17th Conference of Electric Power Supply Industry, MACAU SAR China, 27-31 October 2008

A New Technique for Power Quality Based Condition Monitoring

Mohamed Fuad Faisal Engineering Department Distribution Division TNB, Malaysia mfuadfa@tnb.com.my Azah Mohamed Department of Electrical, Electronic and System Engineering, Universiti Kebangsaan Malaysia azah@eng.ukm.my

Abstract Most power quality disturbances can come from within the facility itself, such as large loads turning on simultaneously, improper wiring and grounding practices, the start-up of large motors, and electronic equipment that can be both a source and victim of power quality phenomena or from externally generated for example lightning strokes on the power lines. These disturbances can interrupt production lines, cause damage to products and equipment, result in lost orders or transactions, corrupt data communication and storage, and cause an overall decrease in productivity in todays global economy. At present, there are no techniques that can effectively correlate the occurrences of these disturbances to the failure of the sensitive equipments. Most of the time, the causes of the equipment failures were termed as nuisance trippings. Unlike a comprehensive electrical system survey, a power quality based condition monitoring focuses on a small set of parameters that can indicate the existence of power quality disturbances and predict possible critical load failures. The term power quality condition monitoring is related to the detection and classification of power quality disturbances. By monitoring the power quality at critical loads, the exact condition of the power supply can be monitored. The condition of the power at specific dates can be used for to predict possible downtime of sensitive machinery. It is important to note that voltage fluctuation, harmonic distortion, and unbalance are good indicators to indicate the existence of these power quality disturbances. These data can also indicate the condition of the load and power system, and can be recorded quickly with little incremental labor using a power quality recorder. Current measurements can also identify changes in the way the load is drawing. All of these measurements can be taken without halting operations and generate numbers that can easily be entered into maintenance software and plotted over time. Identifying the content of digital data forms can be done most reliably by hand, but the large volume of data that is now available makes autonomous techniques necessary. Very few techniques have been applied to the problem of extracting and storing information from the data recorded. In this paper, a new technique for detection and classification of power quality disturbance data for monitoring the condition of the power supply is presented. Keywords: Power quality disturbances, signal processing, S-transform, artificial intelligence, SVM

17th Conference of Electric Power Supply Industry, MACAU SAR China, 27-31 October 2008

1.0 Introduction Each of the power supply problems experienced by the customers has a different cause. Some power supply problems are a result of the shared infrastructure. For example, a fault on the network may cause a power quality event called voltage sag that can affect many customers and the higher the level of the fault, the greater the number affected, or a problem on one customers site may cause a transient that affects all other customers on the same subsystem. Other problems, such as harmonics, arise within the customers own installation and may or may not propagate onto the network and so affect other customers. To understand the condition of the power supply and solve the malevolent power quality problems, the power supply must be monitored using power quality recorders. All the disturbances in the power supply must be recorded and measured accurately according to the definition of each disturbance category and mitigation actions need to be taken to minimize the causes of the power quality disturbances. On line power quality monitoring can provide the fast and important information required to understand power quality problems. In many ways the needs of these facilities are similar to aircraft flight recorders. By providing event tracking and data recording, a re-creation of an actual event can be accomplished. From this data, a determination can be made as to the real cause of an event. As a management tool this provides information for preventing or precluding future "crashes". For power quality, the same principal of recording and tracking of power quality is similar to that of the aircraft flight recorder. The waveform signatures recorded by the power quality recorders can provide excellent information for evaluating the condition of the power and in the analysis of possible causes of the equipment maloperation and to trigger the needs to conduct conditioned based maintenance (CBM) activities. 2.0 Importance monitoring of power quality

Monitoring of the power quality parameters are the pre-conditions for developing counter measures against power quality related problems. Once the disturbances are detected and classified, the sources of these disturbances need to be determined for initiating countermeasures against power quality deterioration. Unlike a comprehensive electrical system survey, predictive maintenance power quality focuses on a small set of measurements that can predict power distribution or critical load failures. By checking the power quality at critical loads, the effect of the electrical system up to the load can be seen. Overall power quality monitoring can address these issues in a number of ways: Evaluation of incoming electric supply and distribution throughout the facility to determine if power quality disturbances or variations are impacting, or have the potential to impact, facility operations and/or manufacturing processes. Identification of power quality trends to provide a baseline for establishing predictive maintenance activities and avoiding interruptions of critical business activities.

17th Conference of Electric Power Supply Industry, MACAU SAR China, 27-31 October 2008

Optimization of power mitigation equipment using a reliability or condition-based monitoring approach. Power parameters can be correlated with process performance and output to locate production defects caused by poor power quality.

3.0 New technique for power quality detection and classification In this paper, a new technique for power quality condition monitoring program involving post signal processing and artificial intelligence is presented. The objective of the proposed technique is to evaluate the condition of the supply systems and identify power quality trends to provide a baseline for establishing predictive maintenance activities in order to avoid interruptions of critical business activities. The new technique can also prompt the users on the needs for the investigations and corrective measures to prevent possible plant failure. The block diagram for the proposed technique is shown in Figure 1.
Measurement Data from PQ recorders Stage 1 Stage 2

1159 [1] b. All types of power quality disturbances will be processed using a signal processing technique. Features for all types of power quality disturbances or unknown events will be extracted. c. The next step is to classify the types of power quality disturbances and unknown using an artificial intelligence technique. The explanation of the technique is presented below. 4.0 Signal processing technique (S-Transform) for detection of events The post processing of the signals will be performed using S-Transform. The S-transform is a time-frequency representation known for its local spectral phase properties [2]. A key feature of the S-transform is that it uniquely combines a frequency dependent resolution of the time-frequency space and absolutely referenced local phase information. This allows one to define the meaning of phase in a lltocal spectrum setting, and results in many desirable characteristics. The S-transform can be considered as an advancement of another signal processing technique called the Continuous Wavelet Transform (CWT), which produces a time-frequency representation of a time series signal [3]. It can uniquely combine a frequency dependent resolution that simultaneously localizes the real and imaginary spectra The basis function for the S-transform is the Gaussian modulation cosinusoids. The cosinuosoids frequencies are used for the interpretation and exploiting the resulting time frequency spectrum. In the case of non-stationary waveforms with noisy data, the S-transform provides patterns that closely resemble a

Detection of voltage disturbances Signal Processing Techniques

Classification of voltage disturbances Artificial Intelligence Techniques

Reports on types of power quality disturbances

Figure 1: Block diagram for power quality based condition monitoring The process for the detection & classification of voltage is explained as follows: a. The power quality recorders will be programmed to record all types of power quality disturbances as defined in IEEE

17th Conference of Electric Power Supply Industry, MACAU SAR China, 27-31 October 2008

disturbance type. The S-transform for a function, h(t) is given by,

S ( , f ) = h(t ) g ( t , f )e 2ft (1)

in which g(,f) is the Gaussian modulation function, given by,

g ( , f ) =

f
2

e (t

/ 2 2 )

(2)

In this paper, the S-transform will be used in the detection of the power quality disturbances. Features to characterize the respective power quality disturbances are extracted using the S-transform. The list of the features selected are shown in Table 1. Table 1 Extracted features using S-transform

nonlinear classification, function estimation and density estimation which has also led to many other recent developments in kernel based methods in general [4]. Originally, it has been introduced within the context of statistical learning theory and structural risk minimization. A special property of SVMs is that they simultaneously minimize the empirical classification error and maximize the geometric margin; hence they are also known as maximum margin classifiers. Viewing the input data as two sets of vectors in an n-dimensional space, an SVM will construct a separating hyperplane in that space, one which maximizes the "margin" between the two data sets. The overall process flowchart for the classification of power quality disturbances using SVM is shown in Figure 2.
List of Power Quality Disturbances

Features Description F1 Minimum value of the magnitudetime spectrum from S-matrix F2 Maximum value of the magnitudetime spectrum from S-matrix F3 Mean value of the magnitude-time spectrum from S-matrix F4 Standard deviation of the magnitude-time spectrum from S-matrix The features will then be used in the classification of the power quality disturbances using artificial technique called SVM.

Classification process

Voltage Sags

Non voltage Sags

3-phase voltage sags

Harmonics

Transients

Unknown

2-phase voltage sags

Implusive

1-phase voltage sags

Oscillatory

Notches

Figure 2: Process flow for the classification process The first step is to differentiate between a voltage sag and non voltage sag. The next step for voltage sag classification is to classify for three phase sags, two phase sags

5.0 Classification technique using Support Vector Machine (SVM)

Support Vector Machines is a powerful methodology for solving problems in

17th Conference of Electric Power Supply Industry, MACAU SAR China, 27-31 October 2008

and single phase sags. For non voltage sags, the voltage events will be classified for harmonics, transients and unknowns. Examples of the classification accuracy for the SVM which was conducted based on 100 sets of power quality data monitored at one of the utilitys substations are shown in Table 1. Table 1: Classification accuracy of SVM PQ events Number of events based on visual inspection 55 12 6 9 1 14 2 % accuracy of data classification 100% 100% 100% 100% 100% 100% 100%

overall condition of the power supply network need to be evaluated. A power quality recorder was install in the plant for 3 month. The data recorded were then processed using S-Transform. From the results, 2 voltage sags (Figure 3), 1 transient (Figure 4) and 12 unknown events (Examples are shown in Figure 5 & 6) were recorded. The causes of the voltage sags were due to lightning strokes on the utility power lines. The cause of the transient events (Figure 4) was found out after the implementation of a thermal scan at the main switch board conductors due to loose connection one of the blue phase conductors. The results in Figures 5 and 6, showed frequent occurrences of incipient faults at the red phase of the factory supply schemes. Overall 12 events were recorded at the red phase.

Voltage sag Voltage swell Oscillatory transient Impulsive transient Notches Harmonics Interruption

Once all the power quality disturbances have been classified, the next step is to identify their causes using thermal scanning and ultrasound techniques. The source of the problem will then be rectified. Figure 3: A voltage sag event

6.0 A case study for power quality condition based monitoring

To illustrate the use of this new technique, a sample case study is presented. This case study was conducted at one industrial plant in Malaysia. Previously, the plant complained of frequent occurrences of nuisance trippings and damages to the production equipment for the last 1 year and put the blame to the power utility. To first understand the problem, the

Figure 4: Transient events for all three phases

17th Conference of Electric Power Supply Industry, MACAU SAR China, 27-31 October 2008

the power supply.

Figure 5: Incipient faults No.1 at red phase Figure 7: Signs of electrical tracking

Figure 6: Incipient faults No.2 at red phase An ultrasound scan was proposed to be conducted at the factory electrical installations. From the scan, it was found out that one of the cable terminations (red phase) had signs of electrical tracking (Figure 7). Rectification was done by replacing the cable termination. The power quality recorder was left for another 1 month and the results which were downloaded later showed no sign of the incipient faults (Figure 8). The signal processing technique had accurately monitored the condition of the power supply and detected the existence of incipient faults. Figure 8: No signs of any power quality disturbances after repair work had been done Monitoring of the power quality parameters are the pre-conditions for developing counter measures against power quality related problems. Once the disturbances are detected and classified, the sources of these disturbances need to be determined for initiating countermeasures against power quality deterioration. The current method using rms is not able to decipher the information in the signals. Using advanced techniques however, can monitor and identify the power quality trends to provide a baseline for establishing predictive maintenance activities and avoiding interruptions of critical business activities.

7.0 Conclusion
The waveform signatures recorded by the power quality recorders can provide excellent information for monitoring the condition of

17th Conference of Electric Power Supply Industry, MACAU SAR China, 27-31 October 2008

References
[1] IEEE Recommended Practice for Monitoring Electric Power Quality, IEEE 1159:1992. The Institute of Electrical and Electronics Engineers, Inc. 345 East 47th Street, New York, NY 10017-2394, USA [2] Zhao, F., Yang, R. 2007. Power Quality Disturbance Recognition using S-Transform, IEEE Transactions on Power Delivery, 22, No.2:944. [3] Chilukuri, M. V., Dash, P. K., Basu, K. P. 2004. Time Frequency Based Pattern Recognition Technique For Detection And Classification of Power Quality Disturbances. Proceedings of the TENCON 2004, IEEE Region 10 Conference:260-263 [4] Christopher J.Burges, A Tutorial on Support Vector Machines for Pattern, Data Mining and Knowledge Discovery, 2, 121167 (1998). Kluwer Academic Publishers, Boston. Manufactured in The Netherlands

Azah Mohamed received her B.Sc from University of London in 1978 and M.Sc and Ph.D from Universiti Malaya in 1988 and 1995, respectively. She is a professor at the Department of Electrical, Electronic and Systems Engineering, Universiti Kebangsaan Malaysia. Her main research interests are in power system security, power quality and artificial intelligence. She is a senior member of IEEE.

Biographies

Mohamed Fuad Faisal is a Power Quality engineer in the Engineering Department in Distribution Division of TNB. He received his BSc from CWRU in Cleveland Ohio and MSc from UiTM Shah Alam Selangor. He is currently pursuing his Phd in UKM, Bangi, Selangor.