This paper proposes a novel method for voltage dip classification using deep convolutional neural networks. The main contributions of this paper include: (a) to propose a new effective deep convolutional neural network architecture for automatically learning voltage dip features, rather than extracting hand-crafted features; (b) to employ the deep learning in an effective twodimensional transform domain, under space-phasor model (SPM), for efficient learning of dip features; (c) to characterize voltage dips by two-dimensional SPM-based deep learning, which leads to voltage dip features independent of the duration and sampling frequency of dip recordings; (d) to develop robust automaticallyextracted features that are insensitive to training and test datasets measured from different countries/regions. Experiments were conducted on datasets containing about 6000 measured voltage dips spread over seven classes measured from several different countries. Results have shown good performance of the proposed method: average classification rate is about 97% and false alarm rate is about 0.50%. The test results from the proposed method are compared with the results from two existing dip classification methods. The proposed method is shown to outperform these existing methods.

In this paper, a deep learning (DL)-based method for automatic feature extraction and classification of voltage dips is proposed. The method consists of a dedicated architecture of Long Short-Term Memory (LSTM), which is a special type of Recurrent Neural Networks (RNNs). A total of 5982 three-phase one-cycle voltage dip RMS sequences, measured from several countries, has been used in our experiments. Our results have shown that the proposed method is able to classify the voltage dips from learned features in LSTM, with 93.40% classification accuracy on the test data set. The developed architecture is shown to be novel for feature learning and classification of voltage dips. Different from the conventional machine learning methods, the proposed method is able to learn dip features without requiring transition-event segmentation, selecting thresholds, and using expert rules or human expert knowledge, when a large amount of measurement data is available. This opens a new possibility of exploiting deep learning technology for power quality data analytics and classification.

The proper operation of single-phase and three-phase grid-connected power converters depends on the synchronization with utility networks. The major challenge of the synchronization is how to quickly and precisely extract the ac signal and fundamental positive sequence in single-and three-phase power systems respectively. This paper proposes a new detection technique based on a modified Kalman filter and the generalized averaging method (KF-GAM). The method has an open loop structure, and uses the orthogonal signals which are obtained directly from the Kalman filter. The resulted detection system is very simple and robust even in the presence of power quality disturbances, such as voltage imbalance, harmonics, and voltage fluctuations. The proposed technique can detect the fundamental and harmonics frequencies within or less than half a cycle in all situations such as small and considerable frequency variations. Meanwhile, the method guarantees the zero steady-state error in complicated harmonic scenarios, including all typical single-phase and three-phase harmonics. Various case studies are assessed and the performance of the proposed detection method is verified by experiments.

This paper compares different methods for voltagedip characterization. Those methods are based on earlier proposed algorithms for extracting three-phase characteristics (dip type, characteristic voltage, and so-called "PN factor"). The difference between the 12 methods being studied in this paper is in the way in which the time variation of those characteristics is treated to result in single-event characteristics. The methods are applied to 259 measured voltage dips and the performance of the different methods is compared. It is found that small differences in method can result in big difference in results. From the comparison, two methods are selected and recommended for inclusion in international standards.

No abstract