Non-intrusive load disaggregation via a fully convolutional neural network: improving the accuracy on unseen household

Abstract

The application of load disaggregation techniques based on neural networks is often limited to users included in the training dataset. A methodology based on techniques typical of the semantic segmentation of images has been proposed for this task, which allows to obtain a high accuracy and good generalization. We introduce here a novel data augmentation technique for improving forecasts for unmonitored houses that does not require any interaction with the user, nor further measurements of consumption of household appliances.