FIELD OF INVENTION
This invention relates to a method for detecting and classifying power quality (PQ) anomalies in power systems, like sag, swell, harmonics, flicker and transient. This invention further relates to a method quantifying the above mentioned anomalies. The method of the present invention requires less computation, which makes it feasible for real time implementation
BACKGROUND OF THE INVENTION
Electric power quality has become an important problem associated with power systems and electric machines. The interest in power quality (PQ) has dramatically increased over the past decade due to sensitivity of modern loads to irregularities of the power supply system, increased complexity of power networks and the worldwide move towards the competitive power markets.
Signal Processing is needed to extract specific information from the raw data, which typically in a power system are the voltage and current waveforms. Several reports are available on detection and estimation of the signal. Recently, Short Term Fourier Transform (STFT) and wavelet transforms have been extensively used in signal processing of power quality problems.

The existing methods like STFT are well suited for stationary signals. STFT has fixed frequency resolution. If there is any non-stationary condition in the considered window then STFT will not give satisfactory results. Also the existing wavelet analysis techniques use complicated approximation theory which is quite cumbersome.
Indian Patent No: 219 596 of 15.08.2008 relates to a method and system which are used for the visual scrambling of a video sequence and for the reconstruction of the original content thereof from a digital video stream obtained from encoding based on a wavelet transform.
Indian Patent No: 219 549 of 09.05.2008 describes a method of zero tree encoding of wavelet data.
Indian Patent No: 210 672 of 26,10.2007 unveils a technique for generating a mother wavelet from the speech signal for evaluation of wavelet transforms to capture the time variations in the speech signal.
International Publication No. WO/2007/131176 of 15.11.2007 describes a method and device for detecting cardiac signals in a medical device that includes decomposing sensed cardiac signals using a wavelet function to form a corresponding wavelet transform, generating a first wavelet representation.

International Publication No. WO/2009/028763 of 05.03.2009 describes an apparatus for measuring distortion power quality index and method of operating the apparatus which uses THD.
International Publication No. WO/2010/02957 of 25.03.2010 relates to a power quality improvement device. The power quality improvement device is provided in the form of an autotransformer which comprises an iron core having first, second and third legs, and first, second and third coils which are wound in a zigzag fashion around said first, second and third legs.
There is a need to detect power quality anomalies using a simple approximation theory without using any further processing.
SUMMARY OF THE INVENTION
The main object of the present invention is to provide a simple method for detecting and classifying power quality anomalies like sag, swell, harmonics, flicker and transient which consumes less time with more accuracy as compared with the conventional STFT and wavelet techniques.

Another object of the present invention is to provide a method for quantifying the mentioned power quality anomalies.
These and other objectives of the present invention are achieved by providing an approximation theory based simple method for finding the wavelet coefficients. This facilitates in classifying and quantifying the PQ anomalies directly without using any further processing using fuzzy logic or neural networks which is used in most of the known method. The use of a simple method makes it fast and suitable for real time hard ware implementation.
Thus, the present invention provides a method for detecting and classifying power quality anomalies in power systems, like sag, swell, harmonics, flicker and transient, said method comprising the steps of: analyzing a given power quality monitoring signal with a reference power quality signal; calculating with the help of real time hardware implementation the wavelet coefficient of the signals using periodical approximation; and calculating the index value represented by the ratio of wavelet coefficients of the monitoring signal to that reference signal; thereby detecting, classifying and quantifying power quality problems with the help of the calculated index values.

BRIEF DESCRIPTION OF THE ACCOMPANYING DRAWINGS
The invention can now be explained with the reference to the accompanying drawings where
Figure 1 shows in flowchart form the method of the present invention.
Figure 2 shows localization of swell event.
Figure 3 shows localization of sag event.
Figure 4 shows accuracy level in quantizing sag.
Figure 5 shows index values of anomalies against sample number.
Figure 6 shows computation time against moving windows.
Figure 7 shows accuracy level in quantizing against noise (dB).
DETAILED DESCRIPTION OF THE INVENTION
In the method of the present invention, the given signal is analysed with 'dB4' mother wavelet and wavelet coefficients of that signal are calculated. The ratio of coefficients of monitoring signal to the reference signal is calculated. A pure sinusoidal signal of nominal voltage is taken as reference. The ratio serves as the index value. PQ problems are detected, classified and quantified with the help of the calculated index values. In order to localize the event, the index value of the first level decomposition is used.

In finding the wavelet coefficients for a particular signal of length 2n, an edge effect phenomenon occurs. To overcome this problem, the earlier methods used symmetric extension of input signal, anti symmetric extension of input signal, zero padding extension, etc. In the present invention, periodic extension of input signal is adopted.
The calculation of wavelet coefficients Vi require the transformation matrices Mi, Ni (i = 1, 2, 3,.... N) where the dimensions of Mi and Ni are (2i-1 x 2i). Mi and Ni form the low pass and high pass Daubechies filter coefficients respectively.
For the 'd34' mother wavelet, the low pass and high pass coefficients are given as

The elements of the first row of Mi are same as vector c. Whenever the coefficients spill over the number of columns of Mi the remaining elements are wrapped around and the terms in the same column are added. If the length column is longer than the columns of c, the elements in the remaining columns are zero. For example,


To extract the features of a given signal, we need to decompose the wavelet levels after finding the coefficients. As generally dyadic filter bank is used to analyze a signal, at every level the number of coefficients gets reduced by increasing power of two. Hence approximation methods are to be adopted to decompose the wavelet levels. The existing approximations used are approximation by zeros and splines. The method of the present

invention uses linear approximation as it is simple and requires less computation time which makes it suitable for real time implementation. A smooth periodic signal is assumed in decomposition levels and the curve between two definite points is calculated using linear matrix. For example, consider a signal with sampling rate of 1.6 kHz. It has 32 samples/cycle for a 50Hz signal. To localize a event we need to decompose the index value after finding the wavelet coefficients. Then the decomposition matrix is:

A signal with 2n samples can be decomposed into Yi' decomposition levels. In the above decomposition matrix, the last row indicates first decomposition level and the first row indicated 5th decomposition level.
To detect, classify and quantify the power quality problems, a simple method which makes use of above mentioned linear approximation theory for finding wavelet coefficients is used. The method of the present invention does not make use of any heuristic mathematical tools. The flowchart of the method of the present invention is shown in Figure 1, which is self explanatory.

With the wavelet technique of the present invention using 'dB4' mother wavelet, signals with 'swell' and 'sag' are analysed and thereafter, first level decomposition is used to localize the event. This is shown in Fig.2 and Fig.3.
The accuracy level in quantizing 'sag' of 0.9090 p.u by the method of the present invention for different noise levels and different moving windows is calculated and plotted as illustrated in Fig.4.
The index value for different PQ anomalies like sag, swell, harmonics, transient, flicker, spike are calculated and the results are plotted in Fig.5.
The time taken by the method of the present invention to detect PQ anomalies (as an example 'swell') for different moving windows was calculated. This is illustrated in Fig.6. It is found that one-eighth cycle moving window is effective.
The accuracy level in quantizing 'sag' of with different per unit values by the method of the present invention for different noise levels was calculated and plotted as shown in Fig.7. The accuracy level decreases with the increase in both noise level and depth of sag.
The technique of the present invention can also be effectively used in a product for monitoring the rotating diode bridge status of brushless excitation system. In rotating diode

bridge rectifier of generator system, the failure of one of the diodes in the bridge can be detected by a method which uses RMS value and average value. It can be done in mapping the diode bridge current in frequency domain. Moreover, from the generator potential transformer (PT) and current transformer (CT), the signal can be analyzed using wavelet transform as mentioned in the method of the present invention the quality of the generator and the quality of the power that the utility is feeding to the grid can be determined with the help of this signal analysis.
CONCLUSION OF INVENTION
The method of the present invention is able to detect, classify the PQ problems. It also yields quantification of sag and swell with 100% accuracy for pure signals and with 97.7% accuracy for signal with noise up to 20 decibels. It is found that one-eighth cycle moving window is effective. The method does not make use of any sophisticated mathematical tools. It reduces the memory requirement and time computation. Moreover, the method works well even for low sampling rate of 1.6 kHz. Hence it can be concluded that the proposed method is a simple effective tool which can be used in distribution systems with hardware compatibility.

REFERENCES
[1] A text book on "Mixed-signal and DSP design techniques", Walt Kester, Analog Devices, 2003
edition, Newnes publishers, an imprint of Elsevier Science. [2] Ding Ning, Cai Wei Suo Juan Wang Jianwei and Xu Yonghai," Voltage sag disturbance detection
based on RMS voltage method", 978-1-42442644-2487-0, IEEE Transactions, 2009. [3] Nermeen Talaat, W.R.Ibrnahim and George L. Kusic "New Technique for categorization of Power
Quality Disturbances", Power Quality and Supply reliability IEEE conference, pp. 11-16, 2008. [4] Peibing Lu, Shiping Su, Guiying Liu and Haizhou Rong," A New Power Quality device based on
Embedded Technique",DRPT 2008, 6-9 April 2008. [5] Jidong Wang and Chengshan Wang, "Detection of power quality disturbance based on binary
wavelet transforms.". IEEE region 10 conference, Oct30-Nov2, JENCON 2007. [6] A.K. Chandel, G. Guleria and R. Chandel," Classification of power quality problems using wavelet
based artificial Neural Networks", Transmission and Distribution Conference and Exposition, 12
May 2008. [7] Wei Chen, Xiaohong hao and Jie Lin, "Identification of Voltage sag in distribution system using
wavelet transform and SVM", IEEE international conference on control and automation,
Guangzhou, 05 November 2007. [8] Mohammad E Salem, Azah Mohamed, Salina Abd. Samad and Ramizi Mohamed, "Developemnet
of DSP based power quality monitoring instrument for real time detection of power disturbances",
International conference on Power Electronics and Drives Systems, Kuala, pp. 304-307, 18 April
2006. [9] A text book on "Discrete Wavelet Transformations", Patrick J. Van Fleet, A John Wiley & sons,
Inc. Publication, 2008 edition.

WE CLAIM
1. A method for detecting and classifying power quality anomalies in power systems, like
sag, swell, harmonics, flicker and transient, said method comprising the steps of:
- analyzing a given power quality monitoring signal with a reference power quality signal;
- calculating with the help of real time hardware implementation the wavelet coefficient of the signals using periodical approximation; and
- calculating the index value represented by the ratio of wavelet coefficients of the monitoring signal to that reference signal;
thereby detecting, classifying and quantifying power quality problems with the help of the calculated index values.
2. The method as claimed in claim 1, wherein said method yields quantification of sag and swell with 100% accuracy for pure signals and with 97.7 % accuracy for signals with noise upto 20 dB.
3. The method as claimed in claim 1, wherein memory requirements and time for computation are reduced and said method is independent of any need for sophisticated mathematical tool.

4. The method as claimed in claim 1, wherein the approximation step used for calculating the wavelet coefficient is linear approximation.
5. The method as claimed in claim 1, wherein the event is localized with the help of the index value of first level decomposition.
6. The method as claimed in claim 1, wherein said method is capable of working even for a low sampling rate of 1.6 k Hz.
7. A method for detecting and classifying power quality anomalies in power systems, like sag, swell, harmonics, flicker and transient, substantially as herein described and illustrated in the figures of the accompanying drawings.

A method for detecting and classifying power quality anomalies in power systems, like
sag, swell, harmonics, flicker and transient, said method comprising the steps of:
- analyzing a given power quality monitoring signal with a reference power quality signal;
- calculating with the help of real time hardware implementation the wavelet coefficient of the signals using periodical approximation; and
- calculating the index value represented by the ratio of wavelet coefficients of the monitoring signal to that reference signal;
thereby detecting, classifying and quantifying power quality problems with the help of the calculated index values.