Accurate Identification of Electrical Equipment from Power Load Profiles

It is essential for the power industries to identify the running electrical equipment automatically. For power monitoring, the load profile data vary with the equipment’s types. Proceeding from the fundamental features of load time series, we propose a method to identify electrical equipment from power load profiles accurately. Aiming to improve the classification accuracy and generalization performance of convolutional neural network (CNN), we combine the training process of generative adversarial networks (GANs) with CNN, which employs the generated samples to enhance the classification accuracy. The CNN and discriminator in our approach share the first convolution layer for extracting richer features. We evaluate our method on UCR data sets comparing with 12 existing methods. Furthermore, we compare our model with LSTM, GRU and CNN on the electrical equipment load data, which is from industries in certain area. The final results show that our model has a higher equipment identification accuracy than other deep learning models.