Modeling solar irradiance smoothing for large PV power plants using a 45-sensor network and the Wavelet Variability Model

With increasing penetrations of solar photovoltaic (PV) power on the electric grid, the variability of solar irradiance, and therefore power, is important to understand because variable resources can challenge grid operations. Predicting PV variability using one irradiance sensor does not account for the smoothing of irradiance over the area of a power plant. Smoothing is examined using two methods: averaging measurements from many irradiance sensors, and using the Wavelet Variability Model developed by Lave et al. (2013). This work provides new experimental testing: comparison of the smoothing over a 30. MW power plant using the average of 25 sensors to the WVM. The results show that an aggregation of 25 sensors predicts more variability than the WVM on short timescales, suggesting that more than 25 sensors would be required in order to predict the same power variability as the WVM. In addition it is shown that the reduction in daily Variability Index depends on the daily cloud speed scaling coefficient.