A soiled photovoltaic module is modeled using regression and neural networks.

Experimentally obtained data is used for the modeling.

Particle size composition of the soil is used as a quantifying parameter.

Influence of particle size composition on the losses is studied from the models.