Comprehensive assessment of plug-in hybrid electric vehicle emissions in real-world driving: Insights from multi-mode...

Plug-in Hybrid Electric Vehicles (PHEVs) are widely regarded as a pivotal strategy for mitigating transportation emissions; however, a comprehensive understanding of their real-world emission characteristics across various operating modes, particularly the charge-increasing (CI) mode, remains insufficient. This research employed a Portable Emissions Measurement System (PEMS) to investigate the actual emission characteristics of a China VI-b compliant PHEV across urban, suburban, and highway conditions. Furthermore, multiple machine learning models were developed based on On-Board Diagnostics (OBD) data to predict pollutant emissions. Findings reveal that during cold-start, CO and HC emissions peaked in Charge-Sustaining (CS) mode, followed by CI mode, with Charge-Depleting (CD) mode exhibiting the lowest levels. Total NO_x emissions, however, remained consistent across all three modes. Vehicle Specific Power (VSP) demonstrated a strong positive correlation with HC, NO_x, and CO₂ emissions in all propulsion modes. However, CO and Black Carbon (BC) emissions exhibited the highest sensitivity to VSP changes in CI mode, whereas this correlation was significantly weaker in CS and CD modes. Further analysis revealed that emission disparities among the three propulsion modes were most significant under urban conditions. Notably, NO_x emissions in CI mode consistently surpassed China VI-b standards across all road types. Among the models evaluated, Random Forest (RF) and LightGBM models exhibited superior performance, with fuel consumption identified as the most critical factor influencing pollutant emissions.