D2 - Detailed Information: ID and Title

agerasiaetherealAI and Robotics

Nov 24, 2013 (3 years and 11 months ago)

93 views

Emerging Technology Evaluation Process


D2
-

Detailed

Information
: ID and Title


Page
1

of
12



Title
Non
-
intrusive Load Monitoring



Name of Person Completing this Form



Dave Kresta

Proposer Contact Information

Please list the name,
organization, and contact
information of the primary person supplying the
information for this form (usually the person who
originally proposed this emerging technology).

NEEA

dkresta@neea.org

503
-
688
-
5459 (office)

503
-
442
-
9667 (cell)


1

Detailed Description

A
class of products which

monitor a
building’s

energy usage from a single point of access
and disaggregate

energy uses into indiv
id
ual components
.

Potential applications for NILMs includes 1) load
monitoring/forecasting/research by utilities and energy efficiency organizations, 2) energy audit tool, 3)
component of an energy management system to provide actionable information. Potential sectors inclu
de
residential, commercial, and industrial, although most current efforts appear to be focused on residential
applications. Access methods and degree of intrusiveness vary, including devices which plug into standard
electrical outlets, devices requiring cu
rrent transformers (CTs) at the electrical panel, devices which are
inserted into the utility meter, and some methods which are software
-
only, utilizing SmartMeter data. Basic
operations include: 1) signal processing (for methods with hardware) is performe
d, 2) the signal is uploaded to
a computer server (typically in “the cloud”), 3) analysis using pattern matching and similar algorithms to
identify and classify individual loads, and 4) downloaded to the user into a home energy management display
or simila
r system. Approaches to identification of loads varies and is an ongoing area of research. Some
methods require significant user training (turning devices on and off to “teach” the device), whereas other
methods seek to leverage pre
-
loaded signature libra
ries that gain experience as they are exposed to more
loads. The accuracy of products varies depending on the targeted usage, from a goal of very accurate load
identification and characterization (at least as accurate as other metering technologies), to m
ethod’s seeking
to simply identify the top energy uses to 10 or 20% accuracy.








2

Standard Practice

Existing

monitoring techniques are invasive and expensive
,
requiring monitoring devices for each appliance or
plug
-
load, typically requiring thousands of dollars per home and the installation of numerous pieces of
equipment. In residential applications, whole home energy usage information is readily available alon
g with
comparative and trend data (e.g. oPower), but information on individual loads is not available making it
difficult to ascertain which loads are responsible for energy usage. Conditional Demand Analysis (CDA) is a
modeling technique which can be used

to disaggregate end uses, but this requires time intensive field studies
and surveys along with relatively large samples to support the statistical characterizations. Note that CDA
does not support real time information, but is useful only for longer ter
m studies.




3

Development Status and History

Pioneering NILM work was first introduced by George Hart in the 1980s. Enetics has offered products in this
area since 1996, based on work by Hart and EPRI. Their product is dated (e.g Win 95 interface), requires an
electrician to install at the utility m
eter, requires manual appliance identification, and does not provide real
-
time or automated load identification (although according to them they are “working on this”). Their business
model and target market is utilities, requiring an investment of $1200/m
eter, $8000 for software. New
developments in this area are focused on less intrusive monitoring, more sophisticated signal measurements
and signal processing, lower costs, real
-
time load identification, and interfaces with energy management
displays. The
se newer products are in the pre
-
commercialization phase. Companies currently developing
products in this area include (in no particular order): Intel, Emme, Belkin, Enetics, Navetas, PlotWatt,
Verdigrist
, . Many other companies are offering products which do not currently provide disaggregating
capabilities, but are focused on related areas such as metering, home area networks/home energy
management systems such as TED, PowerHouse Dynamics, Blue Line Inno
vations, Energy Hub, Tendril, and
many others.



4

Non
-
Energy Benefits

May provide real
-
time intelligence useful for O &

M activities. For example a sporadic compressor load may
indicate imminent failure, or higher than expected energy usage by an appliance could point to incorrect setup
or faulty equipment.



5

End User Drawbacks

The pioneering work in NILMs occurred in the 1980s, but nobody has successfully driven this technology into
the marketplace. Microsoft and Google have entered and exited the home energy management market, and
there are a lot of startups as well as establi
shed companies trying to penetrate this market. More volatility in
the market is to be expected, making it difficult for consumers as well as business partners to choose a way
forward. The ability of the technology to successfully differentiate energy sour
ces in a real environment still
needs to be proven through lab and field studies. Also, the information must be presented and integrated into
a system that provides actionable information (such as a home energy management display) that enables end
users to

make decisions that reduce their energy consumption. There is a danger of providing too much
information, as well as too little, which could stifle the ability of this technology to penetrate the mass market.
Research into the effectiveness of home ener
gy management systems must accompany research into the core
NILM technology to ensure energy savings benefits.



6

Energy Savings

What is the anticipated range of per
-
unit energy savings (and demand savings of this technology if relevant)
relative to standard practice? This could be expressed as a percentage of baseline energy use.
TBD.



7

Cost

Fully metering a home is cost prohibitive, running into the thousands of dollars. Pricing information on NILMs
is scarce because of the pre
-
commercialization status. EMME is a Portland company working on a product that
would cost in the $300
-
$400 range, wi
th a $400 additional piece of display equipment required.


8

Effective Life

How

long are the existing and proposed technologies expected to last?
TBD.



9

Cost Effectiveness

What is the simple payback of this technology? Other forms of cost
effectiveness (such as ROI) may also be
useful here. If actual cost savings aren’t known, this can be a qualitative assessment.

TBD.


10

Competing Technologies

What are competing technologies and products?
See #2 above. There are several products on the
market which do not disaggregate, but rather rely on low cost monitoring devices to plug in
between appliances and the wall socket. These devices communicate wirelessly to a central unit and
display energy us
es for devices which are plugged into these monitoring devices. While these
products do not contain elegant disaggregating technology, they could provide “good enough”
楮獩sh琠楮瑯⁰物m慲a⁥n敲ey⁵獥猬 p慲瑩捵c慲ay⁩ ⁡ low
-
co獴Ⱐnon
-
楮瑲畳tv攠me瑨tT⁴o⁩
co牰o牡瑥
睨w汥⁨ me⁥n敲ey⁵獡 攠慮T⁤敤 捡瑥T⁣楲捵c琠lo慤猠⡳畣(⁡猠H噁䌬⁨o琠睡w敲Ⱐov敮⤠e慳⁩a捬cT敤e
睩瑨⁴h敳攠proTu捴c⸠



11

Advocates

What organizations or individuals have concluded that this
emerging

technology has strong potential for
improved and cost
-
effective energy savings?

Some points of contact:



Dave Kresta, NEEA



Chris Holmes, EPRI


plan to address NILM in 2012



Michael Branmbley PNNL



Rich Brown, LBNL Scientist



Michael Zeifman, Fraunhofer Cent
er



Kurt Roth, Fraunhofer



12

Emerging Technology Synopsis

Non
-
intrusive load monitoring devices (NILMs) m
onitor a
building’s

energy usage from a single point of access
and disaggregate

energy uses into indiv
id
ual components
, eliminating the need for detailed sub
-
metering.
NILMs require integration into an energy management display system, enabling end users to make energy
efficiency decisions based on actual energy usage, providing insight into how their behavior and decisio
ns
impact energy usage on a load
-
by
-
load basis.


Products in this area are for the most part in the pre
-
commercialization phase, with ongoing work to drive
down costs, improve ability to differentiate loads, integration with home energy management systems
, and
minimizing “intrusiveness” of the installation.


Lab and field testing is required in order to validate/confirm operation of these devices, particularly their
ability to differentiate and accurately measure loads in a real
-
world environment.




Utility meter

AMR utility meter

Utility meter add
-
on (kWh
intervals)

Plug (house)

CT (at meter)

CT (at panel)

CT (each circuit)

Plug (device plugged into it)
Sensor

kWh/mo

kWh/minute

Current

Voltage

Real and reactive power

Harmonics

Electromagnetic noise
Signal

Detect “events" (watts)

Real and reactive power

Waveforms (i vs time, power vs
time)

Harmonics

Electromagnetic "noise"
detection
Signature

Realtime Display, whole house

Web, whole, trends

Web ,end use
-
rough est

Web, end use
-
specific

Emerging Tech

M&V

Load shapes

System impact
Use

Questions to Guide Discussion:

1.

What types of signals/ sensors/ info are we most in need of?

2.

Are there technologies ready for additional testing (is the time right; are we the right entity to do it)?

3.

What kind of research could be done in the next year?


Sensors
Signal interpretation
(eg. “database of
“signatures”)
Accuracy
(including multiple
identical loads,
multi
-
state loads)
for various
purposes
Pair with other
service (e.g. NILM
plus Behavior for
commercial
facility)
Collaborate with
EPRI, California,
Fraunhofer, etc;
Smartgrid, etc
Clarify what’s
commercially
available, costs, etc
Consider lab tests,
field tests, etc






Signals

kWh/mo

kWh/minute

Current

Voltage

Real and reactive power

Harmonics

EFI



Types of analysis

Detect “events”(identify appliance by watt draw increase)

Identify appliances via real and reactive power (e.g Refrigerator at 250W and 200 VAR)

Waveforms (current vs time, power vs time)
-

pattern matching of start up transients)

Magnitude of harmonics

Electromagnetic "noise" detection



Sensors

Utility meter

AMR utility meter

Utility meter add
-
on (reads kWh in smaller time intervals)

Plug (senses

electrical signals for whole house)

CT (at meter)

CT (at panel)

CT (each circuit)

Plug (senses energy use for appliance plugged into it)



Uses of information

Customer: In
-
home display (whole house info)

Customer: Website feedback (whole house, trends)

Cu
stomer: end use information (rough estimates)

Customer: end use information (appliance specific info)

Research:

M&V:

Power Planning:


load shapes for each end use

Power Planning:


identification of types of loads and power system implications (CFLs vs
LEDs, Power factor,
motor load characteristics)



Applications

1.

Residential

2.

Commercial



Developmental stage

1.

Commercially available

2.

Pre
-
commercialization

a.

Beta

b.

Proof of concept



Potential types of research

Sensors

Signal interpretation (eg. “database of
“signatures”)

Accuracy (including multiple identical loads, multi
-
state loads) for various purposes

Pair with other service (e.g. NILM plus Behavior for commercial facility)

Collaborate with EPRI, California, Fraunhofer, etc; Smartgrid, etc

Clarify what’s
commercially available, costs, etc

Consider lab tests, field tests, etc

Utility/consumer issues (access to AMI signals, privacy, data storage, etc)

Test the simplest and the most complex