Automated Photovoltaic Hotspot Detection Based on IV Curve Characteristics and Machine Learning Classification

Imam Faried Assalam; M Sahrul Ramadan; Henra Jasman; Ihlas; Fitriani; Dea Calista

doi:10.31963/elekterika.v23i1.5801

Authors

Imam Faried Assalam Politeknik Negeri Ujung Pandang
M Sahrul Ramadan Politeknik Negeri Ujung Pandang
Henra Jasman Politeknik Negeri Ujung Pandang
Ihlas Politeknik Negeri Ujung Pandang
Fitriani Politeknik Negeri Ujung Pandang
Dea Calista Politeknik Negeri Ujung Pandang

DOI:

https://doi.org/10.31963/elekterika.v23i1.5801

Keywords:

Photovoltaic, Classification, SVM, KNN, Classification, Hotspot, I-V curve, KNN, SVM

Abstract

Hotspot formation in photovoltaic (PV) modules represents a major degradation issue that reduces system efficiency, reliability, and operational lifespan. Conventional diagnostic techniques such as infrared thermography (IRT) and electroluminescence (EL) imaging provide accurate fault detection but are often constrained by high costs and limited scalability for large PV installations. To address these challenges, this study develops a machine learning-based diagnostic framework for hotspot fault classification using electrical parameters derived from IV curve analysis. Field measurements were obtained using IV curve tracers and environmental sensors, then normalized to Standard Test Conditions (STC). Two classification algorithms—Support Vector Machine (SVM) and K-Nearest Neighbor (KNN)—were employed with various kernel and distance configurations, including Linear, Quadratic, Cubic, and Gaussian kernels for SVM, as well as Euclidean, Cosine, and Weighted metrics for KNN. Model performance was evaluated through training/testing accuracy, confusion matrices, prediction speed, and classification metrics such as precision, recall, and F1-score. The Fine Gaussian SVM achieved the highest performance with 91.7% training accuracy, 95.0% testing accuracy, 97.2% recall, and an F1-score of 91.10%, demonstrating its strong capability for reliable and cost-effective hotspot diagnosis in PV systems.

References

[1] M. Dhimish, “Defining the best-fit machine learning classifier to early diagnose photovoltaic solar cells hot-spots,” Case Studies in Thermal Engineering, p. 100980, 2021.

[2] Assalam, I. Gunadin, and S., “Application of Support Vector Machine (SVM) Method for Photovoltaic Condition Classification Based on Characteristic Curve Indicators,” Proc. 2023 Int. Conf. Artificial Intelligence Robotics, Signal and Image Processing (AIRoSIP), pp. 188–193, 2023.

[3] D. Winston et al., “Solar PV’s Micro Crack and Hotspots Detection Technique Using NN and SVM,” IEEE Access, vol. 9, pp. 127259–127269, 2021.

[4] M. Ali, H. Khan, M. Masud, K. Kallu, and A. Zafar, “A machine learning framework to identify the hotspot in photovoltaic module using infrared thermography,” Solar Energy, 2020.

[5] N. S., D. Sripada, S., and M. Aghaei, “Detection of visual faults in photovoltaic modules using a stacking ensemble approach,” Heliyon, vol. 10, 2024.

[6] A. Amiri, H. Oudira, A. Chouder, and S. Kichou, “Faults detection and diagnosis of PV systems based on machine learning approach using random forest classifier,” Energy Conversion and Management, 2024.

[7] F. Wang, Z. Zhen, B. Wang, and Z. Mi, “Comparative study on KNN and SVM based weather classification models for day-ahead short-term solar PV power forecasting,” Applied Sciences, vol. 8, p. 28, 2017.

[8] G. Divya, “Analyzing the performance of SVM and KNN in supervised learning for classification,” Int. J. for Research in Applied Science and Engineering Technology, 2025.

[9] H. Limam, A. Zouhair, and W. Oueslati, “A new hybrid multiclass approach based on KNN and SVM,” J. Inf. Knowl. Manag., vol. 21, no. 2250061, pp. 1–16, 2022.

[10] E. Boateng, J. Otoo, and D. Abaye, “Basic tenets of classification algorithms K-Nearest-Neighbor, Support Vector Machine, Random Forest and Neural Network: A review,” vol. 8, pp. 341–357, 2020.

[11] L. Demidova, “Two-stage hybrid data classifiers based on SVM and kNN algorithms,” Symmetry, vol. 13, p. 615, 2021.

[12] P. Noi and M. Kappas, “Comparison of Random Forest, k-Nearest Neighbor, and Support Vector Machine classifiers for land cover classification using Sentinel-2 imagery,” Sensors, vol. 18, 2017.

[13] M. Sha’abani, N. Fuad, N. Jamal, and M. Ismail, “kNN and SVM classification for EEG: A review,” pp. 555–565, 2020.

[14] G. Dlamini et al., “Meta-analytical comparison of SVM and KNN for text classification,” Proc. 2021 Int. Conf. Nonlinearity, Information and Robotics (NIR), pp. 1–6, 2021.

[15] Y. Jusman et al., “Comparison of dental caries level images classification performance using KNN and SVM methods,” Proc. 2021 IEEE Int. Conf. Signal and Image Processing Applications (ICSIPA), pp. 167–172, 2021.

[16] A. Roy et al., “Comparative analysis of KNN and SVM in multicriteria inventory classification using TOPSIS,” Int. J. Inf. Technol., vol. 15, pp. 3613–3622, 2023.

[17] P. Princess, S. Silas, and E. Rajsingh, “Classification of road accidents using SVM and KNN,” pp. 27–41, 2020, doi: 10.1007/978-981-15-3514-7_3.

[18] M. Kushalatha, H. Prasantha, and B. Shetty, “Comparison of image classification techniques and its applications,” Asian J. of Convergence in Technology, 2021.

[19] R. Pharswan and J. Singh, “Performance analysis of SVM and KNN in breast cancer classification: A survey,” pp. 133–140, 2020.

[20] Y. Lin and J. Wang, “Research on text classification based on SVM-KNN,” Proc. 2014 IEEE 5th Int. Conf. Software Engineering and Service Science, pp. 842–844, 2014.

[21] N. Azkona, A. Llaria, O. Curea, and F. Recart, “Detection, characterization and modeling of localized defects and thermal breakdown in photovoltaic panels from thermal images and I-V curves,” Electronic Materials, vol. 3, no. 2, pp. 154–172, 2022.

[22] G. Li, F. Wang, F. Feng, and B. Wei, “Hot spot detection of photovoltaic module based on distributed fiber Bragg grating sensor,” Sensors, vol. 22, no. 13, p. 4951, 2022.

[23] M. Santhakumari and N. Sagar, “A review of the environmental factors degrading the performance of silicon wafer-based photovoltaic modules: Failure detection methods and essential mitigation techniques,” Renew. Sustain. Energy Rev., vol. 110, pp. 83–100, 2019.

[24] IEC, IEC TS 63126:2020 ED1 Guidelines for qualifying PV modules, components and materials for operation at high temperatures, International Electrotechnical Commission, 2020.

[25] M. Bressan, Y. El Basri, A. G. Galeano, and C. Alonso, “A shadow fault detection method based on the standard error analysis of I-V curves,” Renew. Energy, vol. 99, pp. 1181–1190, 2016.

[26] W. Hermann, W. Wiesner, and W. Vaassen, “Hot spot investigations on PV modules—New concepts for a test standard and consequences for module design with respect to bypass diodes,” Proc. 26th IEEE Photovoltaic Specialists Conf., Anaheim, CA, USA, pp. 1123–1132, 1997.

[27] M. Köntges et al., “Performance and reliability of photovoltaic systems subtask 3.2: Review of failures of photovoltaic modules,” IEA PVPS Task 13 External Final Report, 2014.

[28] J.-H. Lin and Y.-K. Wu, “Simulation and fault diagnosis using current-voltage characteristics of photovoltaic systems—A case study,” Engineering Proceedings, vol. 92, no. 1, p. 79, 2025.

[29] S. Gulkowski and E. Krawczak, “Thin-film photovoltaic modules characterisation based on I-V measurements under outdoor conditions,” Energies, vol. 17, no. 23, p. 5853, 2024.

[30] IEC, IEC TR 61853-4:2021 Performance testing and energy rating of photovoltaic modules – Part 4: Standard reference climatic profiles, 2021.

[31] A. Ul-Haq, H. Sindi, S. Gul, and M. Jalal, “Modeling and fault categorization in thin-film and crystalline PV arrays through multilayer neural network algorithm,” IEEE Access, vol. 8, pp. 102235–102255, 2020.

[32] K. Zhan, Z. Wang, and W. Shen, “Deep learning-based automatic detection of multitype defects in photovoltaic modules and application in real production line,” Prog. Photovolt. Res. Appl., vol. 29, pp. 471–484, 2021.

[33] N. Mekhaznia and R. Khenfer, “Diagnosis of PV module based on neural network using performance indices,” Int. J. Power Electron. Drive Syst. (IJPEDS), 2023.

[34] R. Mandal, N. Anand, N. Sahu, and P. Kale, “PV system fault classification using SVM accelerated by dimension reduction using PCA,” Proc. 2020 IEEE 9th Power India Int. Conf. (PIICON), pp. 1–6, 2020.

[35] C. Sowthily, S. Kumar, and M. Brindha, “Detection and classification of faults in photovoltaic system using random forest algorithm,” pp. 765–773, 2020.

[36] R. Eltuhamy et al., “Fault detection and classification of CIGS thin-film PV modules using an adaptive neuro-fuzzy inference scheme,” Sensors, vol. 23, 2023.