Dynamic solar irradiance estimation for vehicle thermal management using a multi-modal machine learning framework

Accurate solar irradiance estimation for moving vehicles is essential for effective thermal management and energy optimization, particularly in electric vehicles (EVs) where cabin climate control directly impacts battery usage and driving range. However, existing approaches often overlook the dynamic environmental conditions encountered during vehicle movement, limiting their applicability in real-world scenarios. This study proposes a novel machine learning framework that integrates satellite remote sensing and sky camera imagery to estimate solar irradiance around moving vehicles in real time. The framework was validated using comprehensive field data collected from an instrumented vehicle equipped with solar irradiance sensors, GPS tracking, and environmental monitoring systems across various routes and seasonal conditions. By fusing multi-source environmental data, the model delivers high spatiotemporal resolution predictions of solar irradiance under diverse urban and atmospheric conditions. Results show strong performance, with normalized root-mean-square errors of 14.61% in summer and 17.10% in winter. These accurate predictions enable adaptive, location-aware thermal management strategies that proactively adjust climate control based on anticipated solar exposure. The integration of satellite and ground-based data sources ensures scalability across regions and vehicle types, establishing a foundation for intelligent thermal management systems that enhance energy efficiency and passenger comfort in both conventional and electric vehicles.