Solar Irradiance Ramp Classification Using the IBEDI (Irradiance-Based Extreme Day Identification) Method

The inherent variability of solar energy presents a significant challenge for grid operators, particularly when it comes to maintaining stability. Studying ramping phenomena is therefore crucial to understanding and managing fluctuations in power supply. In line with this goal, this study proposes a new classification approach for solar irradiance ramps, categorizing them into four distinct classes. We have proposed a methodology including adaptation and extension of a wind ramp classification to solar ramp classification titled the Irradiance-Based Extreme Day Identification method. Our proposal includes an agglomerative algorithm to find new ramp class boundaries. The strength of the proposed method relies on that it allows its generalization to any dataset. We assessed it on three datasets from distinct geographic regions—Oregon (northwestern United States), Hawaii (central Pacific Ocean), and Portugal (southwestern Europe)—each with varying temporal resolutions of five seconds, ten seconds, and one minute. The class boundaries for each dataset results in different limits of Z score value, as a consequence of the different climatic characteristics of each location and the time resolution of the datasets. The “low” class includes values less than 0.62 for Portugal, less than 2.17 for Oregon, and less than 2.19 for Hawaii. The “moderate” class spans values from 0.62 to 3.51 for Portugal, from 2.17 to 5.01 for Oregon, and from 2.19 to 5.88 for Hawaii. The “high” class covers values greater than 3.51 and up to 6 for Portugal, greater than 5.01 and up to 10.72 for Oregon, and greater than 5.88 and up to 8.01 for Hawaii. Lastly, the “severe” class includes values greater than 6 for Portugal, greater than 10.72 for Oregon, and greater than 8.01 for Hawaii. Under cloudy sky conditions, it is observed that the proposed algorithm is able to classify the four classes. These thresholds show how the proposed methodology adapts to the unique characteristics of each regional dataset.