http://www.diva-portal.org
This is the published version of a paper presented at 2019 IST-Africa Week Conference (ISTAfrica.
Citation for the original published paper:
Kelati, A., DHAOU, I B., Kondoro, A., Rwegasira, D., Tenhunen, H. (2019)
IoT based Appliances Identification Techniques with FogComputing for e-Health
In: 2019 IST-Africa Week Conference (IST-Africa Narobi, Kenya: IEEE
https://doi.org/10.23919/ISTAFRICA.2019.8764818
N.B. When citing this work, cite the original published paper.
Permanent link to this version:
http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-241293
IST-Africa 2019 Conference Proceedings
Paul Cunningham and Miriam Cunningham (Eds)
IIMC International Information Management Corporation, 2019
ISBN: 978-1-905824-63-2
IoT based Appliances Identification
Techniques with Fog Computing for e-Health
Amleset KELATI1,2, Imed BEN DHAOU4,5, Aron KONDORO1,3, Diana RWEGASIRA1,3,
Hannu TENHUNEN1,2
1
KTH Royal Institute of Technology, Sweden
2
University of Turku, Finland
3
University of Dar es Salaam, Tanzania
4
Unaizah College of Engineering, Qassim University, Saudi Arabia
5
University of Monastir, Tunisia
Email: smleset@kth.se, phd.imed.benhaou@ieee.org, kondoro@kth.se, dianasr@kth.se, hannu@kth.se
Abstract: To improve the living standard of urban communities and to render the
healthcare services sustainable and efficient, e-health system is experiencing a
paradigm shift. Patients with cognitive discrepancies can be monitored and observed
through the analyses of power consumption of home appliances. This paper surveys
recent trends in home-based e-health services using metered energy consumption
data. It also analyses and summarizes the constant impedance, constant current and
constant power (ZIP) approaches for load modelling. The analysis briefly recaptures
both non-intrusive and intrusive techniques. The work reports an architecture using
IoT technologies for the design of a smart-meter, and fog-computing paradigm for
raw processing of energy dataset. Finally, the paper describes the implementation
platform based on GirdLAB-D simulation to construct accurate models of household
appliances and test the machine-learning algorithm for the detection of abnormal
behaviour.
Keywords: e-health., home management system, Internet of Things (IoT), fogcomputing, non-intrusive load monitoring and identification (NILM), smart-meter.
1.
Introduction
Currently, 55% of the population of the world live in the main cities [1]; however, by 2050
the proportion is expected to reach 68%. Additionally, as reported by the United Nations
Population Fund, UNPF, the number of elderly people will rise to 2 billion in 2050[2].
Seniors tend to live independently and without direct interaction with assistance. However,
caregivers need to monitor the health of the elderly people. The fact that seniors have a
tendency of falling and are prone to physical injury, their lives are in constant. Those facts
called for a paradigm shift in designing sustainable and affordable health services.
A smart city is an emerging concept that aims at integrating a plethora of advanced and
smart services for the communities. The pillars of the smart-city model along with the
number of indicators are described in Fig. [3]. Smart-living incorporates factors such as
health, safety, housing, tourism, and culture.
Smart-home technologies allow for the integration and development of advanced health
services through the daily monitoring of the activities of the end-user. Selection of sensors
in the design of smart homes is one of the challenges as discussed in [4]. The fast
communication between high- and low-level nodes of sensors make smart home a vibrant
solution for monitoring the health of the occupant. Power consumption and the operation
mode of the household appliances provide meaningful information for determining and
Copyright © 2019 The authors
www.IST-Africa.org/Conference2019
Page 1 of 11
predicting of the health and cognitive status of the user. Activity of daily living, ADL, is the
best approach for determining the wellness of the person. Numerous published reports have
proposed an ADL system for e-health, e.g., [5][6], and [7].
Figure 1: EU Model for the Smart City
Home management system (HEMS) has been the focus of some research work over the
last decade [8]. The aim of HEMS is to lower the electricity bill and reduce the peak power
consumption. The integration of HEMS with the e-health is an understudied subject where
the aim is to develop algorithm and platforms that can be used to determine the health of
the end-user [9]. Home integrated health monitoring (HIHM) is a real-time platform
elaborated in [10] that measures, store and process vital health parameter of the patient.
Smart home-based health monitoring is an active research area where IoT based
technologies are used to improve and maintain the healthcare of elderly and sick people.
One of the key things that IoT has enabled has been the provision of critical contextual
information in the healthcare process. The contextual information is vital since it provides
understanding of the patient’s environment, and improves the efficiency of the health
monitoring process. This information includes data about the location, time, identity, and
environment of the patient’s surrounding. The load profiling methods on smart meter
energy consumption data can be applied for the human behavioural model. Smart meter
technology is a corner stone for monitoring, predicting, and controlling the energy
consumption of the appliances. Collecting information on the daily activity of the seniors at
home using smart meter provide a cost-efficient solution for e-health [11] [12] [13]. Here,
the smart meter dataset information includes consumer’s energy consumptions, the
appliances usage information such as duration of appliances usage is an important factor for
remote monitoring e-health system by promoting the independent living of seniors or
patients with self-limiting health status.
There are a wide variety of smart home-based health monitoring systems that have been
proposed in the literature [14]. In general, each of these systems has three main
components: sensors, communication, and processing. Sensors are the devices that
physically interact with the patient's environment and collect relevant health information to
be used for decision-making. There are strong research efforts in both academia and private
sector to develop new sensor devices that can be worn or embedded into patient’s bodies
and collect information that is more accurate. The communication component is the whole
infrastructure that allows collected data to be transferred and stored for further processing.
Different communication protocols for Home Area Networks (HANs) are being proposed
to facilitate robust and efficient transport of vital health information.
Copyright © 2019 The authors
www.IST-Africa.org/Conference2019
Page 2 of 11
A non-intrusive load monitoring and identification (NILM) system is a technique, which
has been widely used for appliance recognition and load forecasting. NILM can be used to
collect ADL of the occupant. This work surveys recently published algorithms on nonintrusive load monitoring and describes an architecture for earlier detection of dementia and
abnormal behavior. The rest of the paper is organized as follows. Section 2, describes load
modelling approaches using non-intrusive load monitoring. Section 3 states the architecture.
Section 4, surveys algorithms for appliances recognition and monitoring. The architecture
of the e-health system described in Section 4. The implementation platform is discussed in
Section 5. Section 6 describes the future research directions and concludes this work.
2.
Load Modelling
The equivalent circuit model for the determination of AC appliances is depicted in Figure 2,
is the impedance of the kth appliance.
where
Figure 2: Circuit model of AC appliances in a given house
Let
be the equivalent impedance of the circuit show above. The complex power
consumption (apparent power) is given by (1).
, (1)
Where is the average power consumption and is the reactive power consumption.
The determination of the complex power consumption is constrained with finding the
appropriate impedance model of the appliance. In the literature, AC appliances have been
grouped in three classes: predominantly resistive, predominantly inductive and
predominantly capacitive [15].
The constant impedance, constant current, and constant power (ZIP) have been widely
used to model AC appliances [16]. The ZIP model is referred to as the polynomial model,
and it is static. The core methods for determining ZIP coefficients are shown in (2) and (3).
(2)
(3)
Where
and
are, respectively, the rated active and reactive power. The ZIP
coefficients for the active power are , , and . The rest are the ZIP coefficients for the
reactive power [17].
GridLAB-D, a defacto and open source simulation tool for power system developed by
the US department of energy, uses ZIP load models. The ZIP equations for the active and
reactive power are given in, respectively, (3) and (4).
(3)
(4)
Copyright © 2019 The authors
www.IST-Africa.org/Conference2019
Page 3 of 11
is the phase angle of constant impedance fraction,
is the phase angle of
Where
is the phase angle of constant power fraction, is the percent
constant current fraction,
of load that is constant impedance, is the percent of load that is constant power, and is
the percent of load that is constant impedance.
The ZIP models are static and hide paramount information that can be used to determine
both the active and reactive power of home appliances. To address the shortcomings of the
ZIP models a multi-state approach can be used to model the end-user load as advocated in
[18]. The transition between states renders the system dynamic, in which a state can be
modeled using ZIP or physical model.
In order to represent accurately the load characteristics of end user appliances, there
have been attempts of developing new load models for modern equipment that appear in
residential and industrial settings. These attempts fall under two categories: measurementbased and component-based. The measurement-based approach involves a direct collection
of power measurement from loads that are subjected to varying voltage conditions. The
component-based approach involves the analysis of components that constitute a load to
derive the model that describes it. Both categories of approaches lead to a ZIP model that
estimates the power consumption of electrical appliances under different voltage
conditions. [19] develop new load models for new types of customer equipment using a
component-based approach. Several types of loads were tested in a lab environment under
varying voltage conditions. Power was obtained from the grid and the voltage was varied
using a variable autotransformer. The research develops a constant impedance-currentpower (ZIP) model for these new devices. The new models facilitate the more accurate
estimation of load profiles of customers with these devices under different voltage
conditions. The component-based approach is also used to develop new load models for
residential and industrial equipment [20]. They tested a variety of modern household’s
appliances and industrial equipment to develop ZIP models that describe their steady-state
behavior. Based on a survey conducted in homes and businesses available in New York
City, different types of loads that accurately reflect reality was collected and assembled in a
lab environment. These devices were analyzed and categorized into different classes
depending on their operational characteristics. They were subjected to different voltage
conditions in a lab environment and power measurements were collected. Derived models
were then validated against actual measurements of loads in the power system.
Load signature approach has been proposed as an alternative to the ZIP model. The
chief idea for load signature is to determine the feature of an AC appliance. A
comprehensive review on this topic is reported in [41]. Signatures can be determined at the
micro-level (sampling rate less than 1 sample/ cycle) and macro-level (sampling rate higher
than 1 sample/ sec) [39]. Traditionally, active and reactive power are used to determine the
load signature [21], the authors of [Signature1] added six more signatures: eigen value,
current waveform, harmonics, instantaneous power waveform, instantaneous admittance
waveform, and switching transient waveform. A total of 27 home appliances have been
characterized using the seven features. Although, accurate, these seven models are
computationally heavy and cannot be used for real-time monitoring. Additionally, the
current waveform (CW) is not unique for some appliances. To address these shortcomings,
CW has to be decomposed into active and non-active current [40].
In summary, three techniques are widely considered for appliances modelling. These
techniques are: ON/OFF, finite state machine (FSM), and continuously variable.
Table 1summarize the appropriateness of each model [21].
Copyright © 2019 The authors
www.IST-Africa.org/Conference2019
Page 4 of 11
Table 1: Appliances Modeling
Model
ON/OFF
FSM
Continuously variable
3.
Sample Appliances
Toaster, light bulb, water pump, coffee
machines, etc.
Processor-controlled appliances, washing
machines, dishwasher, refrigerator, etc.
Light dimmers, variable-speed drive, heat
pumps, etc.
Algorithms for Appliances Identification and Monitoring
In the past, non-intrusive load monitoring, NILM, has been used for load forecasting, price
regulations, appliance designs, etc., [21]. Both manual and automatic NILMs have been
used. The former requires a setup period in which a lookup table is created to recognize the
signature of the appliance that latter automatically builds the signature table and needs
priory information such as type and name of the appliance.
The NILM records the appliances usage time and usage frequency [11]. The assumption
was to detect a behavioral change using electric load intelligence [39].
Detecting human activities and sudden changes can be utilized for the health provider’s
attention[22]. The activity prediction depends on the analysis of the usage of multiple
appliances with the Bayesian network. The time series based multi-label classifier has been
used for the prediction of the home energy usage using appliances at homes [23] to get
results with high-level accuracy. The approach taken was the k-means for clustering
algorithms using the pattern analysis that related to sudden changes and the Bayesian
networks for activity prediction related to health activities based on historical data.
A non-intrusive load monitoring and identification, NILM, system is used for
investigating the electricity usage of the residents [24]. The machine learning method can
analyze and classify appliances with a unique signature. It can distinguish anonymous
appliances and arrange relevant actions to enhance the classification accuracy. Machine
learning algorithm strategy can identify appliances with an arbitrary accuracy if the power
is measured every 10 sec or less [25]. The same approach was used by [26] in India to come
up with four feature which is important. These are load signatures of active and reactive
powers, harmonic components and their magnitudes. The algorithms have three
assumptions: single-phase appliances, the algorithm is based strictly on the specifications of
IEW (Internal Electrical Wiring) and depend on steady-state values of features and Table 3
summarize the signature-based approaches for appliances identification.
Table 2: Load Monitoring Techniques are Categorized with Two Main Methods
Method
Intrusive Load
Monitoring
(ILM)
Description
The measurement is taking
inside the house by connecting
each appliance physically to the
measurement device and are
connected to a hub or smart
meter. The method has an
accurate measurement but
expensive
[27].
Non- intrusive
Load
Each appliance the operation
state and energy consumption is
Copyright © 2019 The authors
Method
Supervisory Control and data
acquisition (SCADA).
Identification is done with an
algorithm that help to
distinguish each appliances a
known feature energy
consumption. Direct or indirect
monitoring [27] [28], based on
power demand [29]. of a typical
appliance from the data
Appliances identification is by
monitoring the change in active
www.IST-Africa.org/Conference2019
Page 5 of 11
Monitoring
(NILM)
depend on the power
measurement of the aggregate
load of the building. [30], [31]
[32].
and reactive power and by
setting the operating state of the
appliances. Hidden Markov
Model (HMM) [33] [35],
identification by switch-on and
off the appliances and changes
in the pre-processed from
steady-state or transient to detect
the appliances [34].
Table 3. Methods for Signatures on Appliances Identification
Signatures
Energy Disaggregation
Fundamental Frequency
Harmonic frequency
Steady-State Signatures
Description
Depend on the individual appliance in the total load [32],
switch ON and OFF power change in steady and the
event is the basis for frequency or harmonic frequency
signatures [30].
Measuring current, power, or the normalized admittance
power is the change as signatures [30].
Appliances produce harmonic current and for some
appliances not easy to distinguish using only reactive or
real power the signature is based on transient and steady
state [35].
Connected to the states on the operating level [30].
x reactive and active and power and current
waveform, include temporal motif mining
approach [36], combined HMM load model [37]
x Current waveform characteristics, Different
current measurements used for load identification
[35] and the method is often used on reactive and
real power and transients power [38].
The energy management system is equipped to provide reliable monitoring and
appliances identification. Smart grid technology enables consumers to know and manage
their electrical loads and smart meter enable the communications between the electrical
utility, energy demand, and pricing. Individual appliances load is measured with the
connected plug load meters that are resalable for real-time energy management. The
measures can identify the on /off state, the change in the state and the information on the
peak power consumption of the appliances [34].
Figure 3. Different Appliances Types Based on their Energy Consumption Pattern [35].
Copyright © 2019 The authors
www.IST-Africa.org/Conference2019
Page 6 of 11
The appliances power consumption can be measured with low frequency to obtain an
accuracy of classifications. The measurements of the appliances power consumption can be
described with the known basic state of mode as describe in the literature [25].
To consider accurate appliances identifications and classifications, the measured power
consumption will undergo different steps: pre-processing, feature extraction, splitting,
post-processing, and training & classification. For stable and accurate appliance
identification, considering the methodology of machine learning algorithms for processing
will be beneficial. The process follows the procedure, which include collecting the readings
of the appliances feature from the dataset and performing feature extraction to figure out
and understand the power consumption dataset to be used for the appliances identifications.
The real-time classification of appliances is set according to energy consumption and
can be used in diverse applications scenarios. One application is associated with human
activity observation and detect abnormalities that related to health conditions. A smart
meter is connected to the household for real-time load profiling of the power consumption
of the appliances through the remote controlling to apply for several and different
applications. The measurements and extraction of the power consumption pattern of each
appliance have a unique pattern that depends on the data acquisition. The complicated
behavior of the appliances acquires different activity and functionality and external
circumstance influence the process of identification and label the appliances signature.
4.
System Architecture
Load profiling and recognition necessities a constant learning process in order to improve
efficiency and reduce the false alert ones. The system architecture as depicted in Figure 4,
consists of the following entities. A smart plug through which a home appliance gets the
power. The smart plug is an embedded system that houses sensors along with the
communication and DSP processing blocks. The plug can also control the operation of the
appliance to support the demand response scheme. The plugs within the house are
connected through a various communication link to the smart meter. The smart meter
aggregates the data from all plugs, process it and ferry it to a server for health-care
diagnosis and other home energy management functions.
Figure 4. Smart Home Architecture
Copyright © 2019 The authors
www.IST-Africa.org/Conference2019
Page 7 of 11
New architectural designs based on fog computing that utilize existing communication
devices in homes are also being proposed to improve performance and deployment. The
processing system includes all backend devices and algorithms that enable the automation
of health-based decisions. New architectural designs based on fog computing that utilize
existing communication devices in homes are also being proposed to improve performance
and deployment. The processing system includes all backend devices and algorithms that
enable the automation of health-based decisions. New approaches based on big data and
machine learning are being used to detect and predict health anomalies. Figure 5. shows,
the three layers of fog hierarchies; the nodes at bottom layer gather the smart meter
measurements from home appliances and facilitate an early warning that needs to detect the
abnormal status immediately by communication to the cloud platform. The architecture is
composed of the following three layers. The Smart gird layer has a smart meter and
responsible for communicating the appliances and smart meter, and other smart grid
devices. The Fog layer communicates smart meter measurements and data to the cloud by
separating different types of user data. The cloud layer data can be stored for further
analysis mainly on identification or some prediction and sent to Fog computing. It is
possible the detailed and private data are normally kept in cloud servers.
CloudComputing
(storage,Analysis
PredictionAnalysis)
Power consumption
data analysis with
Fog Computing
CLOUD LAYER
AnalysersStörige
FOG LAYER
FogNode(Analytics&Storage
IntegrationServices&Powermanagement
SMART GRID LAYER
SMART METER
Figure 5. Energy Consumption Analysis in Fog / Cloud Computing
5.
Implementation Platform
There exist multiple candidates (hardware and software) for the implementation of the
home-based e-health system. First, the power consumption dataset of the appliance
gathered by using wireless sensor networks. At this stage the communication protocol is on
the Smart-plug with an IoT can be used at this level. The purpose lies on the detecting the
behavioural changes on the appliances usage of the consumers at home in the case of the
appliances regularly monitored as they are used by the occupants in the house. The activity
Copyright © 2019 The authors
www.IST-Africa.org/Conference2019
Page 8 of 11
with the interaction of the appliances is stored in the computer or laptop in order to
determine the status of health in our case for instance for elderly people. The information
from the sensor and smart plug is equipped with how, when or for how long the interaction
and activity of the appliances. The data processing part of the smart-plug is accomplished
with the help of a microcontroller such as Arduino. The smart meter is the device that
gathers data from the smart plug. There are multiple platforms for implementation of the
smart meter. To cope with the heterogeneity of communication protocols, IoT middlewares
are received an ample interest for the realization of the smart-meter devices. The smart
meter measurement data describe timely dependant load profiling for load forecasting and
apply the NILM method for household appliances from consumers.
To test the NILM for e-health the implemetation is done with GridLAB-D which is a
promising simulation platform that has abundant accurate models of the home appliances.
The features of GridLAB-D have a user friendly and the open source code that is available
for edit as needed. The platform has a multi-state time variant scheduling load models that
are flexible load profiling simulation related to the end user appliances voltage. The
platform can run on multiprocessor machine and use multi-agent techniques and allows
deploying new control algorithms. Our implementation for load modelling is described
briefly in section 2. The implementation uses the GridLAB-D IEEE 13 node stranded for
the current, voltage control and load flow of the and energy storage for the appliances
measurement on its simulation.
Table 4 summarizes hardware and software tools to simulate the proposed prototype
Table 4: Implementation platforms
Device/tool
Current sensors
Microcontrollers
Middleware
Power system simulator
6.
Functions
Measures the instantaneous
power consumption and
RMS value
Filtering, calculation of ZIP,
communication, appliance
identification,
smartmetering
Security, heterogeneity
Simulation, modelling
Candidates
ACS712
Atmel
89S52,
AVR465,
MCF51EM256,
TMPM411F20
KAA
GridLAB-D and
Matlab
Conclusion
E-health using electric load intelligence is a promising technology for home-based
healthcare system. The idea behind the system is to identify the type and operation mode of
the electric appliance using aggregated energy measurement, then determine the health of
the occupant using a predefined mode of operation. This paper has surveyed the latest
published algorithms for appliances identifications. It also discussed the load modelling
techniques. Dynamic ZIP model is the most suitable candidate for load modelling. The
construction of the models needs a priori knowledge of the appliance to measure the
significant parameters such as rated active power and reactive power, current, and
impedance. Finally, the present work devised a fog-based architecture for the e-health
system. Our future work will focus on devising an algorithm for appliances identification
using machine-learning techniques and the algorithm implementation on FPGA to obtain
efficient identification of hardware architecture.
Copyright © 2019 The authors
www.IST-Africa.org/Conference2019
Page 9 of 11
Acknowledgement
This work is funded by the Swedish Funding Agency, SIDA and supported through iGrid
Project at the Division of Electronics, School of Electrical Engineering and Computer
Science, Royal Institute of Technology (KTH).
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
United
Nation.
(2014).
World
Urbanization
Prospect.
[Online].
Available:
https://www.un.org/development/desa/publications/2018-revision-of-world-urbanization-prospects.html.
N. K. Suryadevara, S. C. Mukhopadhyay, R. Wang, R. K. Rayudu and Y. M. Huang, "Reliable
measurement of Wireless Sensor Network data for forecasting wellness of elderly at smart home," 2013
IEEE International Instrumentation and Measurement Technology Conference (I2MTC), Minneapolis,
MN, 2013, pp. 16-21.
Rudolf Giffinger et al., “Smart cities Ranking of European medium-sized cities”, Research Institute for
Housing, Urban and Mobility Studies (OTB), October 2017, www.smart-cities.eu,http://www.smartities.eu/download/smart_cities_final_report.pdf.
V. Nathan et al., "A Survey on Smart Homes for Aging in Place: Toward Solutions to the Specific Needs
of the Elderly," in IEEE Signal Processing Magazine, vol. 35, no. 5, pp. 111-119, Sept. 2018.
H. Pazhoumand-Dar, M. Masek and C. P. Lam, "Unsupervised monitoring of electrical devices for
detecting deviations in daily routines," 2015 10th International Conference on Information,
Communications and Signal Processing (ICICS), Singapore, 2015, pp. 1-6.
S. Rahimi, A. D. C. Chan and R. A. Goubran, "Usage monitoring of electrical devices in a smart home,"
2011 Annual International Conference of the IEEE Engineering in Medicine and Biology Society,
Boston, MA, 2011, pp. 5307-5310.
N. Noury, M. Berenguer, H. Teyssier, M. Bouzid and M. Giordani, "Building an Index of Activity of
Inhabitants From Their Activity on the Residential Electrical Power Line," in IEEE Transactions on
Information Technology in Biomedicine, vol. 15, no. 5, pp. 758-766, Sept. 2011.
N. Javaid, I. Khan, M. N. Ullah, A. Mahmood and M. U. Farooq, "A Survey of Home Energy
Management Systems in Future Smart Grid Communications," 2013 Eighth International Conference on
Broadband and Wireless Computing, Communication and Applications, Compiegne, 2013, pp. 459-464.
A. Yassine, S. Singh and A. Alamri, "Mining Human Activity Patterns From Smart Home Big Data for
Health Care Applications," in IEEE Access, vol. 5, pp. 13131-13141, 2017.
J. S. Kim, B. O. Kim and K. S. Park, "Development of HIHM (Home Integrated Health Monitor) for
Ubiquitous Home Healthcare," 2007 29th Annual International Conference of the IEEE Engineering in
Medicine and Biology Society, Lyon, 2007, pp. 363-365.
J. Liao, L. Stankovic, and V. Stankovic, ``Detecting household activity patterns from smart meter data,''
in Proc. Int. Conf. Intell. Environ. (IE), vol. 6. Jul. 2014, pp. 7178.
A. Yassine, A. A. Nazari Shirehjini and S. Shirmohammadi, "Smart Meters Big Data: Game Theoretic
Model for Fair Data Sharing in Deregulated Smart Grids," in IEEE Access, vol. 3, pp. 2743-2754, 2015.
H. Ghayvat, S. Mukhopadhyay, B. Shenjie, A. Chouhan and W. Chen, "Smart home based ambient
assisted living: Recognition of anomaly in the activity of daily living for an elderly living alone," 2018
IEEE International Instrumentation and Measurement Technology Conference (I2MTC), Houston, TX,
2018, pp. 1-5.
A survey on health monitoring systems for health smart homes, Haider Mshali a, Tayeb Lemlouma b, *,
Maria Moloney c, Damien Magoni a, International Journal of Industrial Ergonomics, pp. 26-56.
A. Ridi, C. Gisler and J. Hennebert, "A Survey on Intrusive Load Monitoring for Appliance
Recognition," 2014 22nd International Conference on Pattern Recognition, Stockholm, 2014, pp. 37023707.
Alcalá JM, Ureña J, Hernández Á, Gualda D. Assessing Human Activity in Elderly People Using NonIntrusive Load Monitoring. Sensors (Basel). 2017;17(2):351. Published 2017 Feb 11.
"Standard load models for power flow and dynamic performance simulation", IEEE Transactions on
Power Systems, vol. 10, no. 3, pp. 1302-1313, 1995. Available: 10.1109/59.466523.
K. Schneider, J. Fuller and D. Chassin, "Multi-State Load Models for Distribution System Analysis",
IEEE Transactions on Power Systems, vol. 26, no. 4, pp. 2425-2433, 2011. Available:
10.1109/tpwrs.2011.2132154.
F. L. Quilumba, W.-J. Lee, H. Huang, D. Y. Wang, and R. L. Szabados, “Load model development for
next generation appliances,” in 2011 IEEE Industry Applications Society Annual Meeting, 2011.
A. Bokhari et al., “Experimental Determination of the ZIP Coefficients for Modern Residential,
Commercial, and Industrial Loads,” IEEE Trans. Power Delivery, vol. 29, no. 3, pp. 1372–1381, 2014.
Copyright © 2019 The authors
www.IST-Africa.org/Conference2019
Page 10 of 11
21. G. Hart, "Nonintrusive appliance load monitoring", Proceedings of the IEEE, vol. 80, no. 12, pp. 18701891, 1992. Available: 10.1109/5.192069.
22. A. Yassine, S. Singh and A. Alamri, "Mining Human Activity Patterns From Smart Home Big Data for
Health Care Applications," in IEEE Access, vol. 5, pp. 13131-13141, 2017.
23. K. Basu, V. Debusschere and S. Bacha, "Appliance usage prediction using a time series based
classification approach," IECON 2012 - 38th Annual Conference on IEEE Industrial Electronics Society,
Montreal, QC, 2012, pp. 1217-1222.
24. Rossier, Florian, et al. “Near Real-Time Appliance Recognition Using Low-Frequency Monitoring and
Active Learning Methods.” Energy Procedia, vol. 122, 2017, pp. 691–696.
25. Srinivasan Iyengar, David Irwin, and Prashant Shenoy. “Non-intrusive model derivation: automated
modeling of residential electrical loads”, In ACM e-Energy. 2016
26. Sunil Semwal, R S Prasad, Pradeep Juneja “Identifying Appliances using NIALM with Minimum
Features”, December 2014 International Journal of Electrical and Computer Engineering 4(6):909-922.
27. M. Zeifman, K. Roth, “Non-Intrusive load Monitoring: Review and outlook”, Journal IEEE Transactions
on Consumer Electronics, Volume 57 Issue 1, February 2011 , Page 76-84.
28. Yoshimoto, K., Nakano, Y., Amano, Y., and Kermanshahi, B.. “Non-Intrusive Appliances Load
Monitoring System Using Neural Networks.”, IEEE Proceedings of International Symposium on Circuits
System, Vol. 2, pp.183-194.
29. Breed, J.A., and Delport, G.J. (2001). Non-Intrusive Load Monitoring of Residential Appliances.
Proceeding of the 7th Conference on Domestic Use of Energy, pp.33-38.
30. Ehrhardt-Martinez, K. Donnelly, K.A. and John. A. (2010). Advanced Metering Initiatives and
Residential Feedback Programs: A Meta-Review for Household Electricity-Saving Opportunities.
Technical Report E105 for American Ccouncil for Energy Efficient
31. Zoha, A., Gluhak, A., Imran MA., and Rajasegarar, S. (2012). Non-Intrusive Load Monitoring
Approaches for Disaggregated Energy Sensing: A Survey,” Open Access Sensor ISSN 1424-8220,
pp.16838-16866.
32. Z. Wang and G. Zheng, "Residential Appliances Identification and Monitoring by a Nonintrusive
Method," in IEEE Transactions on Smart Grid, vol. 3, no. 1, pp. 80-92, March 2012.
33. Phial, H. (1998). Non-Intrusive Appliance Load Monitoring Systems Based on a Modern kWh-Meter,”
VTT Energy, Thesis Submitted for Official Examination for The Degree of Licentiate in Technology,
Espoo.
34. Barges, M.E., Goldman, E., Matthews, H.S., and Lucio Soibelman, L. “Enhancing Electricity Audits in
Residential Buildings with Nonintrusive Load Monitoring.” Journal of Industrial Ecology, Vol. 14, No.5,
pp.1-15, Oct. 2010.
35. Y. Du, L. Du, B. Lu, R. Harley and T. Habetler, "A review of identification and monitoring methods for
electric loads in commercial and residential buildings," 2010 IEEE Energy Conversion Congress and
Exposition, Atlanta, GA, 2010, pp. 4527-4533.
36. Shao, H., Marwash, M. and Ramarkrishnam, N. (2012). A Temporal Motif Mining Approach to
unsupervised Energy Disaggregation: Application to Residential and Commercial Buildings. 1st
International Workshop on Non-Intrusive Load Monitoring, pp, 1-7.
37. Kolter, J. Z. and Jaakkola, T. (2012). Approximate Interface in Additive Factorial HMMs with
Application to Energy Disaggregation,” Proceeding of the Fifteenth Conference on Artificial Intelligence
and Statistics, pp.1472-1482.
38. M. Ito, R. Uda, S. Ichimura, K. Tago, T. Hoshi and Y. Matsushita, "A method of appliance detection
based on features of power waveform," 2004 International Symposium on Applications and the Internet.
Proceedings., Tokyo, Japan, 2004, pp. 291-294.
39. J. Liang, S. K. K. Ng, G. Kendall and J. W. M. Cheng, "Load Signature Study—Part I: Basic Concept,
Structure, and Methodology," in IEEE Transactions on Power Delivery, vol. 25, no. 2, pp. 551-560, April
2010.
40. T. D. Huang, W. Wang and K. Lian, "A New Power Signature for Nonintrusive Appliance Load
Monitoring," in IEEE Transactions on Smart Grid, vol. 6, no. 4, pp. 1994-1995, July 2015.doi:
10.1109/TSG.2015.2415456.
41. N. Sadeghianpourhamami, J. Ruyssinck, D. Deschrijver, T. Dhaene, C. Develder, "Comprehensive
feature selection for appliance classification in NILM", Energy and Buildings, vol. 151, pp. 98, 2017.
Copyright © 2019 The authors
www.IST-Africa.org/Conference2019
Page 11 of 11