An Improved Fault Diagnosis in Stand-Alone Photovoltaic System Using Artificial Neural Network

• Published in: Iranian journal of science and technology. Transactions of electrical engineering 2024-03, Vol.48 (1), p.325-336
• Main Authors: Sabri, Nassim, Tlemçani, Abdelhalim, Chouder, Aissa, Merrouche, Walid
• Format: Article
• Language: English
• Subjects: Artificial neural networks; Electrical Engineering; Engineering; Fault diagnosis; Faults; Multilayers; Neural networks; Photovoltaics; Real time; Research Paper
• ISSN: 2228-6179; 2364-1827
• DOI: 10.1007/s40998-023-00671-0

This paper proposes an improved fault diagnosis for stand-alone photovoltaic (SAPV) system using artificial neural network (ANN) and power loss parameters as inputs. Unlike the classical power analysis approach that fails to find precisely the fault type, this method can accurately identify the fault class and can be used in real-time applications. The development of the ANN fault diagnosis model goes through data of both normal and faulty operation of the SAPV system. These data are obtained from either a real measurement to represent the normal operation where simplicity and safety are concerned or from the simulation in which the data of faults could hardly and costly be obtained. Three common types of ANN were trained, tested, and compared to choose the most efficient network to predict the faults in the system. The results indicate that multi-layer-perceptron network type is the most accurate network to recognize the faults with 95%. In addition, some test has been carried out in real-time to show their effectiveness.