Table 7.

Comparison of recently proposed models.

Reference	Process of generating data	Pre-processing Technique	Defect analysed	Use of ML technique	Brief summary
[245]	In Lahore, Pakistan, infrared thermal images were obtained utilizing the PV string's modules.	The utilization of a data fusion methodology for the feature extraction of RGB texturing.	Hotspots	SVM model achieved a training and testing correctness of 96.80 and 92.00 % resp.	SVM model is employed for the purpose of classifying thermal images of PV panels into 3 categories: strong, non-defective hotspots, and defective hotspots.
[246]	The I–V curve data obtained from photovoltaic (PV) modules is specifically centred on the analysis of hotspots.	Minimum-maximum averaging	Hotspots	Quantitative analysis of DT, SVM, KNN, and DC. DC showed the best detection result, whereas DT showed the worst.	Four ML models are used to diagnose early-stage hotspots in PV modules.
[240]	The design and modelling of photovoltaic systems using PSIM and MATLAB. The Agilent 34,970 datalogger is utilized for the purpose of capturing data from a 9.54 kWp Algerian grid-connected photovoltaic (PV) system.	The Canonical Artificial Bee Colony technique is utilized for the extraction of the constraints of the one diode model.	PV fault	PNN is a type of ANN that utilizes feed-forward and backpropagation algorithms for training and learning. The proposed PNN demonstrates a detection efficiency of 82.34 % and a diagnosis efficiency of 98.19 %.	The proposed PNN model has been designed for the purpose of fault identification and diagnosis inside the DC side of PV.
[247]	I–V curve data for 960 W PV array obtained from RELab JiJel university in Algeria using Prova 210 IV tracer.	The technique of dimensionality reduction via PCA and standardization.	The phenomenon of partial shade, line-to-line deterioration, and dust deposition.	The following machine learning algorithms were utilized in the study: NB, KNN, SVM, LR, DT, RF, and NN.	This investigation focuses on the application of multiple individual and combined ML for the purpose of detecting and classifying various types of PV problems.
[248]	The design of a PV with MATLAB and Simulink is being considered. The present study aims to analyse the real-time irradiance and temperature data obtained from a grid-linked PV system located in Agartala.	The technique of array capture loss was employed in the training of the machine learning algorithm.	Common flaws include Line Ground, Line, OC, arc, shading, and deterioration.	This article covers Cat Boost, LGBM, and XGBoost. The LGBM performed best, followed by CatBoost.	The PV system was modeled using Simulink and real-time data to assess and identify frequent issues.
[249]	The images obtained from photovoltaic (PV) modules that are put outside.	The technique of Gaussian blurring is commonly employed in the field of image processing.	PID and LeTID	PCA and KNN were employed in the analysis. The obtained accuracy rate with KNN was 89 %.	Modelling PCA-KNN for PID and LeTID prediction. Field-installed modules were utilized to capture EL pictures.