Implementation of an IEC 61215-oriented photovoltaic module test emulator with integrated predictive maintenance capabilities
Open Access
- 30 September 2026
- journal article
- Published by Politechnika Lubelska in Applied Computer Science
- Vol. 22 (3) , 196-218
Abstract
This paper presents a hybrid software-based framework for photovoltaic (PV) testing, monitoring, and predictive maintenance that combines a physically validated PV emulator with data-driven diagnostic models. The emulator reproduces selected IEC 61215-inspired test scenarios, including electrical characterisation, irradiance/temperature variations, simulated accelerated ageing, and electrical fault scenarios—without relying on real solar irradiation, enabling realistic I–V and P–V curve generation under controlled settings for performance assessment and algorithm validation. Note that the emulator is a complementary validation tool and does not substitute for formal IEC 61215 certification. Building on this emulator, a two-stage intelligent monitoring architecture is proposed. First, an unsupervised autoencoder learns normal PV behaviour and detects anomalies by analysing reconstruction errors. It shows strong discriminative capability, achieving a high ROC–AUC while remaining robust across varying environmental conditions, thereby enabling early detection of abrupt faults and progressive degradation. Second, a supervised bidirectional gated recurrent unit (BiGRU) model performs fault classification by explicitly exploiting temporal dependencies in electrical signals. The BiGRU classifier achieves an overall accuracy exceeding 99%, with high class-wise precision and recall across multiple fault categories, including resistive faults, degradation, soiling, and severe electrical failures. The framework is validated on emulator-based and analytically derived datasets (Sandia, Madison, Cenerg models), demonstrating strong generalisation and stability across operating conditions; field validation on real PV installations is identified as a key next step. Overall, integrating physical emulation, unsupervised anomaly detection, and supervised temporal classification provides a robust and scalable predictive maintenance solution for PV installations, particularly in environments with high climatic variability.Keywords
This publication has 33 references indexed in Scilit:
- Policy-driven expansion of renewable energy in Cameroon: A technical and sustainability-centered analysis of growth trends and cross-sectoral impacts (2015–2024)Energy Strategy Reviews, 2025
- A multi-stage review framework for AI-driven predictive maintenance and fault diagnosis in photovoltaic systemsApplied Energy, 2025
- Data-driven Detection of Hot Spots in Photovoltaic Modules Using Information FusionInternational Journal of Control, Automation, and Systems, 2025
- Comprehensive study on photovoltaic cell's generation and factors affecting its performance: A ReviewMaterials for Renewable and Sustainable Energy, 2025
- Strategic Global Deployment of Photovoltaic Technology: Balancing Economic Capacity and Decarbonization PotentialAdvances in Atmospheric Sciences, 2024
- Assessment of concentrated solar power (CSP) generation potential in Cameroon using a Multi-Criteria Analysis Method and Geographic Information System (GIS)Heliyon, 2024
- Advanced hardware-in-the-loop testing chain for investigating interactions between smart grid components during transientsElectric Power Systems Research, 2023
- Fault diagnosis of photovoltaic systems using artificial intelligence: A bibliometric approachHeliyon, 2023
- Photovoltaic system fault detection techniques: a reviewNeural Computing and Applications, 2023
- A real-time co-simulation of PV power generation system using transmission line model interfaceEnergy Reports, 2022