High-accuracy classification of power quality disturbances using hybrid binary neural network

The identification and classification of power quality disturbances has always been a challenge for researchers. The rapid and accurate classification of different disturbance signals plays an important role in improving power quality. In order to ensure reliable and high-quality supply of electricity, this paper proposes a hybrid one-dimensional binary convolutional neural network (BNN), which solves the problem of low accuracy of BNN through a special network structure. BNN is applied to power quality disturbance classification for the first time. While retaining the efficiency of BNN, the classification accuracy reaches the level of full-precision convolutional neural network (CNN), which provides a basis for the deployment of power quality disturbance classification on embedded devices.