A major goal of this paper is to consider a new paradigm of monitoring and predictive maintenance decision support of distributed hybrid renewable energy power units, driven by data, economic and ecologic requirements. A renewable energy power unit is a multi-component multistate complex system. In this paper, colored stochastic Petri nets are suggested to model, evaluate and forecast the health conditions of components, and to drive and update the data collection process through the unit lifetime. The approach allows monitoring the health conditions of the unit and predicting the reliability of its components, on the base of a non-parametric model.