Jack Kelly

Proportion of *variation* in mains explained by each submeter

1

Consider a house which has a large (unmetered) vampire (always-on) load. But every other load is submetered. In this situation, the `proportion_of_energy_submetered` might be low, even though every variation in...

Don't bother keeping info from each 'true' meter. Instead pretend that we only have one meter. So we expose per_period and combined as usual. Done for EnergyResults although it needs...

Metrics need to handle case where disag happens *across* houses

7

at the moment, metrics assume that meter 1 in the disag output should look the same as meter 1 in the ground truth. This will not be true if disag...

If the house used for disag is different to the house used for training then the appliance instance number is likely to be wrong. So need to use new instance...

`Electric.get_activations()` can be given parameters such as `min_off_duration` and `min_on_duration` to smooth over short periods when an appliance's power draw might be below `on_power_threshold` but the appliance is still 'on'...

All the other NILM algorithms have been adapted to use `Disaggregator._save_metadata_for_disaggregation()` instead of having to 'manually' generate the metadata.

Format docstrings so they pass `pydocstyle`'s criteria

1

## Detailed Description We plan to use [`pydocstyle`](http://www.pydocstyle.org/en/stable/) to automatically check that our docstrings conform to a standard format. It would be amazing to have help tweaking our existing docstrings...

https://wandb.ai/openclimatefix PyTorch Lightning has easy integrations with Weights and Biases.

https://developer.nvidia.com/tensorrt