Intelligent fault diagnosis of photovoltaic array based on variable predictive models and I–V curves

Abstract

Photovoltaic (PV) arrays are installed in outdoors and continuously exposed to harsh environmental conditions, which are susceptible to suffer from abnormal and various kinds of faults, reduce the efficiency of power conversion and the lifetime of PV modules, and even cause electric shock and fire. In recent years, many methods of machine learning and deep learning have been successfully applied in the field of fault diagnosis for PV array. However, these methods still exist some issues: neural network training requires a large number of data samples and is prone to overfitting and underfitting; network structure and hyperparameters are difficult to determine, mainly rely on extensive experiments; The kernel extreme learning machine (KELM) and support vector machine (SVM) require kernel functions and hyper parameter adjustment, and the selection of kernel functions and hyperparameter will directly affect the accuracy of fault diagnosis; To solve the above issues, an intelligent fault diagnosis method for PV array based on the variable predictive models and I–V curves is proposed. Firstly, the impact of different operating states on the I–V curve of PV array under standard test condition (STC) is analyzed. Then, the five-dimensional effective feature variables are extracted from the key points of the measured I-V curve as input of the fault diagnosis model. Since the output of the electrical characteristics of the PV array is affected by faults and environmental factors, the extracted feature variables are normalized to eliminate the influence of environmental factors. Finally, the normalized feature variables are used as the input of variable predictive model to identify the fault types of the PV array, which reduces the computational cost and avoids the problem that the kernel function and hyperparameter are difficult to determine. Moreover, the proposed method is verified by simulation and measured data and compared with other machine learning algorithms. The results indicate that the proposed fault diagnosis method has high accuracy and reliability.