Washing Machine Detection

The Washing Machine algorithms has been developed extensively throughout the past few years. From a model that was using the power data to extract information based on algorithmic logic it became a Machine Learning model, based on 1 second and 10 second data. Depending on the data rate the N2G devices receive data the WM module is set up automatically to the 1 or 10 second data rate model.

Before the prediction phase, there is a pre-processing one, where we calculate the first differences of the power measurements, then data are set between a min and a max value and then are normalised.

The process makes a prediction every 5 minutes.

Payload of 17 bytes is:
spike_lower (4) | t_spike_upper (4) | t_ripple (4) | t_cons_off (4) | phase (1)

Right now LIBN2G supports only phase change from the platform if that is the case in the feature maybe the format of the payload will change as well.