Implementations of NILM disaggegregators using Neural Networks, using NILMTK and Keras.
Most of the architectures are based on Neural NILM: Deep Neural Networks Applied to Energy Disaggregation by Jack Kelly and William Knottenbelt.
The implemented models are:
- Denoising autoencoder (DAE) as mentioned in Neural NILM (see example)
- Recurrent network with LSTM neurons as mentioned in Neural NILM (see example)
- Recurrent network with GRU. A variation of the LSTM network in order to compare the two types of RNNs (see example)
- Window GRU. A variation of the GRU network in that uses a window of data as input. As described in Sliding Window Approach for Online Energy Disaggregation Using Artificial Neural Networks by Krystalakos, Nalmpantis and Vrakas (see example)
- Short Sequence to Point Network based on the architecture in original paper (see example)
Note: If you are looking for the LookbackGRU folder, it has been moved to http://github.com/OdysseasKr/online-nilm. I try to keep this repo clean by using the same train and test sets for all architectures.