Paper presented at the 8th International Conference on Ubiquitous Computing & Ambient Intelligence UCAmI 2014.
Abstract: This paper proposes an innovative method based on wavelet transform (WT) to decompose the global power consumption in elemental loads corresponding to each appliance. The aim is to identify the main entities that are responsible of total electricity consumption. The research demonstrates that the WT could be used to identify simpler electrical consumption patterns as a part of total consumption curve. Real power measurements has been used in this work. The results obtained have shown the accuracy to decompose consumption curves using WT. This work could be used to develop new energy management services that will improve ambient intelligence.
Using Wavelet Transform to Disaggregate Electrical Power Consumption into the Major End-Uses
1. Francisco J. Ferrández-Pastor
Juan M. García-Chamizo
Mario Nieto-Hidalgo
Vicente Romacho-Agud
Francisco Flórez-Revuelta
Department of Computing Technology
mnieto@dtic.ua.es
3. Main electric panel
I (t) = I1(t)+ I2 (t)+ I3(t)+ I4 (t)+..+ Im (t)
Im(t) = Im1(t)+...+ Imn (t)
I4 (t) = I41(t)+...+ I4n (t)
I3(t) = I31(t)+...+ I3n (t)
I2 (t) = I21(t)+...+ I2n (t)
I1(t) = I11(t)+...+ I1n (t)
Department of Computing Technology
mnieto@dtic.ua.es
4. main electric panel
I (t) = I1(t)+ I2 (t)+ I3(t)+ I4 (t)+..+ Im (t)
current
transformer
CT
data
acquisition
da
wavelet transform
WT
Ida (t) =
I (t)
CT
Im(t) = Im1(t)+...+ Imn (t)
I4 (t) = I41(t)+...+ I4n (t)
I3(t) = I31(t)+...+ I3n (t)
I2 (t) = I21(t)+...+ I2n (t)
I1(t) = I11(t)+...+ I1n (t)
Department of Computing Technology
mnieto@dtic.ua.es
5. current
transformer
data
acquisition
da
CT
wavelet transform
WT
Ida (t) =
I (t)
CT
I (t) = I1(t)+ I2 (t)+ I3(t)+ I4 (t)+..+ Im (t)
Department of Computing Technology
mnieto@dtic.ua.es
6. Supervised phase The events that produce electrical connection and dis-connection
of appliances (lighting, microwave, television, etc.) are
classified as adapted wavelets. When profiling the appliances to build
the knowledge base, we make controlled connections and
disconnections that generate specific signatures for the various power
consumption modes for each appliance or device.
Monitoring phase The aggregate curve of electrical consumption
(captured during a monitoring process) is processed applying a
wavelet transform using adapted wavelet functions Ψi. Actual and
recorded data are used to identify power events (connection/
disconnection of appliances).
Department of Computing Technology
mnieto@dtic.ua.es
8. Wavelet forms acquisition: different adapted wavelets
are obtained
Washing machine
0 20 40 60 80 100 120 140
2
1.5
1
0.5
0
−0.5
−1
−1.5
time (sec)
Refrigerator
0 2 4 6 8 10 12 14 16 18 20
4
3.5
3
2.5
2
1.5
1
0.5
0
−0.5
−1
time (sec)
Washing machine Fridge
Electric hob
0 100 200 300 400 500 600
1.5
1
0.5
0
−0.5
−1
time (sec)
Plasma TV
0 10 20 30 40 50 60 70
1.4
1.2
1
0.8
0.6
0.4
0.2
0
time (sec)
Electric oven Plasma TV
Incandescent light
0 5 10 15 20 25 30
2
1.9
1.8
1.7
1.6
1.5
1.4
1.3
time (sec)
Led TV
0 5 10 15 20 25 30 35 40 45
0.5
0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
time (sec)
Microwave Air conditioned
Department of Computing Technology
mnieto@dtic.ua.es
9. We calculate coefficients of WT with each adapted wavelet
Ψ
a,b
(t)
Washing machine
0 20 40 60 80 100 120 140
400 600 800 1000 1200 1400 1600 1800
18
16
14
12
10
8
6
4
2
Time (sec)
Electrical current (Amper)
Sampling time: 1 second
b2
b3
b1
adapted wavelet location on the signal
Example of a wavelet Ψ
a,b
(t) of fixed dilation at three different locations on the signal. A large
positive value of coefficients is returned in location b2.
2
1.5
1
0.5
0
−0.5
−1
−1.5
time (sec)
Department of Computing Technology
mnieto@dtic.ua.es
10. Analyzed Signal
Percentage of energy for refrigerator adapted wavelet
Time
Scales a
2
4
3
2
0 0.5 1 1.5 2
4
4
x 10
1
0.5
0.45
0.4
0.35
0.3
0.25
0.2
0.15
2.4
2.2
2
1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
Adapted wavelet
Analyzed Signal
Percentage of energy for washing machine adapted wavelet
Time
Scales a
0 0.5 1 1.5 2 2.5
x 10
1
0.1
Adapted wavelet
Example of adapted wavelet transform with pattern
functions Ψi and analysis of energy Eψi (Scalograme)
Department of Computing Technology
mnieto@dtic.ua.es
12. } This method has been tested in a real environment.
◦ Using an energy meter in a house
◦ During 7 days
◦ Sampling time 1Hz
} A set of seven signal forms (f1 to f7) has been taken.
◦ For each form an adapted wavelet (ψ1 to ψ7) is built.
} Wavelet Transform for each adapted wavelet ψ1 to ψ7
is calculated when an event is detected.
◦ A vector of energy coefficients: [wcf1, wcf2, wcf3, wcf4,
wcf5, wcf6, wcf7] is obtained.
◦ The argmax {wcfi } provides the detected form fi.
Department of Computing Technology
mnieto@dtic.ua.es
13. Accuracy of 92.2%
Department of Computing Technology
mnieto@dtic.ua.es
14. } Wavelet transform with adapted signals is a
technique with great potential for power
consumption analysis
} This work shows that data captured by power
meters, in a non-intrusive way, can be treated
with wavelet analysis to identify activities and to
disaggregate the total electricity into the major
end-uses
} This method could be able to recognise
behaviour of people and may be used to develop
new services and in energy management
Department of Computing Technology
mnieto@dtic.ua.es
15. Francisco J. Ferrández-Pastor
Juan M. García-Chamizo
Mario Nieto-Hidalgo
Vicente Romacho-Agud
Francisco Flórez-Revuelta
Department of Computing Technology
mnieto@dtic.ua.es