MY DISAGGREGATION
Identifying the list of appliances active on a circuit is a task that could be done by an algorithm;However, an algorithm will never be 100% accurate. On the other hand, this information can be provided very quickly and accurately by the end user-this is the reason why Voltaware decided not to rely on an algorithm for this part.
If an appliance is not registered, the underlying machine learning model is not activated and cannot be detected. The list of appliances can be changed at any time using the mobile app.
The algorithm assumes that the client has provided correct and accurate information. Therefore, when someone reports that they have an electric car, the algorithm is going to search for electric vehicle signatures in the signal.
Obviously, if you have an electric car, you might not charge it every day; the algorithm is capable of deciding if an electric car is present or not on a particular day.
However, during the training phase (which takes one month), the algorithm expects to see all appliances in use. This is crucial for the algorithm to be able to learn the particularities of each appliance (your electric car is not exactly the same as your neighbour's). Again the situation where you have not charged your electric car over the last month is possible - although much less probable -and, in any case, the models are regularly retrained to ensure that no appliance is missed during the training period.
Naturally, an error about a single appliance in the registration is going to cause little disruption on the results, while adding five appliances that are not on the circuit in the registration is going to affect the results a lot.
The list of appliances can be changed at any time from the mobile app.
The appliances we have in the app's property profile are the ones we currently support. These appliances are chosen because of our client's expectations and some of the appliances will be added over time.