Prediction and validation of solar power generation for solar electric vehicles using shading statistics on urban roads

This study aims to analyze shading patterns on urban road segments using Google Street View panoramic images and to predict solar energy generation during vehicle operation through Monte Carlo simulations. Urban roads were divided into regular intervals, and Google Street View panoramic images were collected at each point. The study area was Nam-gu, Busan, where 3504 panoramic images were collected at 50-m intervals. These images were converted into hemispherical images. A deep learning model was applied to classify the sky and shading obstacles within these hemispherical images, and the annual solar trajectory was overlaid to calculate solar irradiance reduction rates by season and time of day. The road segments were categorized into seven types of shading patterns, and the hemispherical images were aggregated to analyze the statistical reductions in solar irradiance. Monte Carlo simulations were employed to derive multiple reduction rates for each shading type and predict the maximum, median, and minimum solar energy generation values. Additionally, three routes were subdivided into 10-m segments based on shading patterns, and solar energy generation estimates were made for each segment. Validation through real-world experiments demonstrated that the predicted energy generation closely matched the actual measurements across all routes.