Detection and Classification of Multiple Power Quality Disturbances based on Temporal Deep Learning

Power quality consideration in smart and sustainable environment is regarded as a challenging task, which brings multiple power quality disturbances associated with the same event. In this paper, we propose temporal deep learning based power quality disturbance detection and classification method based on an encode-decoder temporal convolutional neural network (EDTCNN). The proposed method fuses feature extraction and classification into a single block and is capable of capturing regular and uncertainty patterns from long-range sequences such as transient and short duration voltage signals. The proposed method is tested on over 29 classes of power quality disturbances, including single power quality disturbances and multiple power quality disturbances with high level of noises (20dB and 40dB). The superiority of EDTCNN is validated by comparing with several known shallow-based and deep-based methods.