A fault severity quantification approach of photovoltaic array based on pre-estimation and fine-tuning of fault parameters

Harsh outdoor operations may cause various abnormalities or faults of photovoltaic (PV) array, decrease the energy yield and lifespan, and even cause catastrophic events. Recently, many approaches have been successfully applied to the fault diagnosis for PV arrays. However, few studies investigate the evaluation and quantification of fault severity. The quantified fault severity can facilitate the fault severity-dependent maintenance of PV system. In this paper, a fault severity quantification approach based on pre-estimation and fine-tuning of fault parameters is proposed. The key features of the I–V characteristics under different faults are determined to train a backpropagation neural network for estimating the preliminary diagnosis and quantification results. Then, the particle swarm optimizer is further used to locally optimize the estimated results to improve the accuracy of quantified fault severity. Compared with other diagnosis approaches, the experimental results verify that the proposed fault diagnosis and quantification approach obtains higher accuracy with decent computational speed. The proposed method is suitable for the fault severity-dependent maintenance of the PV systems.