Hi Scott,
In SAM, the meaning of the kW of electricity generated by the system in each hour depends on the solar resource data. For hourly data from the NREL National Solar Radiation Database (NSRDB), the W/m2 of solar irradiance for each hour is an instantaneous value at the midpoint of the hour, so the kW of electricity generated by the system in a given hour is the power output at the midpoint of the hour. If the W/m2 data from PVGIS represents hourly averages, then the kW of electricity that SAM calculates for each hour would be the average over the hour.
The load data should match the weather data. If you are using weather data with instantaneous solar irradiance data like that from the NSRDB, the load data kW should be measured power values at the midpoint of the hour. If the solar irradiance data represents an average over each hour, the load data should also represent an average kW over the hour.
I think the method you use to calculate the self-consumption rate from the SAM results is correct. For the PVWatts / Residential model, The value of Electricity from System to Load in kW for each hour is calculated as follows:
IF
System Power Generated <= Electricity Load
THEN
Electricity from System to Load = System Power Generated
ELSE
Electricity from System to Load = Electricity Load
The annual values for each of those variables is the sum of the 8760 hourly kW values.
This answer is for hourly simulations, but would be similar for smaller time steps. SAM's simulation time step depends on the weather data time step, and the load data time step must match the weather data time step. SAM can use weather data with a simulation time step as small as one minute, so if you had access to 15-minute weather data and 15-minute load data, you could do this analysis based on a 15-minute simulation over the year.
Best regards,
Paul.