Review on Deep Neural Networks applied to Low-Frequency NILM

This repo contains data and code that has been used for the publication "Review on Deep Neural Networks applied to Low-Frequency NILM" submitted @ MDPI Energies doi.org/10.3390/en14092390.

This work is a considerable extension of the presentation "DNN for NILM on low frequency Data" that has been done at the NILM workshop 2019. You can find the corresponding presentation here

Content:

• DNN-NILM_Publication-List.xlsx contains the list of the DNN-NILM publications that have been reviewed in the mentioned publication. It corresponds with minor differences in columns and nomenclature to table 2 in the publication and is provided to allow for easy searching and filtering. Abbreviations are explained in the publication.
• Visualize_MAE.ipynb and Visualize_F1.ipynb are the jupyter notebooks that have been used to generate the visualizations in the paper, i.e. figures 3 and 4. Please be aware that citation numbering might have changed in the final publication.
• DNN-NILM_low-freq_Performance.xlsx contains the list of metrics extracted from the reviewed publications. Publications that did
  • not report metrics,
  • report metrics other than F_1-score or MAE or
  • not report metrics according to the relevant evaluation scenario might not appear in the list. The file is the basis for the figures generated with the jupyter notebooks. Some explanations on the columns can be found in the tab Explanations. Please do not expect that all columns are filled up consistently.

In case you are an author of one of the publications and feel that erroneous information has been compiled in our list, do either contact patrick.huber@hslu.ch or open a pull request with your suggested changes. We will appreciate your feedback!