The Deep Jerk-regularized Extreme Learning Machine for Complex Power Quality Disturbance Classification

In order to make full use of a large number of unlabeled measured samples in complex power quality disturbance classification, a semi-supervised disturbance classification method based on the deep Jerk-regularized extreme learning machine (DJRELM) is proposed. The method stacks the Jerk-regularized ELM autoencoder (JRELM-AE) to realize the automatic feature extraction while exploring the intrinsic geometric structure of the unlabeled data. The combination of semi-supervised Jerk-regularized ELM and threshold learning ELM in the classification layer extends the multi-layer network to multi-label semi-supervised learning. Experimental results show that the proposed method outperforms several ELM-based supervised and semi-supervised algorithms as well as the state-of-the-art multi-layer ELM algorithms and deep convolutional neural networks under different noise environments. The new method has both academic and practical value.