The 4 th International Power Engineering and Optimization Conf. (PEOCO2010), Shah Alam, Selangor, MALAYSIA: 23-24 June 2010 Investigation on the Effectiveness of Classifying the Voltage Sag Using Support Vector Machine Hanim Ismail, Noraliza Hamzah, Zuhaina Zakaria & Shahrani Shahbudin Faculty of Electrical Engineering, Universiti Teknologi MARA, 40450, Shah Alam, Selangor, Malaysia. noralizah@salam.uitm.edu.my Abstract - Voltage sags are currently one of the vital issues in power quality today. Voltage sag is a short duration reductions in RMS voltage caused by fault, induction motor starting and transformer energizing. Aim of this paper is to classify the caused of sag either by fault or induction motor starting using SVM. In this paper, the voltage sag was analyzed using PSCAD model. Then, a method to identify voltage sag using mother wavelet Daubechies 4 and support vector machines are used. The waves were discomposed into 10 levels using wavelet transform, afterwards, the selected energy features that were extracted from different levels, were employed as the inputs of the Support Vector Machines to classify the voltage sag. Keywords- Voltage Sag, PSCAD, Wavelet Transform, Support Vector Machines I. INTRODUCTION Power quality disturbances such as voltage sag, transient, flickers, swell and notches may lead to the malfunctions of any sensitive electric facility such as computer-based processes or automatic systems. Voltage sags have become an important power quality issues in the whole world over the past several years. It is defined as the temporary reduction of RMS voltage between 0.1 and 0.9 p.u and with duration from half cycle and 1 second [1, 2]. In despite of the very short duration of voltage sag, it can cause serious problems to a wide range of consumers. Especially industrial customers suffer from regular production stoppages due to voltage sags. Voltage sag can bring tremendous economic losing. In manufactories, it may reduce productivity, malfunctions and induce the quality problems. In order to enhance the diagnosis of power quality in power system, it is desirable to develop method for identifying the causes of power quality disturbance which include voltage sag. Reference [3] proposed a SVM classification for voltage disturbance. Support Vector Machine have been used for classification of the power quality disturbance such as swell, flicker, harmonics and voltage sag [4] in which the inputs are wavelet energy for nine period window and its beginning time. Another method which used the SVM and wavelet decomposition to classify the PQ disturbance is reference [5] in which a multi class classification has been used to classify the disturbance. Based on the encouraging results using SVM [3-6], this paper proposes the application of SVM as a technique to classify the cause of voltage sag as either it is cause by faults in power distribution system or by starting of induction motor. . II. SIMULATION OF VOLTAGE SAG A. Voltage Sag caused by Fault Faults are the main cause of severe voltage sag today. During short circuit faults, the fault current flows through the fault impedance. As soon as fault clearing device interrupts the flow of current, the voltage returns to normal. Even though the voltage returns to normal, many sensitive types of equipment will experience a production outage. Fig 1 below shows the rms voltage waveforms during ground fault. As shown above, voltage sa As shown above, voltage sag occurs at 0.3s and ends at 0.4s after the fault clearing devices operates. As we can see, there is a sharp drop in voltage (correspond to the fault) and followed by a period with a constant voltage and a sharp recovery (correspond to the fault-clearing device). B. Voltage Sag caused by Induction Motor Starting During starting of induction motor, the motor draws approximately 5 times their full-load current, and with a very low power factor. This current causes shallow voltage sags. The magnitude of the voltage sags depends on the characteristics of the induction motor. Motor starting leads to a sudden simultaneous voltage drop in the three phases followed by a slow recovery. Since the motor is a three phase symmetric load, each phase will draw the same inrush Figure 1. RMS voltage waveforms due to fault 978-1-4244-7128-7/10/$26.00 ©2010 IEEE 402