Widespread deployment of photovoltaic (PV) energy production systems is currently limited by uncertainty about how high-frequency variability in PV output can be accommodated by the large-scale power grid. Solar flux variability at a point has been shown to cause ramps as high as 1000 Wm^-2 over one minute intervals. Several studies have shown that the correlation between the solar irradiance measured at two locations decreases as the distance between the measurement sites increases, indicating that PV array output is smoothed when the array size is increased or the outputs of several stations are combined. However, the level of correlation between sites is a function of meteorological conditions in addition to site separation.

Here we present results of a study of irradiance data collected at a high-density radiometer network on Oahu Island. This network contains 17 instruments over an area of about 1 km x 1 km, with station separations ranging from 100 to 1200 m. This unusually high-density network makes it possible to compare irradiances and ramps at the scale of individual PV installations. Using a complete year of measurements, we show ramp statistics for both individual radiometers and simulated plants with capacities of 5-60 MW in terms of absolute irradiance and relative changes. Some seasonal variability is found despite the fairly uniform weather conditions at this location. We also examine the correlations among the time series from all possible combinations of stations at scales from 10 to 300 seconds. A range of results is found from different site pairs at the same distances. For this reason, we separately examine correlations for site pairs oriented along and across the prevailing trade wind direction. Significant differences are found for these two cases, which depend on time scale as well as separation distance. A possible explanation for this finding will be presented in addition to the overall statistical results.