Varying shading in vehicle-integrated PV for different public transport case studies

Abstract

With advancements in the photovoltaics (PV) market, involving increased PV module efficiency and reduced costs, the logical progression is the integration of PV into various surfaces, including vehicles (VIPV). For a driving VIPV, the constant change in irradiance presents a significant challenge. The location, weather, and time are key factors that affect the irradiance prediction. An accurate prediction of solar energy generation is crucial for realizing the full potential of VIPV. In this study, the incident irradiance loss is referred to as shading loss.

This thesis is conducted in collaboration with Delft University of Technology (TU Delft) and the Netherlands Organisation for Applied Scientific Research (TNO). TNO has an existing shading loss model that accounts for the surrounding obstructions and weather effects using a fixed value, where the only variable input is the day of the year. On average, it assumes 27% shading losses yearly. This study has improved TNO’s VIPV shading assessment by incorporating more variables into a new model, thereby contributing to a better estimation of the VIPV energy yield. The inputs of the new model are based on key factors causing irradiance loss, where location data is obtained using a Digital Surface Model (DSM). The new contribution to the DSM-based approach is the Web Coverage Service (WCS), which retrieves height data directly through a URL, eliminating the need for the previous manual operations and building upon the model’s scalability. The improved model determines shading losses along a route using two varying shading indicators: Sky View Factor (SVF) and Shading Factor (SF). It can input various irradiance datasets, including or excluding weather effects, and precisely determines the Sun’s position based on the time and day of the year. The DSM-based approach is validated using a dynamic and static validation dataset, with a SVF≤1. The accurate representation of surrounding obstructions during validation contributes to a good-quality model. Both validation studies differ in inputs based on location, time, and weather conditions. The differences in measured and simulated irradiance values are larger in the dynamic dataset than in the static dataset. Deviations are caused by uncertainties and errors, such as local weather effects, sensor inaccuracies, traffic shading, and LiDAR discrepancies, which is the inability to simulate open or changing structures. Weather variability is the primary factor affecting model quality in this validation study, resulting in no fixed inaccuracy factor for model validity. The validated model can accurately represent the environment, but weather variability should be minimized when comparing measured and simulated irradiance. Therefore, averaging is chosen to compare results in this work as it removes unknown or untraceable data points.

Beyond previous research studies, this work focuses on two public transport case studies, given the systematic routing with a fixed timetable. The first case study includes a car and a bus, and the second, more extensive case study focuses solely on buses. The choice of inputs in both case studies centers on key factors influencing VIPV irradiance loss. In the Bus 40 study, simulations under clear sky conditions and a fixed time around solar noon resulted in shading losses of 22.7% for the bus and 27.7% for the car. Moreover, the simulated shading loss shows a 2.6% relative difference compared to the yearly average TNO value. For the case study considering 27 bus routes in Amsterdam, simulations are performed using averaged KNMI irradiance and all hours a day, where the annual average shading loss resulted in 24.8%. The SF and clear sky irradiance correlate in the first case study and the SVF and average KNMI irradiance correlate in the second case study. However, using clear sky conditions and uniform overcast conditions does not represent realistic weather conditions and leads to a bias in the correlation. Moreover, classified terrain types do not correlate with SVF, in contrast to the reversed approach by Araki et al.

Finally, the model can effectively characterize the shading losses on the route level. However, this methodology falls short when reaching finer resolution, such as when examining smaller route segments or partial shading. The determination of shading losses at the route level can be of further benefit to various stakeholders, such as bus operators who want to determine the optimal bus routes for maximum solar energy generation. In addition, this model can be helpful for policymakers, to predict the required charging infrastructure in a scenario when vehicles are equipped with solar panels. Moreover, this model improves shading loss estimation by accepting a range of datasets instead of TNO’s current approach. By inputting irradiance data for clear skies to complete overcast skies, the shading losses can be predicted in a safer range instead of over- or underestimation based on one shading value.