EnerGAN++: A Generative Adversarial Gated Recurrent Network for Robust Energy Disaggregation

Energy disaggregation, namely the separation of aggregated household energy consumption signal into its additive sub-components, bears resemblance to (source) problem and poses several challenges, not only as an ill-posed problem, but also, due unsteady appliance signatures, abnormal behaviour that is usually detected in appliances operation existence noise signal. In this paper, we propose EnerGAN++, a model based on Generative Adversarial Networks (GAN) for robust disaggregation. We attempt unify autoencoder (AE) GAN architectures single framework, which achieves non-linear power source separation. EnerGAN++ trained adversarially using novel discriminator, enhance robustness noise. The discriminator performs sequence classification, recurrent convolutional neural network handle temporal dynamics time series. particular, proposed architecture leverages ability Convolutional Neural (CNN) rapid processing optimal feature extraction, among with need infer data character dependence. Experimental results indicate method's superiority compared current state art.