A LSTM-based deep learning method with application to voltage dip classification

Abstract

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.

Power quality disturbance detection and classification using signal processing and soft computing techniques: A comprehensive review

“…[219][220][221][222][223] Ma et al 219 proposed stacked autoencoder (SAE) as a DL architecture to extract advanced features from PQ disturbance signals for automatic PQD&C. Moreover, variances of signals and a PSO algorithm were employed to support PQD&C process. It has been effectively used in various diverse research fields, such as speech recognition, human face recognition, computer vision, signal, image, and information processing.…”

“…The application of DL algorithms on automatic power system fault detection and classification approaches was carried out by various authors in the previous studies. [219][220][221][222][223] Ma et al 219 proposed stacked autoencoder (SAE) as a DL architecture to extract advanced features from PQ disturbance signals for automatic PQD&C. Moreover, variances of signals and a PSO algorithm were employed to support PQD&C process. Liu et al 220 presented a new PQD&C method based on singular spectrum analysis (SSA), curvelet transform (CT), and deep convolutional neural networks (DCNNs).…”

“…convolution neural network (CNN), recurrent neural network (RNN), identity RNN (I-RNN), long short-term memory (LSTM), gated recurrent units (GRU), and CNN-LSTM were considered in this article. Balouji et al 222 presented a DL-based approach for automatic voltage sag recognition. The technique includes a powerful LSTM architecture, which is a variant of RNNs.…”

Recognition method of voltage sag causes based on Bi‐LSTM

“…Recently, only a small number of works have been published on deep learning methods of the recognition of voltage sag causes [16,17]. Reference [16] proposes a new method for the recognition of voltage sag causes based on Long Short-Term Memory (LSTM), and Ref. [17] shows that Convolutional Neural Network (CNN) is a suitable tool for recognizing voltage sag causes.…”

“…For Ref. [16], using the raw record data instead of the root mean square (RMS) sequence as the input to the LSTM may be a better choice, and how to improve the performance of the model requires further study. For Ref.…”