Is this spike in the generation? If it is, this occurs during cloudy weather, and are spikes in the actual generation, not a monitoring anomoly. The theory (I don't know if it has been proven, but is the accepted explation) is that the clouds act as reflectors as they approach the sun, increasing the sun hitting the panels, then the reduce it as they pass in front of the sun. This creates a spike followed by trough.
I know it seems like an outlandish explanation, I myself wouldn't believe it except that I heard it from Bill Brookes and have seen the phenomenon in several different systems, only on cloudy days.