1

A Nonlinear Adaptive Filter for Improved

Power System Operation and Protection
R. M. Naidoo, Member, SAIEE, and P. Pillay, Senior Member, IEEE

terms of voltage and frequency deviations.

Abstract-- Fundamental voltage, current and phase
magnitude estimates are required for a wide variety of power III. EXTRACTION OF FUNDAMENTAL FREQUENCY CURRENT
system applications. In this paper, a review of a novel non-linear AND VOLTAGE
adaptive filtering algorithm for power system applications is
presented. The algorithm has the ability to track and estimate the Fundamental voltage, current and phase magnitude
parameters of a sinusoid in real-time. This makes it suitable for a estimates are required for a wide variety of power system
wide range of applications in electrical engineering. It is applications. Examples are active filtering, detection of
compared to other methods of tracking power signals such as the voltage peaks for VAR compensation and the need to extract
Fourier and Wavelet analysis. Suitability of the algorithm for the fundamental frequency component for transient stability
application in power quality, voltage and frequency control and
under frequency relaying has been identified.
assessment. An algorithm capable of extracting and estimating
a specified single sinusoidal component of its input and
Index Terms—nonlinear filter, power quality, dips, harmonics, tracking the variations of amplitude, phase, and frequency of
active filter, frequency control such a sinusoid can be useful for numerous applications. The
potential exists for also tracking multiple frequencies, which
I. INTRODUCTION may exist in cases of instability.

T he ideal power system consists of sinusoidal voltages and

currents at a constant frequency. Power system loads and
system faults affect the shape and frequency of the current
A. Formulation of the Algorithm
A complete mathematical proof is presented in [1]. The
governing set of equations for the algorithm is
and voltage waveforms. The deviation of the waveform from
a pure sinusoid can cause maloperation of sensitive equipment •
and even downtime. The problem has been aggravated with A = µ1e sin φ (1)
the increasing use of power electronic loads. The problem
•
becomes one of identification and classification of power ω = µ 2 eA cos φ (2)
system disturbances. This paper explores the practical
•
application of of the nonlinear adaptive filter proposed by
φ = µ 3 eA cos φ + ω (3)
Ziarani and Konrad [1] for improved power systems operation
and protection. The algorithm may be used for utility power y (t ) = A sin φ (4)
systems as well as finite inertia shipboard power systems,
which may experience significant frequency shifts. e(t ) = u (t ) − y (t ) (5)

II. TRACKING AND PROCESSING OF POWER SIGNALS in which u(t) and y(t) are the input and output signals to the
Current and voltage transducers located at various points in core algorithm, respectively. State variables A, φ and ω
the power system, send proportional signals to measurement directly provide estimates of amplitude, phase, and frequency
and protection equipment. These continuous signals are of u(t). Parameters µ1, µ2, and µ3 are positive numbers that
converted at an appropriate sampling rate into discrete form determine the behavior of the algorithm in terms of
by analogue to digital converters (ADC) of micro controllers convergence speed and accuracy. Specifically, parameter µ1
or digital signal processors (DSP). The data is then processed controls the speed of the transient response of the algorithm
at different levels to derive useful information [2]. with respect to variations in the amplitude of the interfering
signal. Parameters µ2 and µ3 mutually control the speed of the
[3] defines power quality as: “Any power problem manifested transient response of the algorithm with respect to variations
in voltage, current, or frequency deviations that result in in the frequency of the interfering signal [1].
failure or misoperation of customer equipment.” Power
quality phenomena can therefore be categorized mainly in
 2

IV. COMPARISON OF METHODS FOR TRACKING POWER problems in the processing of short duration intervals such as
SIGNALS transients. In [11], the algorithm is used prior to the
application of the wavelet transform to successfully detect
broken rotor bars in induction motors. This approach needs to
A. Voltage Deviation be investigated for improvement in diagnosing and
characterizing power quality transients.
Disturbances
c) Waveform Distortion
a) Voltage Dips and Swells
RMS and peak voltage methods are used to calculate (1) Fourier Transform
voltage dips. Both of these methods use a windowing The Fourier Transform (FT) has several disadvantages
technique to calculate the dip depth and duration. The such as aliasing and spectral leakage [2].
algorithm can be used to track voltage dips in real time
without using a windowing technique. The amplitude A (2) Algorithm
predicts the peak magnitude of the voltage or current The core algorithm is intended to extract one single
waveform. Figure 1 shows the ability of the algorithm to track sinusoidal component from its input signal. A multiplicity of
voltage dips. core units may be employed in parallel, or alternatively
cascaded to decompose a multi-component input signal into
A forty percent dip is simulated after nine cycles. The its constituent sinusoidal components as shown in figure 2
algorithm has the ability to track the dip. It is not known how below.
the algorithm will respond in a noisy environment. Further
research will address methods for minimizing the errors. Input
Core Unit #1 Output #1
Input
Core Unit #1 Output #1

1 error

Predicted Sine
Predicted Amplitude
0.8
Core Unit #2 Output #2 Core Unit #2 Output #2

0.6 error

0.4

0.2

Core Unit #n Output #n Core Unit #n Output #n

0

-0.2 Parallel Cascade

Figure 2: Two ways of decomposing an input signal into a multiplicity of core

-0.4
units
-0.6

-0.8
Frequency Deviation
For power system control and protection, it is important to
-1
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 carry out real time calculations of the supply frequency and
track its variations [4]. Variations in the supply frequency
Figure 1: Tracking a voltage dip in real time from the steady state value can be indicative of unexpected
power system disturbances for which corrective action needs
b) Transients to be taken.
Transients can be categorized as either impulsive or
oscillatory. Oscillatory transients can be classified into three As frequency estimates are normally required in real time,
subclasses coinciding with the common types of switching in processing time is critical and hence computationally
power systems. expensive algorithms and time delays due to pre and post
filtering to remove distortion are undesirable [2]. The zero-
• High-frequency crossing methods and DFT techniques often used assume
• Medium-frequency sinusoidal waveforms of one frequency and are unreliable in
• Low-frequency the presence of distortion.
•
(1) Wavelet A sharp notch filter is capable of separating and extracting
The wavelet transform has been used successfully in a desired sinusoidal component of a given signal if the
diagnosing transient events [9,10]. frequency of the signal remains constant. However, when a
transient occurs on the power system, the frequency may vary
(2) Algorithm with time. Therefore, for changes in the frequency of the
The limited convergence speed of the algorithm poses input signal, errors using a notch filter may be large.
 3

IX. REFERENCES
One of the strongest features demonstrated by the [1] A.K. Ziarani, “Extraction of Nonstationary Sinusoids”, PhD Thesis,
algorithm in the literature, is the estimation of frequency University of Toronto, Canada, 2002
[2] Arrilaga, J., Watson, N.R., Chen, S., Power System Quality Assessment,
[5,6,7,8]. Large variations of the frequency around the center
Wiley, England, 2001
frequency can be measured. Reference [1] compares different [3] McGranaghan, M.F., Dugan, R.C., Beaty, H.W., Santoso, S., “Electric
methods of frequency deviation with the algorithm. The Power Systems Quality”, McGraw-Hill, 2002
algorithm displays a high degree of noise immunity and [4] Kundur, P., Power System Stability and Control, McGraw-Hill, New
York, 1994
directly provides an estimate of the amplitude and phase of [5] A. K. Ziarani and A. Konrad, “A method of elimination of interferences of
the extracted sinusoid. Implementation of the algorithm for quasi-periodic nature,” Proceedings of the 44th IEEE Midwest Symposium
under-frequency relaying and frequency control needs to be on Circuits and Systems, Fairborn, Ohio, 14-17 August 2001, pp. 363-366.
[6] A.K Ziarani, A. Konrad, “A nonlinear adaptive method of elimination of
researched. power line interference in ECG signals,” IEEE Trans. Biomedical
Engineering, Vol. 49, No. 6, June 2002.
V. ADVANTAGES OF THE ALGORITHM [7] A.K. Ziarani, A. Konrad, “A nonlinear method of frequency estimation in
noise,” Proceedings of 2002 IEEE International Midwest Symposium on
• Phase lock loop (PLL) techniques are not required for the Circuits and Systems, Vol. 3, Tulsa, Oklahoma, 4-7 August 2002, pp. 316-
extraction of sinusoidal signals having a time varying 319.
[8] A. K. Ziarani, I. M. Blumenfeld, A. Konrad, “Experimental verification of
phase. PLL’s have a limited frequency lock in range
a novel method of extraction of nonstationary sinusoids,” Proceedings of
within which the variations of the desired signal are 2002 IEEE International Midwest Symposium on Circuits and Systems,
tolerated. Vol. 1, Tulsa, Oklahoma, 4-7 August 2002, pp. 455-458.
[9] P. Pillay, A. Bhattacharjee, “Application of wavelets to model short-term
• It has a simple structure and is easy to implement.
power system disturbances”, IEEE Trans. Power Systems, Vol. 11,
• No windowing of data is required. November 1996.
• Less processing power is required as compared to FFT [10] S. Santoso, P. Hoffman, “Power quality assessment via wavelet transform
analysis”, IEEE Trans. Power Delivery, Vol. 11, No. 2, April 1996.
and Wavelets. [11] P. Pillay, H. Douglas, A.K. Ziarani, A., “Detection of broken rotor bars in
induction motors using wavelet analysis” Electric Machines and Drives
VI. DISADVANTAGES OF THE ALGORITHM Conference, 2003. IEMDC'03. Volume: 2, 1-4 June 2003

• The limited convergence speed of the algorithm poses

X. BIOGRAPHIES
problems in the processing of short duration intervals.
Raj Naidoo has a BSc and Msc in Electrical Engineering from the
• The coefficients of the algorithm have to be optimized for University of Natal and Witwatersrand respectively. He is currently working
a particular application. towards a PhD degree at the University of Cape Town. His interests include
• The errors associated with the algorithm are not known power quality, power system planning and power system stability. He is a
registered professional engineer in South Africa. His working experience
and needs to be investigated and quantified. includes six years at Eskom.

VII. POTENTIAL APPLICATIONS IN POWER SYSTEMS Dr Pragasen Pillay (S'84-M'87-SM'92) received the Bachelor's degree from
the University of Durban-Westville in South Africa in 1981, the Master's degree
• Generator frequency control. from the University of Natal in South Africa in 1983 and the Ph.D from the
• Under-frequency relaying Virginia Polytechnic Institute & State University in 1987, while funded by a
• Dip identification and mitigation. Fulbright Scholarship. From January 1988 to August 1990 he was with the
University of Newcastle upon Tyne in England. From August 1990 to August
• Early detection of sags implies quicker response time for 1995 he was with the University of New Orleans. Currently he is with Clarkson
sag mitigation devices. An example is a UPS. University, Potsdam, NY 13699, where he is a Professor in the Department of
• Harmonics and active filtering. Electrical & Computer Engineering and holds the Jean Newell Distinguished
Professorship in Engineering. Since 1999, he has been an adjunct Professor at
• Flicker quantification.
the University of Cape Town. He is a Senior Member of the IEEE and a member
• Noise elimination for power line communication. of the Power Engineering, Industry Applications, Industrial Electronics and
• Demand and power factor control. Power Electronics Societies. He is a member of the Electric Machines
• Power line communication Committee, Past Chairman of the Industrial Drives Committee within the
Industry Applications Society and Past-Chairman of the Induction Machinery
Sub-Committee in the Power Engineering Society. He is a Fellow of the IEE,
VIII. CONCLUSION England and a Chartered Electrical Engineer. He is also a member of the
The algorithm has been compared to several standard Academy of Science of South Africa. His research and teaching interests are in
modeling, design and control of electric motors and drives for industrial and
methods of identifying power quality phenomena. Potential alternate energy applications and power quality.
applications have been identified for use in power systems.
The ability of the algorithm to track a sag in real time was
demonstrated. This work forms the basis for further research
in the identification of power quality events and the
subsequent improved protection and operation.