Phase difference in hourly data of ERA5's ssrd

Why is there a phase difference of about two hours between the hourly data of ERA5 ssrd and the observation data? I integrate the observation data for one hour and then divide it by the integration time (consistent with the data processing method of ERA5), with the unit of W / m ^ 2

Hi, I have the same problem with you, have you solved this problem?.

My reaction is a bit late, but maybe it helps somebody else.

Just from the top of my hat, based on mistakes I made myself in the past:

Are both data sets using the same time system? ERA 5 uses UTC, are your observed data in local time?

ERA5 sets the timestamp at the end of the averaging period, if you set the timestamp at the start this would account for a shift of one hour.

Is there a mistake in the geographical co-ordinates (especially the longitude)?

The plots seem to show the cumulative data over a longer period. Plotting the data directly and scrolling through the plot could give additional insights. You could also plot the clear sky irradiance and or top of atmosphere irradiance in the same plot.
You could use the PV-lib library (Python and Matlab) for this.