Satellite based snow identification and its impact on monitoring photovoltaic systems

Kurzfassung

Satellite based earth observation allows the detection of snow cover and the discrimination of cloud from snow cover using multispectral measurements. Using this data enables photovoltaic (PV) plant management to distinguish failures due to snow coverage on a PV system from other error sources. It is also possible to improve yield estimates in solar siting. This paper gives an overview on satellite-based snow cover information which is available in different spatial scales. Results of a validation study from January to April 2006 with ground measurements from German and Swiss meteorological stations are presented. Quality measures introduced are the „false alarm rate“, the „error due to underestimation“, the „availability” of the data sets, and the „classification accuracy“. Depending on how defensively satellite measurements are evaluated, the error shifts from too much recognized snow (error due to underestimation of surface solar irradiance) to too little recognized snow causing a false alarm in PV monitoring. For pure power plant operations monitoring the data record of LSA SAF is the most suitable as it has a symmetrical and small error pattern (false alarm rate 26 % / error due to underestimation 23 %), but the data availability is low (65 %). The IMS data set has a low false alarm rate (4 %) and good data availability (100 %), but a large error due to underestimation (59 %). Also, the DLR data set has a rather small symmetrical error pattern (false alarm rate 37 % / error due to underestimation 26 %) and a good availability (99 %). If a cumulative snow cover algorithm is applied to achieve information on every day, the preciseness of all datasets declines and both the DLR and the LSA SAF datasets become comparable in their results.