Nowadays plant monitoring and control of renewable energy sources can take advantage of the use of unmanned aerial technologies. This manuscript aims to develop an image post-processing tool for remote aerial images able to help operation and maintenance operators in photovoltaic (PV) defects detection using light unmanned aerial vehicles (UAVs). In particular, PV systems deployed in the field in the last ten years show often critical behaviour with a range of failures able to compromise the performance and energy yield of the power plants, thus they can greatly benefit of such novel technologies and tools. The described procedures are here tested on real plant data, with different kind of sensors, in order to find out potential advantages in fast fault detection tasks, using an automatic system for PV plants' mapping. The results of this research will be reported in order to provide an understanding of potential impact of image processing techniques based on UAV in the renewable energy sector.

Photovoltaic (PV) plant monitoring and maintenance has become an often critical activity: the high efficiency requirements of the new European policy have often been in contrast with the many low-quality plants installed in several countries over the past few years. In actual industrial practices, heterogeneous information is produced, and they are often managed in a fragmented way. Several software tools have been developed for obtaining reliable and valuable information from the PV plant's raw data. With the aim of gathering and managing all these data in a more complex and integrated manner, an information managing system is proposed in this work-it is composed of a structured database, called the Photovoltaic Indexed Database, and a user interface, called the Digital Map, that allows for easy access and completion of the information present in the database. This information managment system and PV plant digitalization process is able to analyze and properly index the IR in the database, as well as the visual images obtained in photovoltaic plant monitoring.

An accurate forecast of the exploitable energy from Renewable Energy Sources is extremely important for the stability issues of the electric grid and the reliability of the bidding markets. This paper presents a comparison among different forecasting methods of the photovoltaic output power introducing a new method that mixes some peculiarities of the others: the Physical Hybrid Artificial Neural Network and the five parameters model estimated by the Social Network Optimization. In particular, the day-ahead forecasts evaluated against real data measured for two years in an existing photovoltaic plant located in Milan, Italy, are compared by means both new and the most common error indicators. Results reported in this work show the best forecasting capability of the new "mixed method" which scored the best forecast skill and Enveloped Mean Absolute Error on a yearly basis (47% and 24.67%, respectively).

No abstract

No abstract