This project aims to advance in the digitalization of renewable energy control and maintenance operations in collaboration with companies in the energy sector. The ultimate goal is to achieve an ecological transition based on the production of electricity from renewable and clean sources such as sun.
In this way, the project proposes the development of machine learning and big data algorithms, as key areas and pillars of the digital transition, for predictive maintenance in renewable energy plants.
One of the main difficulties of streaming data analysis is the detection of anomalies and, in addition, it constitutes one of the problems with the largest number of real applications. Both in industry 4.0 and in the field of health, the so-called Internet of Things (IoT) proposes that sensors provide data for analysis. This analysis can be carried out in near real time or in batch depending on the characteristic of the problem to be solved. Anomalies are occurrences of events that generally must be warned either before they occur or immediately after they occur. One anomaly detection problem is proposed in this project, in particular the predictive maintenance of photovoltaic panels. In this case, the objective is to predict when a malfunction will occur in a photovoltaic solar plant. For this purpose, two different approaches are proposed: one based on determining when the behavior of a string of photovoltaic panels differs from the actual generation of the other panels at the same time, and the other based on when the actual energy production deviates from the expected production