Study on Data Preprocessing Models for Development of Photovoltaic Module Fault Diagnosis Algorithm

With the global push for climate change mitigation and sustainable energy transition, photovoltaic (PV) systems have emerged as a leading renewable energy source.As of 2023, global PV installations reached a record 444 GW, with continued growth expected in both international and domestic markets.However, the widespread adoption of PV systems has highlighted the importance of efficient maintenance and fault diagnosis.Performance degradation and failures in PV modules can result from various factors, including environmental conditions, manufacturing defects, and operational stress, leading to reduced energy output and economic losses.To ensure accurate fault diagnosis, reliable preprocessing of time-series data, such as current, voltage, and power output, is essential.This study presents a preprocessing model that combines outlier removal and missing value imputation using methods such as interquartile range, standard deviation, K-nearest neighbor, and mean-based approaches.The effectiveness of each preprocessing combination was evaluated based on data reliability (accuracy), which was visualized through graphs and metrics.The proposed pipeline provides a robust foundation for developing intelligent fault diagnosis algorithms and enhances data integrity for further analysis.