2005 IEEE PESC (Power Electronics Specialist Conference), 12-18 Jun 2005, Brazil. Submitted.

A Taxonomy of Load Signatures for Single-Phase Electric Appliances

Authors: K.H. Ting 1, Mark Lucente2, George S.K. Fung1, W.K. Lee1 and S.Y.R. Hui3
(1) Department of Electrical & Electronic Engineering, University of Hong Kong
(2) CLP Research Institute, Ltd. (Hong Kong)
(3) Department of Electronic Engineering. City University of Hong Kong.

Abstract:
Load signature analysis is an emerging research topic highly relevant to electric utility companies.
The knowledge of the nature and taxonomy of electric loads would provide invaluable
information for power industry to understand how their low-voltage power sources (ac mains) are
used. In this project, a new methodology in load signature analysis is reported. Using input current
waveforms and 2-dimensional voltage-current trajectories, a framework for taxonomy of electric
loads based on over 120 single-phase office and domestic electric appliances is proposed. This
research study indicates that the front-end power electronic circuits and the mechanical
characteristics of some electromechanical loads play critical roles in both load signature and
taxonomy of electric loads.
References
1. C. Laughman, K. Lee, R. Cox, S. Shaw, S. Leeb, L. Norford and P. Armstrong, Power
Signature Analysis, Power and Energy Magazine, IEEE, Vol. 1, pp.56-63, 2003
2. G. W. Hart, Nonintrusive appliance load monitoring, Proc. of the IEEE, Vol. 80, Dec, 1992,
pp.1870-1891
3. L. K. Norford, S. B. Leeb, Non-intrusive electrical load monitoring in commercial buildings
based on steady-state and transient load-detection algorithms, Energy and Buildings, vol.24,
pp.51-64, 1996
4. F. Sultanem, Using appliance signatures for monitoring residential loads at meter panel
level, IEEE Transactions on Power Delivery, Vol. 6, pp.1380-1385, 1991
5. S. Drenker, A. Kader, Nonintrusive Monitoring of Electric loads, Computer Applications in
Power, IEEE, Vol. 12, pp.47-51, 1999
6. A.I. Cole and A. Albicki, Nonintrusive identification of electrical loads in a three-phase
environment base on harmonic content, Proc. of the 17th IEEE Instrumentation and
Measurement Technology Conference (IMTC 2000), vol.1, May 1-4, 2000, pp.24 - 29
7. C.M. Ho, W.K. Lee, and Y.S. Hung, Signature Representation of Underground Cables and its
Applications to Cable Fault Diagnosis, Proc. of IEE Int. Conf. on Advances in Power System
Control, Operation and Management, Dec, 1993, pp.861-865
8. W K Lee, G S K Fung, H Y Lam, F H Y Chan, M. Lucente, "Exploration on Load Signatures",
International Conference on Electrical Engineering, July 2004, Japan.

A Taxonomy of Load Signatures for Single-Phase Electric Appliances

1.

Introduction:
New and developing technologies provide a better understanding of the usage of

electric power, and promise to strengthen the link between electric utility companies and their
customers. The future may include smart meters that can analyze electrical load signatures to
monitor usage and relay this information to the utility. Knowledge of electric load signatures is
the foundation of practical technologies for load monitoring, load diagnostics, power quality
control and power circuit design. Such knowledge can provide benefits to utilities, customers,
regulators, appliance manufacturers and others. A utility can improve planning and operations
and develop new products and services, such as enhanced building audits, optimized operation
and other energy services. Customers benefit from reduced costs and improvements in power
quality and reliability. Regulators can improve the precision and relevance of policies and rules.
Manufacturers of appliances and equipment can improve quality and compliance, while
anticipating market demands and providing more effective and efficient products.
Previously, different application-specific methods have been used to understand
different aspects of loads: power quality and efficiency, reactive power compensation, system
planning, load shedding, etc. At the core of these approaches is the need to understand the load
characteristic - load signatures - of electrical appliances and equipment. Some understanding of
load signatures has emerged from work in non-intrusive load monitoring (NILM) [1-6], echoresonance load monitoring [7], and other methods. Recent research [8] suggests the creation of a
methodology to measure and represent load characteristics, eventually aimed at the creation and
use of a library of load signature information. Such knowledge is essential to most load
monitoring/identification approaches as well as appliance design and deployment. A typical
taxonomy of load types groups the appliances by the characteristics of their load signatures, rather
than by their use (e.g., HVAC equipment, kitchen appliances, consumer electronics, etc.). These
groupings and subgroupings of load signatures correspond to the underlying power circuitry of
related appliances. Initiated by the CLP Research Institute, this joint research team has carried out
a study on the methodology of load signature characterization based on over 120 office and
domestic single-phase electronic equipment and electric appliances. An approach to load signature
classification is proposed. It is found that load signatures depend primarily on (i) the front-end
power electronic stage and/or (ii) the operational behavior of loads. The main factors that affect
load characterization are discussed and illustrated with examples.
2.

Influence of Power Electronics on Load Signatures

In buildings, power is used by a wide range of electronic equipment and electric

appliances. Increasing use of nonlinear electronic and electric loads has led to various power
quality issues such as harmonics and reactive power. The requirements for meeting various power
quality standards such as EN-61000 have forced many electronic designers to incorporate active
and/or passive means to improve the power quality of the loads. Various types of power factor
correction (PFC) circuits have been developed to improve the input current of electric loads. Such
development has made great impact on load signatures. When a PFC circuit achieves near-unity
power factor, the entire load behaves almost like an equivalent resistance thus masking the
nature of the load;it becomes a challenge to differentiate a pure resistive load (e.g., an electric
heater) from an air-conditioner with a front-end PFC circuit. For electromechanical loads such as
a refrigerator, the input current of a load depends not only on the electric circuit but also
mechanical load characteristics. Therefore, knowledge of the nature of electric loads would help
the power industry to understand how their low-voltage power sources (ac mains) are used in
buildings. The way an electronic load or electric appliance uses the ac power supply is defined
here as load signature. Based on over 120 types of office and domestic electronic/electric loads,
this study indicates that the front-end power electronic circuits and the operational behavior (such
as mechanical characteristics of some electromechanical loads) play critical roles in the way
electronic/electric loads drawn power (particularly current) from the ac power supplies.
3:

Methodology and Practical Evaluation

A.

Main Factors for Load Signature Characterization

In this study. the input current of each load provides load signature information . The

50Hz voltage and current waveforms were captured at a high sampling rate of 50 kHz per channel
which allows us to analyze appliances through raw waveforms in addition to the traditional
electrical measurements. The voltage of the ac mains is not purely sinusoidal; in this study it has a
(maximum) THD of 2.83% and a maximum voltage amplitude deviation of ~8.1V, limiting the
maximum error to 3.68%.
The following factors of the load current are important to characterize each type of
electronic/electric loads commonly used for office and domestic applications: Current is measured
during start-up transient and steady state; this waveform and its envelope are analyzed, as is the
voltage-current trajectory under start-up and steady-state conditions
1.
2.
3.
4.

Start-up current and its envelope

Steady-state current waveform
Transient current
The voltage-current trajectory under start-up and steady-state conditions

B.

Practical Evaluation: (Table 1)

The complexity of the classification cannot be fully explained in a digest. But some

brief examples of load signature classification are given here as examples.

Type 1: Passive and Active resistive loads: These loads behave like resistors. Examples of
passive resistive loads are heating equipment such as hair dryers, kettles and heaters. Examples of
active resistive loads are systems with front-end PFC circuits that operate at near-unity power
factors. In general, the steady-state V-I trajectory (Table 2) of passive resistive load is a straight
line, whilst that of an active resistive load is a zig-zag curve along a straight line. The zig-zag
is due to the switching ripple of the PFC.
Type 2: Rectifier Loads: This group of loads usually draws current pulses in steady state. The
front-end circuits are usually diode-rectifier loaded with an output capacitor or a valley-fill circuit.
Type 3: Motor Loads: This group of loads is commonly used in office/domestic applications.
Examples are electric motor loads such as refrigerators and fans. They are inductive loads and
tend to have a characteristic start-up process (due to the accelerating speed of the motors).
Type 4: General Inductive loads: Unlike Type 3, Type 4 loads do not necessarily have
repeatable start-up behavior. Examples are magnetic ballasts for lighting devices. The igniter of
the ballast consists of a bimetallic switch (S) which operates differently at different times,
depending on the initial voltage and thermal conditions.
C. 2-D Voltage-Current Trajectory (Table 2)
A new approach to classifying load signatures is to use the 2-D V-I trajectory. Table 2 shows the
VI trajectories of a wide range of loads. These 2D trajectories under start-up and steady-state
conditions can be classified in various ways. In the full paper, we will provide the details of the
taxonomy based on eigenfunction decomposition and feature-based analysis.
4.

Conclusion:

Based on the analysis of input current measurements of over 120 office and domestic electronic
equipment and electric appliances, we propose a framework for taxonomy of electric load
signatures. The proposed taxonomy can be used to classify various types of electric loads
commonly used for office and domestic use. It is envisaged that such taxonomy would evolve
continuously with future changes of load characteristics and new types of loads. The framework
proposed here, however, forms the basis for individual load signature identification and taxonomy
study. Active research is being carried out to use signal processing means to differentiate different
types of loads used simultaneously in the same ac mains.

Type of load
1.Passive and
active resistive
loads

Observations
No transient or significant transient:
transient current is nearly the same as
steady state current.

Examples: kettle,
hair dryer

Steady-state current: sinusoidal; little

phase difference between I and V.

2. Rectifier loads

Short transient and a large peak at the

beginning: the magnitude of the first
peak of transient current is at least a
few times (over ten times in some
cases) of steady state peak current, and
the current reaches steady state after a
few cycles (less than 5 cycles in some
cases).

Examples energysaving light bulb,

desktop PC

Waveform examples
Kettle

Desktop PC

Typical equivalent circuit

Pure resistive

Desktop PC

Steady-state current: pulses near

voltage peaks.
3. Motor loads
Examples: fan,
vacuum cleaner,
refrigerator

Slowly decay envelope: transient

current gradually decreases to steady
state because of the increasing back
EMF (e) of motor loads, a function of
motor speed ().

4. General
inductive loads

Non-repeatable: transient current

envelope is different every time.

Examples:
fluorescent lamp
with conventional
ballast

Steady-state current: sinuoidal, out-ofphase with voltage.

Table 1.

Vacuum cleaner

Fluorescent lamp with conventional ballast

Motor with back EMF

Fluorescent lamp with conventional ballast

Example of load types

5

1
5
0
-5
-500

0
17

-0.2
500 -500

-2
500 -500
0.2

0
0
49

-0.2
500 -500

0
0
57

-0.5
500 -500

-0.5
500 -500

65

0
73

-10
500
-500

-5
500 -500

500

-0.5
500 -500

500

0.5

-1
500 -500

500

0.5

0
0
105

0.2

0
106

0
0
113

0.2

-2
500 -500

0
114

500

-5
500 -500

500

-10
500 -500

500

-10
-500

-10
500 -500

-1
500 -500

500

-20
500 -500

0
88

500

0
96

500

0
104

500

0
112

500

0
120

500

500

0.5
0

-2
-500

0
103

-0.5
500 -500
0.2
0

0
111

-0.2
500 -500
0.5
0

0
119

-0.5
500 -500
1

0
0
126

-2
-500

-1
500 -500

0
0

-0.2
500 -500

20

0
0

-0.1
500 -500

0
95

0
0
118

500

-0.5
-500

-0.2
500
-500

0
0
125

0
0

-5
500 -500

-1
500 -500

0.2

0.1

0
0
124

10

0
0

-20
500 -500

0
117

-5
500 -500

0
80

0.2

-0.2
500
-500

0
0
123

10

0
0

-5
-500

0
116

-0.5
500 -500

0
87

0
0
110

500

-0.05
-500

0.2

20

0
0
122

10

0
115

-5
500 -500

0
109

0.5

0
0
121

-5
-500

0
108

0
500

0
-2
500 -500

0
79

0
102

500

0
500

0
-2
500 -500

-20
500 -500

-0.5
-500

-0.2
500 -500

0.5

0
94

0
64

72

0
0
101

500

0.05

20

0
107

-10
500 -500

-0.5
500 -500

500

0.2

-0.1
500
-500

500

0
56

0
0

0
86

500

0.5

71

0
0
93

0
63

0.1

0.5

0
0
100

0
-20
500 -500

-0.5
500 -500

-0.5
500 -500

0
48

0
-0.5
500
-500

0
78

500

0.5

-10
500 -500

0
0
85

10

20

0
-10
500
-500

-0.2
500 -500

-0.5
500 -500

0
55

10

0.5

0
0
92

0
0
99

10

0
-0.5
500 -500

-0.02
-500

0
77

0
40

70

-0.5
500 -500

-1
-500

-0.1
500
-500

-2
-500

0.1

-0.2
500 -500

0.5

0.2

0
0
98

-0.1
500 -500

-10
500 -500

500

0.2

0.5

0
0
84

0
0
91

0.02

0
0
97

-0.1
-500

-10
500 -500

-0.05
-500

69

0.5

0.1

0
0
90

0
76

500

500

0
0
47

0
62

0
32

-5
500 -500

0
0

500

0.05

0
0
83

0.1

0
0
89

-10
-500

0
61

-0.1
-500

0
0
39

0
54

500

-0.5
500 -500

0
-0.1
500
-500

500

0.5

0.1

10

10

0
0
82

0.5

0
75

-10
500 -500

68

-10
500 -500

0
53

0
31

0
46

0
24

-0.2
500 -500

0
-0.5
500
-500

0
0

10

-5
500 -500

-0.5
-500

0
23

0.2

0
38

500

0.1

-1
500 -500

0.5

10

10

0
74

-0.2
500 -500

67

0
0
81

0.5

-2
-500

-10
500 -500

0
60

0
0

10

-0.2
-500

66

10

-0.2
-500

-0.5
500 -500

500

0.2

0
0

10

0
59

-0.5
500 -500

-2
-500

0
16

0
0
30

-10
-500

-0.05
500
-500

0
0
45

0.5

0
0

0.5

-0.5
-500

0
58

0
52

-1
500 -500

0
22

0.5

0.5

0
-0.5
500 -500

0.5

0
51

0.5

0.5

-1
-500

0
50

0.5

500

0
-5
500 -500

-0.5
500 -500

0
0
37

0
0
44

0
-0.5
500 -500

-0.5
500 -500

-1
500 -500

-0.2
-500

500

0.05

-0.5
500 -500

0.5

0
0
43

0.5

0.5

-0.5
-500

-0.5
500 -500

500

0
15

0.5

0
0
29

-10
500 -500

0.2

0
0
36

0.5

0
0
42

-1
500 -500

0
21

8
10

0
14

0
-0.2
500
-500

0
0
35

0.5

0
0
41

-10
-500

-2
500 -500

-1
500 -500

-0.5
-500
0.2

0
0
28

0
0
34

-0.2
500 -500

500

0
0
27

-0.5
-500

-2
500 -500

0
20

-0.1
500 -500

7
10

0
13

0.2

0
0
26

0
0
33

0.5

-0.5
-500

-10
500 -500

0
19

0.2

-0.5
-500

0
18

-2
500 -500
0.1

0
-0.1
500 -500

6
0.5

0
12

0.1

0
-0.5
500 -500

0
0
25

-1
-500

-0.5
500 -500

-5
500 -500

5
2

0
11

0.5

10

-0.5
-500

-1
500 -500

4
5

0
10

10

-1
-500

-1
500 -500
0.5

-10
-500

3
1

0
9

0.5

-0.5
-500

2
1

0.5
0

500

0.5

Table 2. Two-dimensional voltage-current trajectories of 126 electric loads, each with 5 periods
plotted; unit are Amps vs. Volts.