Transcript Eric_Final_Presentat.. - School of Electrical Engineering and

Smart Environments for
Energy Reduction
Eric Torunski
Introduction
• Electricity usage has increased steadily and will continue to
increase for the foreseeable future.
• Electricity production creates carbon dioxide as a by-product,
which causes global warming.
Introduction

The energy grid must provide enough energy for peak demand,
when people turn on their air conditioners, stoves & televisions
when they come home from work.

Ottawa Hydro will introduce time-of-use electricity rates

http://www.hydroottawa.com/smartmeter/index.cfm?lang=e&template_id=357
• Energy usage for a typical home:
• http://www.energystar.gov/index.cfm?c=products.pr_pie
Solution
• Create a control system that helps reduce energy
consumption in the home.
o Provide real-time feedback to the user of current energy
consumption.
o Automatically power down, or enter low-power state for
items that are not being used: television, lights, furnace, air
conditioner.
o Reduce power consumption of items when total
consumption is above a maximum level: turn down air
conditioner, lower home heating, water heater.
o Also, try to schedule energy usage to minimize cost.
Previous work – resource
usage minimization
• The problem of home power control is similar to
computer device control. The problem is to try and
predict when a resource will be used next.
• Similar to screen savers, powering down hard drives,
even virtual memory in a computer.
• Concept of “Break-even” cost. This is the amount of
energy saved by powering down a device versus the
power needed to restore its state.
Artificial Intelligence control
• Fuzzy logic control – system is pre-programmed to
classify inputs into pre-determined membership
categories, with corresponding outputs.
• Neural networks – systems slowly learn the correct
output for given inputs.
• Software Agents – An agent acts on behalf of
something in the system and agents negotiate a
solution. They can be pre-programmed or learn
patterns over time.
Previous work – neural network
• Fuzzy logic and neural network control:
o Neural Network House in Colorado
– A project at the University of Colorado converted an old 1920's
school house to test their ACHE (Adaptive Control of Home
Environment) control system.
– The first goal is to meet user's comfort needs, and the
secondary goal is energy conservation.
– Light and temperature levels are kept at a minimum level, and
then raised to user's comfort levels only when they are about
to enter a give room.
– Hot water tank only heats enough water for predicted use.
Previous work – software agents
– Software agent control systems:
• iDorm in University of Essex, England
• iHome at University of Massachusetts
• University of Karlskrona, Sweden
– When a user enters a room or house, the user's agent
negotiates with the room's agent for lighting levels and
temperature settings.
– Over time, the user's comfort agent learns the user's
comfort preferences.
– For meetings, the agents of people attending negotiate
common settings for light and temperature.
– The user's comfort agent is tied to the house or building.
Previous work - testing
• It is hard to build and test real systems with real users.
• It is possible to simulate people living in a house, and
get “reasonable” results.
• Some papers used tracking devices on people to
record their movements for 1 week. Then, the
researchers can test their algorithms on the recorded
movements.
Previous work – user feedback
• Visual feedback from HCI community:
• By just giving visual feedback to the user, at
least 10% energy reduction is possible.
• http://tiffanyholmes.com/?page_id=28
• http://www.google.org/powermeter/
• The Watt Bot shows the user power usage by
category, although the user is responsible for
taking steps to reduce energy consumption.
Image of WattBot
Power Generation
• Vehicle to Grid
o Hybrid and electric cars will have high-capacity batteries to
move a car for up to 300km on a single charge.
o If the car is plugged in at home, why not use the battery as
an energy storage device?
o Peak load leveling: provides power when demand is above
average, recharges when demand is below.
– Schwarmstrom (Swarm power)
o http://www.treehugger.com/
o A home generator generates electricity and uses excess
heat to heat the home and water.
o Provides electricity that is much cleaner than coal, and is
cheaper and safer than nuclear power.
– Renewable Energy: Solar, Wind, etc.
Proposed new work


Artificial Intelligence controller for home that meets
the user's environmental preferences, while
minimizing energy usage.
Mobile device is the interface to the environment,
and provides feedback to the user. This makes the
user aware of current power usage, and perhaps
compares it to average usage. The mobile device can
also stores the user's comfort preferences (in the
form of software agent) for entering new
environments.
Proposed new work .. 2
• Artificial Intelligence controller lowers electricity
usage cost by using alternate energy sources when
available (vehicle to grid, solar, wind, home
generator).
• Schedule some tasks for when the electricity rates
are cheap (laundry, water heating, recharging car).
• Reduce peak energy usage by using stored
electricity in batteries.
Example of AI Controller and Peak Reduction
JAVA Agent Framework
•JADE (Java agent development environment). Java library
for distributed agents. Implements FIPA agent
specification so it will work other compliant agent
frameworks.
•Facilities for sending messages between agents,
transferring agents amongst platforms.
Distributed Agents
•Every controllable object has an agent. For example,
lights, heaters, window blinds.
•Each room has an agent to interact with objects in rooms.
•The user has a comfort agent to inform the room agent
what the lighting and temperature settings should be. The
room agent picks the objects in the room which can
provide those settings using the least amount of energy.
•When there is no one in the room, the room agent
maintains a minimum power state.
Comfort settings
•The user interacts with the room using mobile device to
turn up temperature, change lighting levels.
•The mobile device learns over time what the preferred
settings are, and can inform new smart environments.
•The mobile device can challenge the user by lowering
temperature and see if the user notices.
Conclusion
•By lowering energy usage, the home owner can save
money. With visual feedback, the user becomes more
aware of how much energy is used.
•By lowering peak energy demand, the most polluting
electricity generating stations (burning coal) can be shut
down. This reduces greenhouse gas emissions and
environmental pollution.
•Hybrid car batteries can store energy from intermittent
alternative energy sources (wind, solar).
References
1. M. Boman et al, Energy Saving and Added Customer Value in Intelligent
Buildings.
2. M. Boman et al, Saving Energy and Providing Value Added Services in
Intelligent Buildings: A MAS Approach.
3. M. Mozer, The Neural Network House: An Environment that Adapts to
its Inhabitants.
4. H. Hagras et al, Creating an Ambient-Intelligence Environment Using
Embedded Agents.
5. T. Holmes, Eco-Visualization: Combining Art and Technology to Reduce
Energy Consumption.
6. D. Petersen et al, WattBot: A Residential Elecricity Monitoring and
Feedback System.
7. Tae-Jung Yun, Investigating the Impact of a Minimalist In-Home Energy
Consumption Display.