Power Output Variability in randomly spaced dutch urban rooftop solar photovoltaic systems

In contrast to several studies on variability within or between (virtual) MW-scale photovoltaic (PV) plants [e.g. T.E. Hoff and R. Perez (2010)] we investigated the aggregated behavior of randomly spaced residential PV-systems in a densely populated urban area (≈ 100 km²) in and around Utrecht, the Netherlands. The behavior of such a realistic distributed PV-network is assumed to be substantially different than equivalent and regularly spaced PV panels in a compact array. Nevertheless, qualifiers and instruments like inter-system Ramp Rate (RR) correlation and Mean Output Variability (MOV) can be used to generate statistically important information about a generic PV-network. The characteristic decorrelation length scale over which variations between PV-systems reach a distance-independent plateau (within 1 - e^-3 ≈ 95%), was found to range from 0 to approx. 3 km, with a mean value of between 0.4(2) km and 1.12(17) km for the full year, depending on the time step of the power output time series. This length scale is typical for an urban environment and is important for reduction of variability in aggregated output variability of PV-systems. The distance-independent variability plateau was found to be linearly dependent on daily MOV values.