Work

Emission prediction in hybrid electric vehicles using football optimization and reinforcement learning

Year: 2025

Type: article

Abstract: Hybrid electric vehicles (HEVs) are considered one of the relevant sustainable transportation solutions to reduce carbon emissions and improve energy efficiency. Therefore, the accurate prediction of emissions in HEVs is considered a significant challenge. To address this challenge, this paper proposes a novel methodology that combines Football Optimization Algorithm (FbOA) with Reinforcement Learning for Time Series Prediction (RLTSP) to improve emission prediction. In specific, the FbOA is employed for selecting the subset of features that significantly impact emissions; whereas, the RLTSP model is used for time series prediction to dynamically adapt and improve the prediction accuracy over time. The proposed FbOA + RLTSP model demonstrated a promising prediction accuracy (MSE = 1.13E-05, RMSE = 1.02E-05, MAE = 1.14E-05, r = 0.977, R2 = 0.983) and revealed significant connections between HEV features and emissions. The strong negative correlation between maximum range and emissions (r = $$-$$ 0.85) highlighted the need to extend electric-only driving ranges, while the moderate positive correlation between vehicle weight and emissions (r = 0.47) highlighted the need for lightweight design to reduce environmental impact. Battery capacity and fuel cell efficiency had fewer relationships, suggesting indirect or context-dependent effects on emissions. Researchers, politicians, and industry stakeholders can use these findings to improve HEV design and performance for sustainability.

Source: Discover Computing

DOI: 10.1007/s10791-025-09875-w