Mitigating greenhouse-gas emissions from the transportation sector remains a key requirement for global climate action, and electric vehicles (EVs) are being deployed at scale as a primary pathway. Widespread EV adoption, however, increases electricit...
Mitigating greenhouse-gas emissions from the transportation sector remains a key requirement for global climate action, and electric vehicles (EVs) are being deployed at scale as a primary pathway. Widespread EV adoption, however, increases electricity demand and heightens reliance on charging infrastructure, which can constrain operational flexibility. Solar electric vehicles (SEVs) integrate photovoltaic modules into the vehicle body to supply electricity during driving and parking, providing an additional onboard energy source when charging access is constrained. In real-world operation, SEV generation varies with solar conditions and spatiotemporal shading from roadside buildings and trees. Realizing SEV benefits therefore requires operational optimization that jointly considers on-road solar generation and charging decisions. This study develops an integrated navigation-based framework that quantitatively assesses SEV solar power generation potential and incorporates it into driving and charging decisions.
To capture spatiotemporal shading induced by roadside environments, a Google Street View (GSV)–based shadow database was constructed and coupled with an SEV-specific power generation model. Deep-learning segmentation was applied to classify the sky, trees, and artificial structures in GSV images. For trees, monthly transmittance coefficients were derived using image-processing techniques to represent partial shading effects from seasonal canopy changes. The mean error of the estimated tree transmittance relative to measurements was 6.4%, and the resulting generation model achieved R² = 0.73 against measured SEV power. Compared with conventional 3D model–based shading approaches, the proposed GSV-based analysis reduces memory use and computational cost while preserving road-level spatiotemporal resolution. This enables efficient estimation of route-dependent generation profiles and facilitates integration into navigation frameworks over large spatial domains.
Using the proposed model, incorporating tree transmittance increased the annual mean estimated SEV solar power generation by 3.2% relative to the no-transmittance assumption, with increases up to 5.8% in winter and 20% on mountainous segments. These results indicate that treating trees as fully opaque can lead to underestimation of SEV solar power generation, and that seasonal variability can be represented more realistically by accounting for partial transmittance. Building-induced shading was also quantified: annual SEV solar power generation decreased by approximately 20% near high-rise buildings compared with low-rise surroundings, and road segments with buildings concentrated on the southern side exhibited 10% lower SEV solar power generation than other orientations. These findings show that weather-only SEV solar power generation estimates are insufficient to capture urban-canyon effects, and that reliable assessment requires explicit consideration of buildings and roadside trees.
An energy-efficient route navigation method was developed to minimize net energy consumption by incorporating the spatiotemporal variability of SEV solar power generation during driving. Under an assumed 2030 fleet size of 107,380 SEVs, selecting the energy-efficient route reduced annual electricity use by 195 MWh and avoided 92 tCO₂eq relative to the shortest route. This demonstrates that embedding segment-level generation variability into route choice can improve both energy efficiency and emissions in dense urban environments.
For long-distance operations, external charging remains necessary; therefore, an SEV-specific driving–charging scheduling navigator was formulated to optimize charging time, location, and amount while accounting for periods and road segments with higher SEV solar power generation potential. The scheduling model was implemented using a hybrid approach combining reinforcement learning and evolutionary strategies. At the 2030 adoption scale (107,380 vehicles), SEV-specific optimization reduced annual external charging electricity by an additional 541 MWh and avoided an additional 257 tCO₂eq compared with the non-optimized SEV baseline. Further benefits were obtained by coupling the SEV-specific schedule with parking-based self-charging: relative to SEV-specific optimization alone, annual external charging electricity decreased by an additional 88 GWh and emissions were reduced by an additional 41.72 ktCO₂eq. These results indicate that adding solar self-charging during parking substantially lowers external electricity dependence and increases fleet-scale emission reductions.
This study quantifies road-environment–driven spatiotemporal shading by integrating GSV-based analysis with tree transmittance modeling, and integrates the resulting high-resolution SEV solar power generation estimates into an optimization-based navigation framework. By jointly accounting for SEV solar power generation variability and battery state-of-charge constraints, the proposed driving and charging navigation improves solar energy utilization and reduces reliance on external electricity. Future work should incorporate real-time traffic and driving patterns, together with forecast-based meteorological information, to provide route-level generation forecasts and advance the framework toward practical deployment.