NASNet-LSTM based Deep learning Classifier for Anomaly Detection in Solar Photovoltaic Modules

Abstract

Many nations are increasing the installations of solar photovoltaic (PV) modules for clean energy production. Such PV modules are considered to be cost effective if the module's operation lifetime is more than twenty years. In real-time, the PV modules undergo degradation due to hotspots, defects and other anomalies resulting in reduced operation lifetime. Infrared (IR) Thermography is a Non-Destructive Testing (NDT) method that can be used in identifying such anomalies present in PV modules. However, the IR thermography requires Artificial Intelligence (AI) based classification techniques to detect the anomalies. This research article proposes a deep learning classifier, based on NASNet-LSTM for the identification of electrical and non-electrical anomalies occurring in PV modules. NASNet is a convolutional neural network (CNN) based classifier when combined with Long Short-Term Memory (LSTM) Networks performs classification with an accuracy of 84.75% considering the raw dataset used in this research. The results are validated by comparing the accuracies with other models. The study concludes that NASNet-LSTM performs well in the anomaly detection of PV modules.