| Title |
Design of a Predictor Based on ANN(Artificial Neural Network) for Diagnosis of DC Junction Block of PV System Using Its Internal Temperature |
| Authors |
김홍성(Hong-Sung Kim) ; 김기복(Ki-Bok Kim) ; 이승요(Seung-Yo Lee) |
| DOI |
https://doi.org/10.5370/KIEE.2020.69.6.937 |
| Keywords |
PV(Photovoltaic) System; DC Junction Block; Diagnosis; Artificial Neural Network(ANN); Bath Tube Failure Rate Function; O&M(Operation and Maintenance) |
| Abstract |
A diagnosis model for DC junction block O&M(Operation and Maintenance) of PV system is proposed in this paper. The proposed diagnostic model is based on a bath tube failure rate function and requires a well-functioning predictor to generate a reference signal which evaluates the operating characteristics of the diagnostic junction block. Among the various ML(Machine Learning) techniques, ANN(Artificial Neural Network) is used as a technique for the predictor with design factors such as input, output variables and learning rates, and sufficient data are required to train the diagnostic model by ANN. However, it is difficult to obtain operational data related to the DC junction block that should be used to train and evaluate the proposed diagnostic model. In order to solve this problem, environmental data on outdoor temperatures and solar radiation are downloaded from the domestic meteorological data center in Korea, and the data needed to evaluate the proposed predictor are generated by appropriate assumptions using various references and numerical simulations. The accuracy and generality of the proposed predictor are evaluated by the ? value and the RMSE(Root Mean Square Error). |