A deep learning-based strategy for fault detection and isolation in parabolic-trough collectors

Solar plants are exposed to the appearance of faults in some their components, as they vulnerable action external agents (wind, rain, dust, birds …) and internal defects. However, it is necessary ensure a satisfactory operation when these factors affect plant. Fault detection diagnosis methods essential detecting locating faults, maintaining efficiency safety This work proposes methodology for isolating parabolic-trough plants. It based on three-layer composed neural network obtain preliminary classification between three types fault, second stage analyzing flow rate dynamics, third defocusing first collector analyze thermal losses. The has been applied by simulation model ACUREX plant, which was located at Plataforma de Almería. confusion matrices have obtained, with accuracies over 80% using layers hierarchical structure. By forcing all layers, exceed 90%.