Abstract
Wind power generation is expected to greatly contribute to the future of humanity as a promising source of renewable energy. However, the high variability inherent in wind is a challenge that hinders stable power generation. To utilize wind power as a primary energy source, integration with a polymer electrolyte membrane water electrolysis (PEMWE) system is proposed. Yet, PEMWE is known to suffer from degradation when exposed to input power patterns with high variability. This poses challenges to its commercialization. This necessitates stress testing with various wind power fluctuations during the production process of the devices. This study investigates representative patterns of wind power fluctuation so that these patterns can be used for the stress testing process. We employ data-mining techniques, including the Swing Door Algorithm and k-means clustering, to identify these patterns by analyzing wind power generation data at a 10-second interval. As a result, the five most representative wind power ramps are presented. This study provides practical guidelines for the development process of expensive devices for wind power generation, thereby promoting the active utilization of wind power generation.