Probabilistic Assessment of Narrowband vs Broadband Solar Irradiance Temporal Variability in Ottawa

Abstract

Using a recently-created database for Ottawa, Canada, a 9-month longitudinal study of the solar irradiance, measured with a custom spectral pyranometer every 250ms, was conducted with a specific focus on comparing the narrowband and broadband response to the temporal variability. Deterministic diurnal and orbital dependencies of the spectral irradiance were removed by clear sky normalization, and the resulting clear sky index was forward differenced across time steps ranging from the sub-second to ~30 minutes. The stochastic behavior of this spectral clear sky index increment was assessed by kernel density estimates of the probability distributions for data from each of the nine narrow wavelength channels of the spectral pyranometer and the derived broadband global horizontal irradiance (GHI). Scaling analyses of their peak densities and their full widths at half maximum (FWHM) with increment time step revealed power law scaling with consistent stationary breaks at very-short and short times. Broadband scaling was consistent with some narrowband dependencies, but not with others, which may reflect the wavelength dependence of different sky conditions. The existence of three distinct scaling regimes, each of which coincides with an operationally significant time period (primary/secondary control, automatic generation control, and real time market dispatch), has implications for short term probabilistic forecasting.