre-programmed heuristics. This schedule is then passed on to the heating unit, similar to inputs received from a generic
programmable thermostat. Most of the design decisions made herein are based on
intuitive heuristics. Future iterations of this design require explicit knowledge of the
heating system being used in the application and the dimensions of each heat zone.
Design Process
The first step in designing such an algorithm is to determine how the inputs from
the thermostat are going to be parsed. The binary input determines the direction in
which the temperature is going to be stepped by the heating system. It was intuitively
decided that the number of consecutive button presses must be parsed to represent, in
a relative way, just how much the user wants the temperature to change. Each button
press is mapped by the algorithm to represent a ‘Temperature Step’. The number
of button presses are also capped at seven Temperature Steps to avoid overshoot.
Intuition also determined that the size of each Temperature Step must depend on
the current temperature. Consider the situation where the current room temperature
happens to be 10◦ Celsius and a button is pressed to increase the temperature in the
room. Intuitively, it is obvious that the Temperature Step is going to be rather large.
However, the Temperature Step must be parsed to be relatively smaller when the
room temperature is 25◦ Celsius and a button is pressed to increase the temperature.
After the button presses have been parsed into a stepping direction, magnitude
of steps (temperature), and number of steps, the next step is to design a feature
which extrapolates a heating schedule by taking in these inputs, the current room
temperature, and the date. The main challenge in designing such algorithm is to
design it in such a way that it will be able to generate a schedule given sparse number
of inputs. The first step in generating such a complicated feature is to break it
down and start by programming in heuristics which make intuitive sense. Typically,
people’s schedules on weekdays differ from weekends. Therefore, if an input is received
to increase the temperature to 23◦ at noon on Monday, it is very likely for this event
to recur on Tuesday, whereas it is less likely for this behaviour to manifest on the
weekends. Similarly, a perturbation on Monday this week is more likely to happen on
the following Monday than other days of the current week. All of these basic features
were implemented in the first iteration of the algorithm
With regards to these intuitive notions, the algorithm takes an input and then
aggregates the schedule by repeating this given input every 24 hours (Figure 3.3).
However, the repetitions are scaled with respect to their similarity to the day in
which that input was originally received; hence the name Relevance Scheduler.
Figure 3.3: Aggregated Temperature Schedule Given One Input
These basic heuristics were then compiled into a crude demonstrable prototype.
The demonstration of the prototype to potential users revealed the opportunity for
further heuristics to be programmed into the prototype. For example, inputs received
at night (between 8pm and 6am) should be held for longer durations than those
received throughout the day. The reasoning behind this is that if the nightly inputs
were held for short period of time (e.g. 2 hours) the user might have to wake up in
the middle of the night to add more inputs to the system. This would violate the user
comfort requirement, and thus it was also added as a heuristic (Figure 3.4). These
improvements formed the basis for the second iteration.
As mentioned before, most of the parameters discussed in this section (i.e. hold
Figure 3.4: Aggregated Temperature Schedule Given Series of Inputs
time and temperature step magnitude) are case specific. Testing is required to develop
values depending on the method of heating, which is beyond the scope of this design
(refer to Chapter 5).
Innovative Aspects
Unlike the design of the thermostat, which had many key innovative features, the
implementation of a Relevance Scheduler is novel in and of itself. Although the
thermostat design is new and innovative, its new features alone are not enough to
fill the niche that this system is targeting. The proposed thermostat alone cannot
deliver any energy savings. The novel features of the thermostat act as selling points
by allowing the system, as a whole, to be easily implementable.
It is the Relevance Scheduler’s unique ability to generate an optimized temperature
profile, given minimal inputs, which makes the HeatScore system innovative. With
this innovative feature, the HeatScore system stands out above the competition.
Further Analysis
Once the basic features have been derived, more functionality may be built into the
algorithm to make it more robust by accounting for pathological cases where outliers
occur among the inputs. Examples of such outliers (an event that only happens
infrequently) are social events. During such events, the temperature only needs to be
adjusted at the time of the input and does not necessarily need to be replicated. The
solution to this issue was to give each preset temperature a decaying lifespan.
As mentioned in the iterative design process, the scheduled temperatures are programmed in relation to the time and day at which the input was given. The relations
are now changed to be programmed in such a way that the signal will only be repeated
for a maximum of three weeks. This will ensure that the outlier has been eliminated
in three weeks. Once this feature was implemented, some interesting unexpected
benefits were observed to exist within this feature, the most important of which was
increased energy efficiency.
This method also increases the efficiency by driving the temperatures to a preset
minimum, which will further reduce heating, and yield an increase in energy savings.
This effect is discussed in detail within the Environmental Analysis section of this
Testing and Evaluation
The testing of the created system can be broken down into several individual components. First of all the functioning of the created system must be tested, which
will be discussed in the first section. The second component is testing the algorithm
that generates the temperature profiles and controls the heat production. This is a
vital component of the created system, and thus a prototype was developed to help
understand and further improve the algorithm. Further rationale for the prototype
as well as the results of the prototype testing will be presented in section 2.2. Finally,
even if the system is functioning the way it was designed, the environmental impact
has to be assessed, which includes the environmental benefits the system offers as
well as a basic lifecycle analysis of the created thermostat.
Functionality Testing
Functionality testing will address most of the requirements that were discussed in the
prior sections of the report. Since our system is made of individual components, each
of these components separately.
Thermostat Units
Requirement: Compatible with 80% of existing wall light switch slots.
To ensure this requirement is met, the design has to fit into a typical slot
advocated by the National Electrical Code in US and Canada, which is at least
3 x 2 x 2.75 inches [11]. As all residential light switches have to comply with
the code, this would ensure that the requirement is met.
Requirement: The wall unit must be powered by 110V.
Since the system was designed to meet the requirement, it has been met.
Requirement: Temperature is read in the rooms to at least 0.5 degree
The created system relies on using prefabricated components such as the temperature sensor. Thus it is believed that the supporting documentation for
those components is reliable. The selected temperature sensor would then be
chosen to meet the requirement at lowest possible cost.
Central Unit
Requirement: Compatible with 80% of existing heating systems.
Statistics Canada suggests that roughly 80% of households use either hot air
furnace or electric baseboards both of which are compatible with the system [1].
In the case of the hot air systems, flaps controlling the airflow to individual
rooms. In the case of electric baseboards, they can also be controlled room by
room using the signal from the central unit.
Requirement: Powered by 110V and has a central on/off switch.
Since the system was designed to meet the requirement, it has been met.
Requirement: The system must be able to transmit the signal with at
most 5% data loss within a 25 meter range on 10 different frequencies.
Once again, the system relies on prefabricated components such as a wireless
communicator in this case. Since it is more complex than a temperature sensor
and is crucial to the system performance, once the desired component is chosen
it would be tested in a basement to attic communication scenario to ensure
consistent transmission over a course of a day.
Requirement: The unit has enough memory to store 12 months of input
and temperature profiles for up to 10 zones with no data loss in case
of power failure.
The data storage component of the system would be chosen to satisfy these 2
settings, with no additional testing being required.
Prototype Testing
A crucial component of the created system is the algorithm that controls temperature
profile generation as well as the extremely simple user interface. In fact, these two
things are the key differences setting the created system apart from others available
on the market today. Due to their importance a prototype modeling a house with 3
rooms has been created to allow further tuning of the design of these 2 vital components. Since a real house was not available to implement the final solution, a testing
environment has been created where the rooms are modeled. These rooms will be
heated by the system-controlled furnace, and will cool naturally due to the outside
temperature. The thermostat will also not include the LED user feedback that will
be present in the final design. Although not very advanced, this prototype allowed
some basic user and algorithm testing to be performed.
User Interface
The requirement to be addressed was the “The interface must be understandable by
95% of users which spend at most 5 minutes looking at the user manual”. This was
tested during the symposium where the prototype was made available to the general
public, ranging from students to professors to people who just happened to be there.
Having only a brief explanation as to what the system did, they were then allowed
to interact with the prototype and see the results of their actions displayed back to
The general feedback that was received was very positive. Most users found the
system to be easy to understand and even when users have attempted to crash the
system by generating random unnatural inputs, the prototype behaved as expected
without generating any unpredicted outputs. The general feedback that was that
our requirement of ease of use was well achieved. The only negative response that
received was the fact that the prototype had no feedback to the user, something that
the final design already addresses.
There are several requirements that the algorithm had to satisfy: first of all the number of inputs had to be at most one per day when the learning cycle is complete.
Second, the algorithm had to produce behaviors that are compatible with what the
user thinks would happen, in order to improve user friendliness. And finally, the
generated temperature profiles should have better heating efficiency. Improving efficiency will be addressed in a later section, so initially the discussion will focus on
how well the system satisfies the first requirement.
This requirement depends on the end user preferences and comfort levels. For
example, a person who is comfortable when the temperature is +/- 3 degrees of what
they wanted will input less than a person who needs it to be +/- 1 degree. Thus,
to test this requirement, a sample schedule is generated for a mock user that goes to
work on weekdays and spends half his time at home during the weekends. Once the
initial profile was set up, it is clear that the design has successfully maintained the
preferences for the user, who use at most one input per day for another 2 weeks. This
was done to a +/-2 degrees comfort level, meaning that input was only provided to the
system if the user was assumed to be in the room and outside their comfort zone. Tests
have shown that 1 input per day is more than enough to keep the temperature profile
within the comfort level. Thus, on a prototype level, it is clear that the system meets
the functional requirements. More testing may be necessary when the real system is
build and installed in a house to verify these findings in real life situations.
The concept of the algorithm being user friendly was tested during the symposium
when the prototype was available to the general public. While the users were interacting with the system, their feedback was collected to see if the system does what
the users are trying to make it do. Once again, the response here was mostly positive;
the system generated the temperature profiles in a way that was not unnatural to
the users. Some of the more interested people asked for details about the operation
of the algorithm that generates the temperature profiles and were supportive of the
design solutions that were implemented. There were also several ideas for potential
improvement that were suggested, such as modifying the algorithm to have different
memory behavior for different rooms, or being able to give the system an initial profile
at initialization different than just a minimum safe temperature at all times in every
room. This feedback means that although the system is already performing to the
specifications that were generated, there is room for future improvement.
Environmental Assessment
Evaluation of System Benefits
The last requirement that needs to be satisfied, thus justifying the designed system,
would be the impact on the heating costs and consequentially on the environment.
Although a numerical estimate of the impact would be impossible without any
statistical testing, there are several observations that can be made. One can compare
the temperature profiles that would be generated by a programmable thermostat to
those generated by the designed system. In Figure 4.1 and Figure 4.2 below, a user’s
preference in the living room is modeled.
It can be seen from the two profiles that although both profiles generate comfortable temperatures in the mornings and evenings when the user requires the house to
be heated, the area under the graph in Figure 4.2 is significantly smaller, indicating
that less heating is required.
Figure 4.1: Sample Profile for Programmable Thermostat
Figure 4.2: Sample Profile for HeatScore system
An additional advantage of the developed system is that the energy savings demonstrated above would be further increased due to the fact that the system generates
individual heating profiles for each zone in the house, meaning only the used area
of the house would be heated with the HeatScore system. Thus, while the profile
generated by the programmable thermostat would remain the same for the bedroom
as for the living room, the profile generated by our system would be different. Since
the user does not require the bedroom to be heated at times other than when he/she
is there, the profile would be able to adjust to heating the bedroom less than the rest
of the house during the day, resulting in an overall savings for the user.
Another advantage that the system provides the user with is being able to adjust
to any changes in the users schedule without the users having to reprogram the whole
system. If, for example, the user changes jobs and now comes back home one hour
later than they did before, the input that set the house to be heated up earlier would
no longer be relevant and would decay out, resulting in a new heating schedule that has
the users new preferences accounted for. In comparison, the user of a programmable
thermostat would have to remember to reprogram it, which may or may not happen.
Since they are not experiencing any discomfort at any period of time, they would
likely just be heating their house for an hour longer than they really need until they
remember to reprogram it.
Our system also offers additional benefit to users with children. Since it is very
simple to use, the operation can be explained even to kids, who would most likely
not be allowed to use the conventional thermostats. This would allow the concept of
“heating where you need, when you need” to be in effect even if the only people at
home are children. In comparison, a programmable thermostat would most likely be
programmed to heat the house from the time children come home, as the adult user
would not really know what part of the house needs heating and would not want to
inconvenience their children.
Finally, the most important advantage of using the system is the built in decay in
the temperature profiles that happens over time. This means that outlier inputs do
not affect the temperature profiles and that the profiles adjust to the users schedule.
There is also another impact that the decaying temperatures have: the user only
provides input to the system when they are outside the comfort zone, and thus the
temperatures are kept at the minimal comfortable level all the time. In comparison,
for a programmable thermostat the user is forced to guess their comfortable temperature ahead of time and considering the fact that the comfortable temperature varies
within several degrees, on average the preset would be higher than the resulting profile
temperature generated by our system.
One can see that the energy savings generated by the system depend on the users.
This makes it impossible to pinpoint the exact value of savings generated by the
designed system. Yet, overall the reasoning presented above makes it clear that the
system provides many opportunities for the user to reduce their energy use. Since
even a 1◦ C reduction in heating can provide up to 10% reduction in energy use [12],
it can be clearly seen that the designed system would definitely meet the 5% energy
reduction criterion set previously.
Lifecycle Analysis
The life cycle of the system would consist of four main stages : manufacturing, installation, maintenance and disposal.
Manufacturing stage The system would be made of prefabricated components
put together. For example, a thermostat unit would be made from a power converter,
a temperature sensor, a wireless transmitter and an array of LEDs mounted onto a
typical light switch. And the central processing unit would be an extremely simple
computer with a storage device, a wireless transmitter and a power adaptor. All of
these components are presently available on the market and thus it is not expected
to see any significant environmental impact from their use or production.
Installation stage One of the key criteria for the system was the ease of installation. All the user would need to do would be to replace their light switches with
the thermostat units in the individual rooms and attach the central unit onto their
existing heating system. This process is extremely simple and would not cause any
significant amount of effort, and thus would not have any environmental impact.
Maintenance stage The device requires no maintenance during its operation.
The only time any interference is needed is if one of the components breaks down.
Since the device itself is very modular (the room units are independent) and made
from a collection of simple and easily replaceable components, the repairs would not
be a complicated job for any electrician.
Disposal stage Since the device itself is a collection of various electronic parts,
they could be easily reused or recycled using the existing electronics collection facilities
available. There are several companies that specialize in this kind of services, such
as Vancouver based Electronics Recycling Canada or Toronto based ADL Process
Inc. Thus the disposal of the system can be done efficiently without producing any
environmental damage.
Summary and Recommendations
Heating makes up a large portion of residential energy use in Canada. Any reduction
in the amount of energy required to heat homes presents a significant financial savings
to users, and has a positive impact on the environment by reducing the amount of
green house gasses emitted in absolute terms. Significant progress has been made in
this area in the past decades, but there is room for further improvement by adding
intelligence to the interface and control of residential heating systems.
The design presented in this report provides an effective means to realize this
improvement. By combining extremely simple user input, and an intelligent control
algorithm which drives to minimize energy use given this input, this design offers
significant energy savings. Because it does this while maintaining or improving user
comfort levels, the value proposition to the user in terms of functionality and cost
savings is excellent. Given the user appeal of this system, the design is successful in
its end goal of being a viable means of helping the environment.
Recommendations for Further Designs
In light of the success of the preliminary design, this report recommends that further
research and design work be conducted to commercialize the system. In particular,
the following areas have not been fully covered in this report and should be addressed:
• Integration with existing heating systems. This includes the mechanics
of controlling a heating system, as well as further development of the algorithm
to take heat dynamics into account, such as heat transfer between rooms.
• Full parameter analysis. The system parameters chosen based on general
research and intuition should be fully tested and verified. Furthermore, machine
learning could be introduced to refine these parameters while online.
• Increased scope of climate control. Active cooling and humidity control
can be taken into account to provide a more complete climate solution.
• Additional Input Mechanisms. Although the fundamental input mechanism
is sound, further control can be given to the user at their option. Methods could
include a computer interface, or an in-depth master control. These should never
be required, but should supplement the simple primary input method.
• Design for Manufacture. The design should be adapted for mass production. Key objectives are low manufacture cost, ease of installation, and minimal
environmental impact.
This report concludes that the design presented herein is a success with respect to
the defined objectives. With the future improvements outlined above, the system has
the potential to be a commercial success, to improve the quality of life of its users,
and to help save the environment.
[1] Office
energy use and ghg emission by energy source.”
[Online] http:
res_on_1_e_3.cfm?attr=0. (current Feb 2009).
[2] Office of Energy Efficiency, Natural Resources Canada, “Heating system stock by building type and heating system type.”
[Online] http:
ca_27_e_3.cfm?attr=0. (current Feb 2009).
[3] B. C. Hall, “Programmable time varying control system and method.” United
States Patent 4,967,382, October 1990.
[4] M. R. Levine, J. T. Russo, A. C. Cairo, and V. H. Rigotti, “Programmable
thermostat.” United States Patent 4,606,401, August 1986.
[5] Natural Resources Canada, Survey of Household Energy Use, 2003.
[6] S. Dushane, T. Zimmerman, G. Bohm, and J. Staples, “Wireless programming
or programmable thermostat mobile unit for multiple separate controller or programmable fixed units and programming transmission method.” United States
Patent 6,196,467, March 2001.
[7] S. L. Carey, “Thermostat with one button programming feature.” United States
Patent 6,814,299, November 2004.
[8] H. Rosen, “Programmable thermostat including a feature for providing a running
total for the cost of energy consumed during a given period for heating and/or
cooling a conditioned space.” United States Patent 6786421, September 2004.
[9] Information Please, All the knowledge you need., “Winter indoor comfort and relative humidity.” [Online] http://www.infoplease.com/ipa/A0001412.html.
(current March 2009).
[10] V. Goyal, “Expert analysis:
User-centric design.” Personal INTERVIEW.
(March 11, 2009).
[11] J. Markell, “Residential Wiring to the 2005 NEC,” p. 6, 2005.
[12] B. Stats, “The Heat is On: Energy Use and Household Heating in B.C.,” 2008.
Download PDF