SPLTRAK Abstract Submission
State of the Art of Modelling Soiling and Snow Losses in PV Systems
Sébastien ARBARETAZ, Murielle STEPEC, Eszter VOROSHAZI
INES/CEA, Le Bourget du Lac, France

The cost of photovoltaic energy has considerably decreased in recent years and is becoming more and more cost-effective. As a result, the installed PV capacity has increased significantly and countries with low irradiance and high snowfall are now investing in photovoltaic. In this context, many studies are focusing on the optimization of production and the modelling of losses due to soiling and snow.
This paper provides a review of soiling and snow losses modelling methods in photovoltaic systems in the last few years. It highlights the main principles of soiling and snow models and the most important parameters that influence soiling and snow losses in photovoltaic plants. It also presents new trends in the methods used for the realization of even more reliable models. A large variety of methods is detailed such as artificial neural network, linear regression, physical approach, data inference, Markov chain and use of data from neighbored plants and historical data.
The most common parameters of models are ambient temperature, irradiance, wind, relative humidity, PM, rainfall and snowfall. Tilt angle is not taken as a parameter in most of the dust soiling models but it is a major parameter for snow losses models. Most of the models are trained and tested with data from the same location or country, thus they describe the environment in which they were constructed. This paper suggests as possible improvement that models could be to train and test in various environment in order to evaluate their flexibility in other environment.