The national target set for 2030 in the National Integrated Energy and Climate Plan (PNIEC) is to achieve an installed photovoltaic capacity of 52 GW with a production of 74 TWh/year.
To achieve this goal, in addition to new photovoltaic installations and the development of more efficient applications and technologies, it is necessary to use tools that can preserve the performance of the plants through efficient operations, which do not increase the cost of the electricity produced from photovoltaics (LCOE – Levelized Cost of Energy) but, on the contrary, reduce it through higher production at the same investment cost.
In recent years, Machine Learning (ML) techniques have been widely used for fault detection in various application domains. In the photovoltaic sector, these techniques are also finding interesting applications in the development of Fault Detection and Diagnosis (FDD) algorithms. Fault diagnosis can provide valuable guidance to operators in charge of operation and maintenance, allowing them to optimize maintenance times and repair costs.
For this purpose, in a previous activity of this project, RSE analyzed the main causes of faults with the greatest economic impact on the life of a photovoltaic system and designed a PV Fault Facility capable of inducing fault conditions in small-scale systems.
With this work, RSE has started the implementation of the PV Fault Facility and analyzed the behavior of system components during induced fault conditions. Data acquired from fault tests are used in the training and validation of FDD algorithms.
The use of real data has been an added value for the development of the FDD algorithm, as it allows for consideration of issues that can arise in the field and are often overlooked when considering data generated from theoretical models.
Finally, as part of a research contract with the Politecnico di Milano, the state of the art of ML techniques was reviewed, and the development of an FDD algorithm was initiated. This algorithm will be completed and validated in a subsequent project activity using data generated by the PV Fault Facility.