Hi MichaGue_01, -For that, i am guessing you should know how the waveform will be for known device and the current unknown graph. -Then you can use correlate vi, with predefined values and best match can be identified.
As LabVIEW is a programming language, you can program such detection. Is there anything already written to do what you want? Probably not.
One thing which complicates the issue is that multiple devices can turn on and off at different times. And the time resolution of the smart meter data is relatively low, unless you can get the raw data from within the meter itself. What I would look at is creating a table of power consumption patterns for each of the devices. In particular look at the turn on and turn off transients as those will tend to be more prominent in the smart meter data. Large items like the oven and air conditioners are relatively easy to spot. Can you distinguish between a few 26 W CFL lamps and a computer and monitor turning on? If the computer and some lights turn on at the same time will you be able to distinguish that from the refrigerator which might use about the same total power?