Abstract
The accurate fault detection and monitoring system is necessary and essential to maintain the optimal performance of the photovoltaic (PV) system, the literature on this topic has identified most of the case detection methods that focused only on the system (PV). The gap found is the main problem of the most previous studies and research and duplication of research in the detection of defects (structure and system). This paper presents the results of electromechanical impedance (EMI) analysis performed for a 36-cell silicon photovoltaic module with bonded piezoelectric transducer (PZT). Utilizing ANSYS software, a comprehensive three-dimensional coupled field finite element model is constructed, facilitating harmonic analysis in the context of high-frequency impedance analysis procedures. However, the conventional examination of electromagnetic impedances proves inadequate in offering a comprehensive understanding of damage localization. This study offers valuable perspectives on the formulation and application of a concise electromagnetic (E/M) impedance technique designed for high-frequency impedance analysis. Moreover, it underscores the inherent limitations associated with relying solely on electromagnetic impedances for accurate damage localization. This study provides useful insights into the development and implementation of the compact E/M impedance method for high-frequency impedance analysis and highlights the limitations of using EMIs for damage localization. The model's training process is undertaken using simulation data to encompass a broad spectrum of scenarios. Subsequently, its validation is executed on a photovoltaic (PV) panel model to establish its viability and demonstrate the model's practical applicability. The extreme learning machine (ELM) based algorithm has been devised to predict damage locations using data from piezoelectric sensors. The trained model demonstrated heightened predictive accuracy in comparison to alternative classification methodologies within the realm of machine learning. The outcomes underscore the superior efficacy of the ELM algorithm in defect detection, boasting an impressive overall accuracy rate of 85%.