You are on page 1of 6

2013 IEEE International Conference on Systems, Man, and Cybernetics

Pulse Width Modulation (PWM) Method for Power


Components Estimation – Active and Reactive
Power Measurement
Anastasios G. Bakaoukas∗ , Kuo-Ming Chao† and Weidong Li‡
∗ Faculty of Engineering and Computing, Department of Computing,
Coventry University, Priory Street, Coventry, CV1 5FB, UK
Email: ab3369@coventry.ac.uk – Anastasios.Bakaoukas@gmail.com
† Faculty of Engineering and Computing, Department of Computing,
Coventry University, Priory Street, Coventry, CV1 5FB, UK
Email: k.chao@coventry.ac.uk
‡ Faculty of Engineering and Computing, Department of Computing,
Coventry University, Priory Street, Coventry, CV1 5FB, UK
Email: weidong.li@coventry.ac.uk

Abstract—This paper presents a Pulse Width Modulation component as it plays a crucial role in the stable operation
(PWM) based method for Active Power (AP) and Reactive Power of Power systems [1]. In a number of resent research papers
(RP) measurements in electric sinusoidal and Non–sinusoidal
different methods have been investigated and presented which
power systems. The method takes advantage of the fact that
frequencies present in a power line are of a specific fundamental take advantage of the periodic nature of the signals. In [2]
frequency range (a range centred around the 50Hz or 60Hz) the application of the FFT is presented in Power measure-
and that in case of the presence of harmonics the frequencies ments, however, this solution is heavily frequency depended
of those dominating in the power line spectrum can be specified as the optimal functionality of the FFT is obtained under the
on the basis of the fundamental. In order to estimate the power
condition that the processed input signals are assumed to be
components in a generalised power line, in the first stage the
elimination of any harmonics present and the identification of constant to a fundamental frequency of either 50Hz or 60Hz.
the power line fundamental frequency are cascaded by the In [3,4] Kalman filtering–based methods have been presented
application, in the next step, of a PWM stage at which a Pulse but the statistical properties of the processed signals are a
Wave is generated based on the frequency of the fundamental. In requirement for optimal estimation of unknown signal param-
the final stage of the method the power components are calculated.
eters and these are sometimes difficult to achieve. The filter–
In the major advantages of the proposed method are included:
a) in contrast to a number of existing methods a time delay or based and the wavelet–based methods [5,6] require the phase
shifting of the input signal is not required, b) the method does not shift of the input signal. Other methods [7,8,9,10,11] have
involve a time delay by π/2 of the Current signal with respect to been proposed, which require either one of: a) the Amplitude–
the Voltage signal, c) the fundamental frequency of the power line pulse modulation together with a phase shift operation, b)
is considered as an unknown parameter, simultaneously estimated
an electronic shifter based on stochastic signal processing,
with the signal’s spectrum using the Fast Fourier Technique
(FFT) and so the method is insensitive to fundamental power line c) the incorporation of a back–propagation artificial neural
frequency changes, and d) complex computations that include network, or d) the Reactive Power measurement using a digital
multiplication of the Voltage and Current sample values by Infinite Impulse Response Filter. In general, even from the
trigonometric functions are avoided. early research years in the field, measurement methods of
Power system components were categorised in those using
Index Terms—Active Power, Reactive Power, Power Measure-
ment Method, PWM. Time Domain [12] and those using Frequency Domain [13,14]
techniques. In the Time Domain analysis, system equations as
I. I NTRODUCTION time functions are involved, while in the Frequency Domain
analysis each harmonic component of Voltage and Current
An extensive research on different methods for the Power
waveforms is taken under consideration. The method presented
components estimation has been taking place for a number of
in this paper is based on a general idea first introduced in
years now. Initiated by the increased use of Power electronics,
[15,16], which comes as a direct result of earlier studies [17].
microprocessor–based devices, and extensive Non–linear loads
Although the algorithm as outlined in [15,16] solves many
incorporated into Power systems. These loads contribute exten-
of the complexity and weakness issues dominated previously
sively to the distortion of the Voltage and Current waveforms
demonstrated methods (phase shift operation of the input
due to harmonics.
signal, time delay of the Current signal to the π/2 with
In particular importance stands the need of an accurate
respect to the Voltage signal, calculation complexity due to
method for the measurement of the Reactive Power (RP)

This work was supported by collaborative project “CASSANDRA Project” (ICT Systems for Energy Efficiency), Seventh
Framework Programme, Grant agreement No: 288429.

978-1-4799-0652-9/13 $31.00 © 2013 IEEE 1040


DOI 10.1109/SMC.2013.181
the spectrum estimation methods used etc.) introduces some P (P ower) = V oltage(V ) Current(I) (1)
weaknesses of its own, like: a) the algorithm can only be
applied to Power lines operating at the 50Hz fundamental
As stated by IEEE standard (1459–2000), in the electrical
frequency (50Hz−240V European Power Line), b) a relatively
circuit with the linear load, a sinusoidal Voltage source:
large error window is introduced when frequency variations
from the fundamental after a large block of load or a large √
V (t) = 2 Av sin(ωt) [V oltage] (2)
generator unit is connected to or disconnected from the Power
system and alters its characteristics, and c) for every harmonic
present as frequency component of the Current waveform the produces a sinusoidal Current:
realisation of the method requires a separate Walsh function √
generator. These weaknesses are the direct result of the fact I(t) = 2 Ai sin(ωt − θ) [Current] (3)
that the method uses Walsh Functions (WF) as its effectively
spectrum analysis basis functions, which to operate optimally
require by their definition the fundamental frequency to be Where Av and Ai are the rms values of the Voltage and the
necessarily exactly at 50Hz. While the third–order WF happens Current respectively. Substituting into Equation 1, we obtain:
to have a frequency matching exactly a 50Hz frequency signal (P ower) = V (V
P√  =
oltage) I(Current)
 √
and the seventh–order to have a frequency matching the third 2Av sin(ωt) 2Ai sin(ωt − θ) =
harmonic of this fundamental, there are no corresponding
WFs to match the 60Hz fundamental frequency Power line = Av Ai cos(θ) − Av Ai cos(θ) cos(2ωt)+
characteristics. +Av Ai sin(θ) sin(2ωt) (4)
The aim of this paper is to demonstrate how the weaknesses
outlined in the previous paragraph can be corrected, offering The Active and Reactive Power components are defined as:
that way a generalised method for the accurate calculation Active Power Component (P) = Av Ai cos(θ)
of Active and Reactive Power, while at the same time the Reactive Power Component (Q) = Av Ai sin(θ)
major advantages of having rectangular pulses with a constant
amplitude of (+1) − (−1) at different frequencies as the basis So, the Instantaneous Electric Power is given by:
functions are maintained.
p(t) = P − P cos(2ωt) + Q sin(2ωt) =
The material in this paper is presented in a total of eight
sections. In section two the Active and Reactive Power equa- = P [1 − cos(2ωt)] + Q sin(2ωt) (5)
tion calculations are presented, the results of which are to
Converting equation (5) into a digital form, we obtain:
form the mathematical foundation of the method. The final
representation of the Instantaneous Electric Power equation 4π 4π
p(n) = P [1 − cos(n)] + Q sin( n) (6)
in a matrix form assists significantly in the direction of N N
understanding the logical mathematical steps leading to the The individual Power components, for the 60Hz fundamen-
evaluation of the method. Section three, based on the results tal frequency Power line case, are as illustrated in Figure 1.
obtained in section two, presents the mathematical formulation
of the method. In section four, the derivation steps of the
actual algorithms mathematical equation is presented with the
final rectification of the Reactive Power Waveform allowing
for the averaging of its value lying at its core. Section five
presents the different software evaluation stages of the method,
clearly demonstrating all the constituent parts, all the way
up to the construction of the, according to the fundamental
frequency characteristics, Pulse Wave for the calculation of
the Average Reactive Power component of the Power line. In
section six the simulation parameters used in the evaluation
of the method are presented and explained, while in section
seven the actual simulation results are presentedand. Finally,
section eight includes the conclusions of the paper.
Fig. 1. The Voltage, The Current, The Instantaneous Power, and The Reactive
Power.
II. T HE ACTIVE & R EACTIVE P OWER C ALCULATION
E QUATION The dominant harmonic frequencies produced by most of
the industrial loads are the odd integer multiples of the funda-
The Power equation for a system has been defined and mental frequency (Figure 2), so we expect these frequencies
relates mutually the Active, Reactive and Apparent Power to be dominantly present in the case the Current waveform is
components (P, Q, S) starting from the Power equation: harmonics distorted.

1041
If we now multiply both sides in Equation 11 by sin(3ωt),
we get:
F (t) sin(3ωt) = B1 sin(1ωt) sin(3ωt)+
+B2 sin(2ωt) sin(3ωt) + B3 sin(3ωt) sin(3ωt) + ... (12)

Looking through Equation 12 it is easy to identify that all


but one term has as much area above as below the horizontal
axis on its Cartesian plot representation, that is, possess zero
net area. The exception is the term sin(3ωt) sin(3ωt), which,
being a square, has to have only a positive area (Figure 3).

Fig. 2. Odd Integer Multiples of The Fundamental (60Hz) Frequency.

If we present Equation 5 in an equivalent matrix form, as:


p(t) = P [1 − cos(2ωt)] + Q sin(2ωt) ⇒
 
  P
⇒ p(t) = 1 − cos(2ωt) sin(2ωt) (7)
Q
becomes immediately apparent that if we could achieve for
Equation 7 a form similar to:
 
  P
p(t) = 1 0 (8)
Q

or
Fig. 3. The Area Under The Right–hand Side Terms of Equation 12.
 
  P
p(t) = 0 1 (9) If we now take the area under both sides of Equation 12, all
Q
the terms on the right-hand side except the sin(3ωt)sin(3ωt)
we effectively achieve calculation of the Active and Reactive term will vanish (Equation 13). This fulfils the requirement
Power components respectively, by elimination of the in every set by Equations 8 & 9, for the calculation of Active and
case unwanted component. In order to achieve this we need Reactive Power. The only unknown parameter, in every case,
to take into consideration the fact that P cos(2ωt) and is the appropriate multiplication factor to substitute sin(3ωt)
Q sin(2ωt) are orthogonal to each other. in multiplying both sides of equation 12.

III. P OWER C OMPONENTS M EASUREMENT A PPROACH 1


 nT
[F (t) sin(3ωt)]dt =
If we assume that one cycle of a Current wave down a 1

nT
nT
0
nT 0 [B1 sin(1ωt) sin(3ωt)]dt+
Power line has some arbitrary shape that we can call F (t),  nT
1
which is a function (F ) of time t, and is given in its general + nT [B2 sin(2ωt) sin(3ωt)]dt +
1
 nT
0
nT 0 [B3 sin(3ωt) sin(3ωt)]dt + ... ⇒
form by:
F (t) = A1 cos(1ωt) + A2 cos(2ωt) + 1
 nT
⇒ nT 0
[F (t) sin(3ωt)]dt =
... + B1 sin(1ωt) + B2 sin(2ωt) + ... (10)  nT
1
= [B3 sin(3ωt) sin(3ωt)]dt (13)
nT 0
then, coefficients A1 , A2 , B1 , B2 , etc., represent the portion IV. P ULSE WAVE I NTEGRATION IN BASIC P OWER
by which each sine or cosine wave contributes within function C OMPONENTS C ALCULATION
F (t). For simplification purposes we can rewrite Equation 10 Let us assume that F (t) and sin(3ωt) in Equation 13 are
excluding the A coefficients, without any loss of generality, substituted respectively for Equation 5 and the analytic formula
and obtain: for a Pulse Wave S(t) with half–amplitude A, period T , and
F (t) = B1 sin(1ωt)+ offset to , is defined as:
[2(t−to )T ]
+B2 sin(2ωt) + B3 sin(3ωt)... (11) S(t) = A(−1) =

1042
  
2π(t − to )
= A sgn sin (14)
T

For real t, sgn(t) is given by:


2 H(t) − 1 (15)
where H(t) is the Heaviside Step Function. For an offset to =
0, so we can consider a purely odd function for the Pulse wave,
the formalism of Equation 13 leads to:
 
p(t) Asgn sin 2πt T =
= [P (1 − cos(2ωt)) + Qsin(2ωt)]
 
Asgn sin 2πt T ⇒ Fig. 4. Fully Rectified Reactive Power Waveform (Fundamental at 60Hz).

1
 nT  
⇒ nT 0
p(t) Asgn sin 2πt T dt =  nT  
1
 nT nT 0
p(t) Asgn sin 2πt
T dt =
1
= nT 0
[P (1 − cos(2ωt)) + Qsin(2ωt)]  nT
  1
Asgn sin 2πtT dt ⇒ = |Qsin(2ωt)| dt (22)
nT 0
1
 nT  2πt 
⇒ nT 0 p(t) Asgn sin T dt =
1
 nT   Solving Equation 22 for Q we obtain:
= nT 0
P Asgn sin 2πt T −
 2πt       
−P cos(2ωt) Asgn sin T + π T
2πt
     Q=− p(t) Asgn sin dt (23)
2πt 2T 0 T
+Qsin(2ωt) Asgn sin dt (16)
T A signal with a constant value of 1 over the period T is
all is required to calculate the Average value of the Active
Power component P [17]. So, Equation 16, with the Pulse Wave
Equivalently, the right–hand side of Equation 16 can be possessing a duty cycle of a 100% (The duty cycle is the
decomposed into its three constituent parts as: percent of the period in which the signal is positive), takes the
 nT      form:
1 2πt    tDuty  
P Asgn sin dt (17)  nT 2π t 100% t
nT 0 T 1
p(t) Asgn sin P eriod
dt =
nT 0 T
 nT     
1 2πt 
P cos(2ωt) Asgn sin dt (18) 1 nT
nT 0 T = P dt = P (24)
 nT      nT 0
1 2πt
Qsin(2ωt) Asgn sin dt (19) Corresponding equations, which can be derived following
nT 0 T
the same lines, for the equivalent digital approach are omitted
From Equation 17, the Average Power P has a constant value here since their derivation involves only the trivial process
throughout the overall period of T , so: of converting the analogue frequency components into their
digital equivalent and the substitution of each integral in the
 nT     
1 2πt equations for its digital summation equivalent.
P Asgn sin dt = 0 (20)
nT 0 T
V. P ULSE W IDTH M ODULATION (PWM) A LGORITHM
D EVELOPMENT
Because the Pulse Wave is a purely odd function, when its
frequency matches the 2ω multiple of the fundamental, is The PWM algorithm is centred on a two–fold core con-
orthogonal with the P cos(2ωt) term, so: sisting of a signal generator and a PW Modulator unit. As
the input Current signal enters the system, a filtering process
      takes place, which purpose is the isolation of the fundamental
1 nT
2πt
P cos(2ωt) Asgn sin dt = 0 (21) frequency component from the possible harmonics or other
nT 0 T distortions present.
The product in the integral of Equation 19 results in the full Because we know beforehand that power lines are operated
rectification of the Reactive Power component (Figure 4). As either at a 50Hz or a 60Hz fundamental frequency a low–pass
a result, Equation 19 gives the Average value of the Reactive filter with a Cut–off frequency (−3db) of 80Hz is adequate
Power Q: for the job at hand.

1043
The fact that the filter design corresponds to a low–pass amplitude of the sine wave is set at 1, while the amplitude
filter with a Cut–off frequency at 80Hz actually means that the value of the sawtooth waveform is set at a magnitude of 0.1
system is immune to frequency deviations throughout the first (Figure 6) to insure a straight forward–sharp edge sawtooth
portion of the spectrum. A fundamental frequency of 55Hz envelope. From this relationship between the two signals we
or 65Hz could be (and can) equally well detected. For our take a PWM waveform with a 50% duty cycle (Pulse Wave)
purposes a Yulewalk recursive IIR digital filter using least– possessing a positive half–cycle maximum amplitude of 1 and
squares fit to the specified frequency response is used. The a negative half–cycle minimum value of −1. As in all the
design allows for a stable IIR filter with good performance PWM cases where this trailing edge modulation technique is
and reliability levels. used, the condition applies that when the sawtooth waveform
Figure 5 present the Current spectrum at the presence of the is less than the sine waveform, the PWM signal is in the high
fundamental and the 3rd harmonic against the filter’s frequency state (1), otherwise it is in the low state (0).
response. After the filtering process has been completed, the The Pulse Wave generated is now at a frequency matching
algorithm can detect the fundamental frequency value out exactly the frequency of the Reactive Power component. Mul-
of the filtered Current frequency spectrum and pass on the tiplication between the Reactive Power component waveform
information to the next algorithm stage. and that of the Pulse Wave results in full scale rectification
of the Reactive Power component, as illustrated in Figure 4,
of the entire Instant Power signal. Integration of the rectified
Reactive Power component is giving us essentially the Average
Reactive Power value.
VI. S IMULATION PARAMETERS
In choosing the appropriate parameters to be used for
simulation and algorithm verification purposes, the theoreti-
cal requirements set by the Power line components analysis
equations were followed. Particular attention has been paid on
choosing such simulation parameters that could produce results
easily comparable with results presented in other studies, so a
simple direct comparison could verify or not results obtained
by the algorithmic approach presented in this paper. A first
Fig. 5. The Current Frequency Spectrum (plus 3rd harmonic) Against The
Yulewalk Recursive IIR Digital Filter Frequency Response With a Cut–off requirement for such a kind of a verification procedure was
Frequency at 80Hz. to make sure that the algorithm maintains immunity char-
acteristics to phase shift variations between the Voltage and
the Current input signals. So, a phase value θ parameter was
introduced in the input Current signal equation to account for
the fact, giving us the following Voltage and Current equations
respectively:

v(t) = Vf sin(ωt) (25)

i(t) = If sin(ωt − θ) (26)

with, Vf = 4, If = 2, ω = 2πf , f = 60 the linear frequency in


Hz, ω = 377 the frequency in RadSec , and θ the phase difference
introduced between the Voltage and the Current waveforms.
Fig. 6. The Generated Sine Waveform & The Generated Sawtooth Waveform. During simulations the phase value has been assigned various
values in the closed interval between [0o · · · 90o ]. According
From the last term of Equation 5 we know that the Reactive to the above given parameters the Instant Power Value is given
Power component of the Power line comes always at double by:
the fundamental frequency (sin(2ωt)). The Term Q can be
safely ignored at this stage as its only contribution is at the p(t) = 8 sin(377t) sin(377t − θ) (27)
amplitude of the Power component. Based on this fact and the
fundamental frequency component value, the algorithm’s next VII. S IMULATION R ESULTS
stage generates a sine waveform and a sawtooth waveform The proposed numerical approach for the calculation of the
at double the fundamental frequency value, matching that Power line components and more specifically for the calcula-
way the requirements for the Reactive Power component. The tion of the Active and Reactive Power components is tested,

1044
through computer simulated tests, against a number of phase The considerable simplification of the number of computing
difference values between the input Current and Voltage signals operations (considerably restricted to summations, instead of
to validate the approach. The results as presented in Figure 7 the more computationally demanding multiplications) is one
provide evidence to the validity of the approach, illustrating of the strong holds of the method. The filtering of the input
optimum performance against most of the previously known to the algorithm signal guarantees that is not sensitive to
approaches. its frequency changes, which, during faults or periods of
The results in Figure 7 emphasise the fact that the maxi- abnormal operations and electrical machines on the line with
mum error value given by the algorithm does not exceed the variable speed, tend to be typical. The algorithm has been
0.0018VAR. In the figure the actual and calculated values (Q in tested through dynamic computer simulated tests, which are
VAR) along with the error values (ΔQ) are presented, against currently under extensive expansion to cover different Power
each phase difference value between the current–voltage input line situations and scenarios.
signals. This figure constitutes the lower error value recorded
ACKNOWLEDGMENT
between results of most of the previously known algorithmic
approaches to the problem. The authors would like to thank editors and anonymous
reviewers for their valuable and constructive suggestions on
this paper.
R EFERENCES
[1] Fairney W, “Reactive Power–real or Imaginary?”, Power Engineering
Journal, Volume 8, Issue 2, April 1994, pp. 69 – 75.
[2] Gherasin C, ”DSP Implemantation of Power Measurements According to
The IEEE Trial-use Standard 1459”, IEEE Trans. Instrum. Meas. Vol. 53
(4): 1086 – 1092, 2004.
[3] Girgis A, Qui J, ”Measurement of The Parameters of Slowly Time Varying
High Frequency Transients”, IEEE Trans. Instrum. Meas. Vol. 38 (6): 1057
– 1063, 1989.
[4] Dash P.K, ”An Extended Complex Kalman Filter For Frequency Mea-
surement of Distorted Signals”, IEEE Trans. Instrum. Meas. Vol. 49 (4):
746 – 753, 2000.
[5] Yoon W.K. and Devaney M. J, ”Reactive Power Measurement Using The
Wavelet Transform”, IEEE Trans. Instrum. Meas., Vol. 49, pp. 246–252,
April 2000.
[6] Moulin E, ”Measuring Reactive Power In Energy Meters”, Metering
International, Issue 1, 2002.
[7] Branislav D, Eddy S, and Bosnjakovic P, ”A High Performance Frequency
Insensitive Quadrature Phase Shifter and Its Application In Reactive
Power Measurements”, IEEE Trans. Instrum. Meas., Vol. 49, pp. 161–
165, February 2000.
[8] Branislav D, Eddy S, and Bosnjakovic P, ”A High Performance Frequency
Insensitive Quadrature Phase Shifter and Its Application In Reactive
Power Measurements”, IEEE Trans. Instrum. Meas., Vol. 49, pp. 161–
165, February 2000.
[9] Makram E. B, Haines R. B, and Girgis A. A, ”Effect of Harmonic Distor-
tion In Reactive Power Measurement”, IEEE Trans. Industry Applications,
Vol. 28, Issue 4, pp.782–787, July–August 1992.
[10] Chow T. W. S. and Yam Y. F, ”Measurement And Evaluation of Instan-
taneous Reactive Power Using Neural Networks”, IEEE Transactions on
Fig. 7. Reactive Power (Q) Simulation Results. Power Delivery, Vol. 9, No. 3, pp. 1253–1260, July 1994.
[11] Ozdemir A. and Ferikoglu A, ”Low Cost Mixed–signal Micro–controller
Based Power Measurement Technique”, IEEE Proc. Sci. Meas. Technol.,
Vol. 151, No. 4, pp. 253–258, July 2004.
VIII. C ONCLUSION [12] Page C. H, ”Reactive Power In Non–Sinusiodal Situations”, IEEE Trans.
One of the major problems restricting the optimum opera- Instrum. Meas., Vol. IM–29, pp. 420–423, December 1980.
[13] Sharon D, ”Reactive Power Definitions And Power Factor Improvement
tion of many of the already proposed algorithms and methods In Non-linear Systems”, Proc. IEE, Vol. 120, No. 6, pp. 704–706, June
of calculating Power components is their strong dependency 1973.
on a stable fundamental frequency for the system. Some other [14] Czarnecki L. S, ”Considerations on The Reactive Power In Non–
sinusiodal Situations”, IEEE Trans. Intrum. Meas., Vol. IM–30, No. 3,
methods and algorithms, as outlined in the introduction section, pp. 394–404, September 1985.
can only operate optimally with one specific fundamental [15] Abiyev A, ”A New Signal Processing Based Method For Reactive Power
frequency for the system. Measurements”, Journal of Computers, Vol. 4, No. 7, July 2009.
[16] Abiyev N. A, ”A New Signal Processing Technique For Evaluation
In this paper a PWM based method for Active and Re- of The Reactive Power In The Power Transmission And Distribution
active Power measurement is presented. More particularly, Systems”, Southeastcon Conference IEEE, Print ISBN: 978–1–4244–
the method is based on the application of a combination of 1883–1, 3–6 April 2008.
[17] Brandolini A, Gandelli A, Veroni F, ”Energy Meter Testing Based
filtering and PWM to calculate the Average value of the Power on Walsh Transform Algorithms”, Instrumentation and Measurement
components. The mathematical foundation of the method is Technology Conference, Conference Proceedings, Vol. 3, pp. 1317–1320,
found on the definition of the Instantaneous Electric Power. 10–12 May 1994.

1045

You might also like