In this work we propose a new sensor concept to evaluate the degradation of PV arrays due to soiling. It is based on I-V curve analysis coupled with artificial vision inspection of a reference PV module to quantify and identify the type of dirt. In order to assess the usefulness of this approach in the automatic scheduling of maintenance interventions in smart-grid PV modules, we developed a Simulink model of a DC nanogrid to test different control strategies. Early experimental results are also shown demonstrating the feasibility of the approach.