slides

Transcription

slides
ENLITEN PROJECT
DATA CHALLENGES
2016-03-03, 1030
Sukumar Natarajan
@suknat
OVERVIEW
•
ENLITEN project overview
•
ENLITEN hardware
•
Data characteristics
•
Data challenges
Primary Aim
To reduce carbon emissions from energy use in dwellings by
developing a low-cost intelligent home energy advisor that will
provide actionable prompts to households that they can use to save
money and energy.
CONTEXT
Estimates of savings from IHDs vary widely
2006
typically: 5-20%
small samples: 12-18%
larger samples: 2-6%
Faruqui
2010
14% with pre-payment meters
7% without
EDRP
2011
2-3% average
Darby
Why?
CONTEXT
Design: No consistent language
Type | Size | Location | Detail
CONTEXT
Literacy: No one understands Watts and kWh
Image source: http://www.whatonearthcatalog.com/graphics/products/large/CL9321.jpg
CONTEXT
Knowledge: No one knows what to do with the information
?
CONTEXT
Motivation: As designers, we conflate knowledge with motivation
High
Knowledge
Apathetic
Activist
High
Motivation
Low
Motivation
The Walking
Dead
Keen but Clueless
Low
Knowledge
CONTEXT
“Knowledge is not enough, we must apply” - Bruce Lee
information
behaviour
The Information Deficit Model
literacy
knowledge
motivation
But things are much more complex
behaviour
Overall Picture
✚
RE
router
DL
AY
E
R
Feedback and
advice
secure
cloud store +
whole building energy model
iBert
In ~40
households
ENVIRONMENTAL + ENERGY SENSING
per home
Exeter city centre
3x
1x
1x
1x
3x
iBERT: ACTIONABLE PROMPTS + VALUE FRAMING
Specific and Actionable
Not
actionable
Without value framing
With value framing
(biospheric, hedonistic, altruistic)
I have noticed that the temperature
in your home is frequently X°C.
This is unusually high.
This might require E kWh more
energy over a whole winter, in
comparison to a temperature of
21°C.
Advice: Consider lowering the
thermostat to 21°C. If you don’t
have a central thermostat, adjust
your radiators. Alternatively, try
changing your heating schedule so
your boiler operates for fewer
hours.
I have noticed that the temperature
in your home is frequently X°C.
This is unusually high.
Over a whole winter the extra
pollution from this compared to
using 21°C is equivalent to the
destruction of T trees.
Advice: Consider lowering the
thermostat to 21°C. If you don’t
have a central thermostat, adjust
your radiators. Alternatively, try
changing your heating schedule so
your boiler operates for fewer
hours.
iBERT EMBODIMENT
Data from the sensor infrastructure
Env Sensors
Energy Sensors
CO2
Temperature
Light
PIR
Humidity
Gas
PUSH by Web-API
PULL by Scripting
Navetas
PUSH by Web-API
App-Status
IP address
PULL by Scripting
WebServer
WebServer
(total)
Electricity
Home ID
(disaggregated)
User Engagement
Electricity
Cloogy DB
Energy Sensors
Structure of Array Data
{"id":SENSOR_UUID, "sensor":SENSOR_CODE, "type": SENSOR_TYPE,
"value":FLOAT_VALUE, "ip" :SENSOR_IP_ADDRESS, "timestamp“:TIMESTAMP}
“id”
Unique ID of the sensor, usually hardware information of the sensor, String
“sensor”
Numerical value assigned to each sensor (e.g. Temperature – 1, Humidity – 4, CO2 –
16 etc), Integer
Textual information of each sensor (e.g. “temperature”, “humidity”, “co2”, “light” etc),
String
“type”
“value”
Sensor reading (e.g. 21 degree temperature, 60% humidity, 850 ppm co2 level), Float
“ip”
IP address of Raspberry PI computer where all sensors are installed
“timestamp” The time and date when the entry is sent to the DB server
This is usually under 128 byte per entry
i.e. maximum 1 Mbyte (> 128 byte X 6750 entries)
per day per home could be stored in the DB
Survey data
Recruitment Installation
Main
Air con
Thermal
Feedback
Length
21 qs
41 qs
484 qs
27 qs
13 qs
15 qs
Type of questions
MCQ
MCQ
MCQ + free
text
MCQ + free
text
MCQ + free
text
MCQ
reduced
thermal
comfort
survey
energy literacy,
thermal
satisfaction,
values
demographics,
psychological
variables, build occupants'
ing chars,
thermal
occupants' atti
comfort,
tudes, values, energy habits,
energy literacy, ventilation beh
energy habits,
aviours
ventilation beh
aviours
Type of data
Demographics,
building chars,
attitudes,
photos, plans
etc.
Sensor
positioning
etc.
Return rate
100%
100%
81%*
46%
40%
100%
Format
paper
paper
paper
paper
tel calls
paper
Stored as
MySQL
MySQL
MySQL
MySQL
XLS
paper + XLS
Size (rows x cols)
200 x 21
100 x 41
53 x 484
37 x 27
80 x 13
50 x 15
* some partials
CHALLENGES!
•
Lots of data, but not problematic (~50GB total)
•
Many kinds of data
•
Multiple formats
•
Different frequencies
•
Multiple personnel involved
•
Commercial providers - do we own our data?
•
Varying levels of quality / completeness
•
Connections between datasets, but not always formalised
•
Long term storage