INTEGRATED INSTANCE-BASED AND KERNEL METHODS FOR POWER QUALITY KNOWLEDGE MODELING Mennan Güder, Özgül Salor and Işık Çadırcı The Scientific and Technological Research Council of Turkey, 06531, Tubitak-uzay Metu Ankara, Turkey Keywords: Machine Learning, Knowledge Discovery, Power Quality Mining, Feature Construction, Feature Extraction. Abstract: In this paper, an integrated knowledge discovery strategy for high dimensional spatial power quality event data is proposed. Real time, distributed measuring of the electricity transmission system parameters provides huge number of time series power quality events. The proposed method aims to construct characteristic event distribution and interaction models for individual power quality sensors and the whole electricity transmission system by considering feasibility, time and accuracy concerns. In order to construct the knowledge and prediction model for the power quality domain; feature construction, feature selection, event clustering, and multi-class support vector machine supervised learning algorithms are employed. 1 INTRODUCTION In order to improve the Power Quality (PQ) in energy generation, transmission and distribution systems, real-time and long-period data have to be investigated and an exhaustive model of the electricity system characteristics has to be constructed. PQ events may cause shut down of processes run by electronics devices. Therefore it is important to detect, classify and model PQ events occurrings on a specific site to take countermeasures against the potential PQ problems. Data mining methodologies on the PQ event data may be used to identify the correlations between the events, sites and transformer substations. The cause and location of any event may also be identified with the use of collected data. The resulting knowledge may be used to avoid the problem in the future. Widespread and long term PQ monitoring and analysis are required to collect such data and contruct the corresponding modelling. To handle the huge amount of PQ data, a considerable amount of effort has been spent previously. Automatic clustering is applied on the harmonics data collected from three year simultaneous measurements of eight sites in a transformer substation (Asheibi, Stirling & Soetanto 2006). SNOB (2010) and AutoClass (2010) data mining tools are used to cluster the collected data, where SNOB implements unsupervised learning using minimum message length principle and AutoClass implements Bayesian classification. In the research (Dash, Chun & Chilukuri 2003), examination is carried on voltage raw data collected for one year. First, data processing by using a phase corrected wavelet transform is applied to extract relevant features. Then the features and if-then-ruled fuzzy neural classifier are used to classify the short duration transient PQ disturbance patterns. Fuzzy multi-layered perception is used to determine the class membership values of the input patterns. The trained fuzzy neural network is also used for rule generation. Another research (Asheibi, Stirling & Robinson 2006), uses ACPro clustering software in order to build predicting models for load forecasting and to discover the relationships between the input and output variables. The other research is based on signal processing techniques. In Gerek, Ece and Barkana’s (2006) research, covariance behavior of several features derived from the event data is used for PQ event detection and classification. Classification of PQ events such as harmonics, sags, and capacitor switching is achieved using time- frequency analysis of the voltage and current waveforms in Wang, Rowe and Mamishev’s research (2003). Neural networks have been used by Uyar, Yildirim and Gencoglu (2008) for PQ disturbance classification, while fuzzy-expert systems are used by Liao and Lee (2003) for the same purpose. Wavelets are used for PQ event classification in the Hu, Zhu and Ren’s research and Wang, Rowe and Mamishev’s research. In these 352 Güder M., Salor Ö. and Çadırcı I.. INTEGRATED INSTANCE-BASED AND KERNEL METHODS FOR POWER QUALITY KNOWLEDGE MODELING. DOI: 10.5220/0003117703520357 In Proceedings of the International Conference on Knowledge Discovery and Information Retrieval (KDIR-2010), pages 352-357 ISBN: 978-989-8425-28-7 Copyright c  2010 SCITEPRESS (Science and Technology Publications, Lda.)