Intelligent Approaches Based Parameters Prediction And Cell Defects Classification Of Photovoltaic Power Generation And Transmission System

Photovoltaic power generation and transmission systems are sensitive to irregular and uncontrolled weather changes.PV power generation and transmission are highly dependent upon the weather parameters such as irradiation and temperature.These weather parameters experience high variation throughout the day and causes high variation in PV power generation and transmission line parameters leading to uncertainty in power production and transmission to the consumer end.Accurate prediction of PV power generation and transmission line parameters can reduce this uncertainty and helps in improvement of power system performance and reliability.Therefore,it is essential to recognize the long-term effects of meteorological parameters on the PV power generation and transmission system.The worldwide growing demand of photovoltaic power generation and transmission system,and their large-scale installation requires an accurate prediction of parameters of photovoltaic power generation and transmission line,and the classification of their cell defects.This parameters prediction and defects classification is necessary to enhance the PV system performance and reliability.In this background,the present thesis focuses on prediction of parameters of PV power generation and transmission line,and cell defects classification of Photovoltaic module.This study consists of three major parts i.e.prediction of PV power generation parameters,and prediction of power transmission parameters,and automatic cell defects classification using machine learning algorithms as discussed below.In the first and second part of this research work,the prediction of PV power generation and transmission line parameters have been conducted.For this purpose,machine learning algorithms are used.The predicted parameters of PV power generation are Open Circuit Voltage(Voc),Short Circuit Current(Ish),Series Resistance(Rs),Shunt Resistance(Rsh),Maximum Current(Imax),Maximum Voltage(Vmax),Maximum Power(Pmax),DC power,AC power,and system efficiency.Whereas predicted parameters of power transmission line include resistance,inductance,capacitance,voltage drop,and power losses.Furthermore,four different seasons based SVM models(i.e.spring,summer,autumn,and winter models)are considered for prediction of PV power generation parameters whereas only two seasons based SVM and ElasticNet models(i.e.summer and winter models)are considered for prediction of power transmission line parameters.The performance results of PV power generation and transmission line parameters' models are presented in terms of Mean Relative Error(MRE)and Root Mean Square Error(RMSE).Additionally,the performance results are to show that which algorithm has better prediction accuracy,practicability,and feasibility.Third part of this research work is carried out for automatic defects classification of PV cell in electroluminescence images.Two machine learning approaches,features extraction based SVM and convolutional neural network(CNN)are used for the classification of solar cell defects.Suitable Hyperparameters,algorithm optimizers,and loss functions are used to achieve the best performance.Cell defects are divided into seven classes that include one non-defective and 6-defective classes.Feature extraction algorithms such as HOG,KAZE,SIFT and SURF are used to train SVM classifier.At the end,the performance results are compared.It was also observed that CNNs accuracy for cell defect classification is 91.58%which outperform the state-of-the-art methods.Based on the performance results of the developed models of this study,the present research work may provide technical guidance to photovoltaic power system and transmission line design engineers and technicians to improve the power system efficiency.