You are on page 1of 6

JOURNAL OF SUSTENABLE ENERGY, VOL. 1, NO.

1, MARCH, 2010

SIGNAL ANALYSIS IN POLLUTED POWER NETWORKS
GHEORGHE Daniel Nicolae*, CHINDRIS Mircea**, VASILIU Răzvan*,
BINDIU Radu*, POP Gabriel Vasile*, GECAN Călin-Octavian*
*
Ph.D. student, Daniel.GHEOHE@eps.utcluj.ro, Mircea.CHINDRIS@eps.utcluj.ro
Razvan.VASILIU@eps.utcluj.ro, **Professor, Technical University of Cluj – Napoca
Technical University of Cluj – Napoca
15 C.Daicoviciu St., 400020, Cluj – Napoca
Tel: +40264 401408

Abstract: In the last decades, commercial electric it . Nonsinusoidal modes cause the presence of different
customers have become increasingly interested in the harmonic components in the current and voltage curves:
relative quality of the power they purchase as their higher harmonics; subharmornics; interharmonics.
businesses rely more heavily on modern high-tech The power must be supplied without interruption, but
processes. The utilization of power electronic control also the voltage and current waveforms must maintain
elements causes unwanted grid perturbations, and nearly sinusoidal shape, constant frequency, and
therefore a reduction of the power supply quality. amplitude at all times to ensure continuous equipment
Spreading of harmonics is one of the main concerns, operation. Low power quality can cause serious problems
caused by switching actions of the power electronic for the affected loads, such as short lifetime,
control elements, and by voltage variations in the malfunctions, instabilities, interruption, and etc. The key
electrical grid. The mentioned grid perturbations lead reason for our increasingly keen interests in power quality
to a functional impairment of the connected power lies in the great economic value directly associated with
electronic as well as electrical devices. In many those disturbances [1].
applications, even small loads reduce power quality Poor power quality is normally characterized by the
and lead to thermal overload of the whole appliance presence of disturbances such as harmonics distortion,
and a reduction of life time of single components, e.g. capacitor switching, high impedance faults, transformer
filter capacitors in converters. Power disturbances inrush currents, lightning pulses, and motor starting
found on typical electric utility distribution systems transients. In order to improve the quality of service,
degrade product quality, increase process downtime, electrical utilities must provide real-time monitoring
and dissatisfy commercial customers. For this reason, systems that are able to identify the signatures of different
monitoring the quality of the supply networks has events and make proper decisions for switching and
become an issue of international interest. maintenance.
This paper describes the main power quality The analysis of power quality in electrical power
issues in electrical networks, aiming to investigate systems includes the study of transient disturbances as
harmonic and interharmonic pollution from a modern frequency variations, sags, swells, flicker or interruptions.
approach. The beginning shows a theoretical According to the periodicity of power disturbances,
description of the distortions occurring in power they can be classified as stationary or non-stationary
networks. The paper continues with a virtual signals. For many stationary signals, Fourier analysis is a
instrument that detects harmonics, interharmonics useful analysis tool because the frequency content is of
and the fundamental frequency of the processed great importance. Practical measurements using Fourier
signal. The virtual instrument is based on signal Transform (FT) assume infinite periodicity of the signal
processing techniques like Fourier and Hartley to be transformed. Furthermore, the time-domain
transforms and a fundamental frequency information in the signal would be spread out on the
approximation algorithm. In the end, the test made whole frequency axis and become unobservable following
with simulated perturbations will be described and the the transformation. Therefore, this method is not suitable
results will be discussed. for analyzing non-stationary signals.
Digital control and protection of power systems
Key words: Fundamental Frequency, Harmonics and require the estimation of supply frequency and its
Interharmonics, Virtual Instrument, Wavelet variation in real-time. Variations in system frequency
Denoising, Zero Crossings from its normal value indicate the occurrence of a
corrective action for its restoration. A large number of
numerical methods is available for frequency estimation
1. INTRODUCTION from the digitized samples of the system voltage[2].
The algorithm proposed in this paper is based on a
The problem of the quality of electrical energy is one Discrete Wavelet Transform filter that attenuates the high
of the most vital problems of today electrical power frequency harmonics which would create false zero
engineering, theory of non-sinusoidal mode being part of crossings near the real one, an adaptive window of search

I.S.S.N. 2067-5538 © 2010 JSE

I. 4. The “Zero Crossings” frequency estimation Fig. and so on. VOL. the remaining coefficients will contain only low frequencies. (3) Original Waveform The mother wavelet chosen is Daubechies 14: 200 Filtered Waveform Magnitude [V] 0  200 0 0. Discrete Wavelet Transform Coefficient 2 + 250V Coefficient 4 +100V Coefficient 5 The following signal contains a fundamental Fig. low frequencies (high 0 scales) correspond to a global information of a signal (that usually spans the entire signal). Magnitude [V] 200 100 2. The transform DWT contains 8 levels of detail.01 0. The last 256 entries represent information at the This signal will be sampled at 512 points and it will be smallest scale. a linear (4) interpolation is performed in order to find an approximate coordinate in time when the signal value is 0. The zero crossing detection algorithm is based on the The transform coefficients are computed : sign difference between two consecutive samples. in terms of frequency. 2067-5538 © 2010 JSE . 1. twice as large. 1. When a group of and samples are found.3  t    17sin(25.S. each representing a frequency band. Daubechies 14 wavelet and scaling function method Where S is the scaling function and W is the wavelet function. high Voltage Waveform 200 scales correspond to a non-detailed global view (of the Magnitude [V] signal). the preceding 128 entries represent a scale stored in vector F. if the low scale coefficients are truncated. Therefore. 2010 and an algorithm that tracks the fundamental frequency 300 and approximates it each period.005 0. Original voltage waveform and filtered voltage waveform B.3) and a high frequency The signal is now decomposed in 10 levels of one (rank 25. whereas high frequencies  200 (low scales) correspond to a detailed information of a hidden pattern in the signal (that usually lasts a relatively 0 0.7).015 0.005 0. MARCH. The first two elements. and low scales correspond to a detailed view. Similarly. are called approximation  f (t)  230sin(t)  23sin3. 0 and 1.02 Time [s] Fig. The parameter scale in the wavelet analysis is similar to the scale used in maps. 4 and 5) component of 50hz and two interharmonics. 1. coefficients. JOURNAL OF SUSTENABLE ENERGY. a low frequency interharmonic (rank 3. The remaining elements are the detail  2 coefficients. Discrete Wavelet Transform coefficients (2.02 short time). NO.S. As in the case of maps.N. FUNDAMENTAL FREQUENCY 0 ESTIMATION ALGORITHM  100 0 200 400 Time [s] A. The signal is reconstructed by taking the Inverse Discrete Wavelet Transform of the remaining coefficients. 2.7  t) (1) coefficients.01 0.015 0. 3. Time [s] The large scale coefficients will contain the Fig. Voltage Waveform fundamental frequency and the low frequency harmonics and the high frequency harmonics will be stored in the The discrete wavelet transform of function f(t): low scale coefficients.

are defined. the axes are translated the quality of electrical energy without analyzing into that location in time so the first non integer cycle is interharmonics distortions seems to be incomplete and truncated. MARCH. induction motors. The following harmonics with high amplitudes. Since the spectrum. voltage (current) harmonic component coefficient as well as permissible values of the power of valve converters. 5.” harmonic frequency is high. discrete Fourier (7) transform (DFT) is still the basic tool. These effects are solved table provides a simple. new algorithms/schemes have been proposed in the literature. the window of search will considerably reduce the number of (5) iterations per period since the samples outside the window are not being read. They can are very close to the “real” crossing points. VOL. Linear interpolation fact that the theory of interharmonics is a relatively new field of knowledge and consequently has not been thoroughly studied as compared with the theory of high (6) harmonics. there will be a negative aspect. Usually these IEC-1000-2-1 [6] defines interharmonic as follows: problems are caused by harmonics and the most difficult “Between the harmonics of the power frequency ones to detect are the ones very close to the fundamental voltage and current. Linear interpolation method : 3. which sometimes totally blurs the true line spectrum . These the first zero crossing when the width and the translation loads include static frequency converters. Furthermore. it will not pass through the Harmonics and interharmonics of a waveform can be low pass wavelet filter. Normalizing of interharmonics is in most cases not carried out. However. 2067-5538 © 2010 JSE . furnaces and all loads not pulsating synchronously with the Multiple zero crossing can occur on a single period fundamental power system frequency. in today conditions of continuous increase of different powerful nonlinear loads analysis of After the first zero crossing.N. In this regard. duration leading to major measuring errors. Fig. However. First zero crossing and the adaptive search Interharmonics can be thought of as the inter- window modulation of the fundamental and harmonic components of the system with any other frequency components and The rectangular search window is dimensioned after can be observed in an increasing number of loads. validity of this approximation is strictly a function of the wave form being analyzed and the signal sequence covered by a window width. Luckily. 2010 Zero crossing condition: expected fundamental frequency value. 1. dividing the number of zero crosses counted and the total Accurate harmonics/interharmonics analysis and duration of the integer cycles. NO. they are produced by low rank steady state over a range of frequencies. INTERHARMONICS Electrical energy quality standards used in different countries identify the norms for higher harmonics levels and distortion coefficient of voltage (current) sinusoidal curve. if the “false” zero crossing which are not an integer of the fundamental. JOURNAL OF SUSTENABLE ENERGY. it is either being developed or exists in way of recommendations.S. This window will be then shifted along the sub-synchronous converter cascades. further frequencies can be observed zero crossing point. 6. the spectral leakage of an analysis of a periodic wave form using DFT may reach an unacceptable level. To tackle the spectral leakage problem. measurement in electrical power systems are of particular importance since a true and exact spectrum of a waveform provides a clear understanding of the causes and effects of waveform distortion. The frequency estimation is determined by insufficient. 1. If the false zero crossings are defined in terms of its spectral components in the quasi- further than the real ones. cycloconverters. This can be explained by the Fig.S. the harmonic appear as discrete frequencies or as a wide-band frequency that produces them is also very high. arc whole signal. DFT is of great interest because it approximates the continuous Fourier transform of the time-domain signal. If the window width of DFT is not properly chosen or if a fixed window width is used uniformly for a spread of frequencies. yet effective mathematical by adjusting a rectangular window of search around the definition: I.

o The same result can be obtained by using the Hartley 30 Transform which.2 28 Rank Fig.fundamental power system frequency  10  40 1 2 5 8 11 14 17 20 23 26 29 The interharmonic analysis is based on the discrete Rank Fourier transform and the signal is windowed at 5Hz or Fig. Magnitude spectrum 0. 1.6 26. Phase spectrum The results in figure 7 and 8 do not contain any T 2 1 2p k   i (11) spectral leakage because the sampled signal has perfectly ak   f (t )  cos(k  1  t )dt   f i  cos( ) T o p i1 p round components. which represent the real and the imaginary part of the transform. like the Hanning bk  T  f (t )  sin( k   1  t )dt   f i  sin( p ) p i 1 (12) window for example.72 with a 5Hz Hanning window I. the 9 6 magnitude and the phase spectrum computed with a 5Hz 3 window are presented below. 2067-5538 © 2010 JSE . spectral leakage appears. (10) Fig.N.5Hz. in its fast version is even faster than 27 24 the Fast Fourier Transform. 7. N 1 F (k )   f (n)W N nk (8) n0 Where W N  exp  j 2 / N  (9) The Fourier coefficients.4 27. lead to the magnitude and phase spectrum. 9.8 25. Interharmonic rank 25.6 26. NO. 260 Harmonic f = nf1 where n is an integer >0 230 200 Magnitude [V] DC f = nf1 for n= 0 170 140 Interharmonic f ≠ nf1 where n is an integer >0 110 80 Subharmonic f > 0 Hz and f < f1 50 20 f1 . MARCH. VOL. This leakage can be reduced by 2 T 1 2p k   i selecting an appropriate window.  k  arctan   E  k  (15) 30 27 Magnitude [V] or 24 k   a  21 2 2  k  arctan   18 Ak  (a k )  (bk )  bk  15 (16) 12 Considering the signal given in figure 1. 0 24 24. The low amplitude interharmonics Ak  (O k ) 2  ( E k ) 2 O  k are no longer visible in the next figure. Interharmonic rank 25.S.4 27.2 28 Rank Hk  HNk Hk  HNk Fig. 2010 Table 1. JOURNAL OF SUSTENABLE ENERGY. If the signal components are real. Magnitude 21 18 N 1 15  x   cos  2   n  k   sin  2   n  k   Hk   n     N     N     12 9 n  0 (13) 6 3 0 The odd and even coefficients are given by : 24 24.72 with a 5Hz Ok Ek 2 2 (14) rectangular window The amplitude and phase spectrum : Using the Hanning window the spectral leakage is significantly reduced.S. 1.8 25. 10. 8.

NO. the measurement error (analytical case). As seen from which is the classic approach and the Hartley Transform. identification module computed the waveform components correctly.5Hz to 58Hz. 9. displaying the magnitude and I.7). 2010 4. one for the fundamental instrument complies with class A equipment where the frequency estimation and the other one for interharmonics maximum allowable deviation for a 10 seconds analysis identification. Within the frequency estimation module. a low frequency interharmonic (rank 3.S.5Hz domain approximated every cycle. 1. Fig. 2067-5538 © 2010 JSE .6 rad. The window slides along the sampled signal the discrete wavelet transform offered important accuracy and the virtual instrument reads only the samples that are improvements and solved most of the false zero crossings inside the window. Fig.3) and a high frequency interharmonic (rank 25. The input waveform and the wavelet filtered waveform are also displayed. 10. necessity of accurate parameters measurement has grown The total numbers of samples was 64000 but the in importance too. the two modules was developed. one for accurate fundamental frequency estimation and the second for interharmonic identification. 1. JOURNAL OF SUSTENABLE ENERGY. The instrument has two modules. VOL. is ±10mHz [5]. The frequency electrical engineering and also two different ways of deviation is calculated as the difference between the interharmonics identification. comparing the analytically input process which can be useful in many other applications in frequency and the estimated frequency. The Fourier algorithm. estimated frequency and the real frequency. the table. duration. The analytical test waveform contains a 50hz fundamental frequency. The new signal processing techniques. Figure 10 presents a set of measurements on the The paper describes briefly the wavelet denoising interval 42. the current or voltage waveform can be either inputted as an array of samples or as an analytical expression.36 rad. The total analysis duration for a 10 The virtual instrument developed performed well in seconds waveform is about 5 seconds so the instrument is high harmonic polluted conditions. Virtual instrument front panel. The instrument returns a figure of Fig. 0. Virtual instrument front panel. instances.N. program performed only 5163 iterations. The window width chosen in the above situation was 5Hz.72mHz. This reduced The algorithm presented and the results obtained are calculation complexity is given by the rectangular search satisfactory. the number The interharmonic identification module performs the of detected false zero crossings and the total analysis Fourier and the Hartley transform. Measurement error on the the frequency approximation in time. According to the 61000-4-30 standard. therefore 128 samples per become a very important issue in power systems so the cycle. the frequency being 42. MARCH.S. a virtual instrument containing 3. the total analysis time (if the waveform is an analytical expression) and the sampling frequency. the total integer cycles time. 11. Frequency estimation CONCLUSIONS The total duration chosen was 10 seconds which corresponds to 500 cycles of 20ms each and the sampling In the last decades power quality estimation has time 6400 samples/second. like window. the maximum error for the analyzed domain is With these techniques. VIRTUAL INSTRUMENT A virtual instrument has been implemented with the help of LabView. The instrument requires the width of the search window. The interharmonic able to perform a real time frequency measurement. each component with a phase shift of 0. Interharmonic identification.7 rad and 1.5Hz – 57.

and Why of Wavelet duration was less than the total waveform duration so the Shrinkage Denoising. 2000. Carmen ş.2: Electromagnetic compatibility (EMC) – Part 4-30: Testing and measurement techniques – [1]. 2010 phase spectrum. 15. vol." Computing In Science And Engineering. Sect. Jan. 2. April 2007 waveform recognition using wavelet-based neural [6]. IEC 61000-4-30 Ed. S. NO. 3. 1: Description of the pp.S. Taswell. How. CEI/IEC 1000-2-1:1990. First Edition. The total analysis [3]. [4]. E. “Power quality disturbance Power quality measurement methods . Golovanov. Dash. 2067-5538 © 2010 JSE . 3. 1. instrument can be used online. Vol. p. no. The frequency estimation module IEEE Transactions on Power Delivery.” Bucharest: Editura Tehnică. May/June 2000. 1. REFERENCES [5]. „Metode moderne de măsurare în electroenergetică. JOURNAL OF SUSTENABLE ENERGY. classifier” IEEE Transactions on Power Delivery. Vol.a. power supply systems. 1990-05 I. VOL. C. Part 2: Environment. J. 222-228. environment – Electromagnetic environment for low- [2].N. K. "The What. 14. P. Santoso. July computed the fundamental frequency of a signal with high 1999 frequency harmonics accurately. “Electromagnetic Compatibility”. MARCH.S. Powers. 12-19. "Frequency Estimation of Distorted Power frequency conducted disturbances and signalling in public System Signals Using Extended Complex Kalaman Filter". No. 2001.