A method for delineating traffic low emission control zone based on deep learning and multi-objective optimization

Abstract

Current methods for defining traffic low emission control zones (TLEZ) often face limitations that hinder their widespread implementation and effectiveness. This study addresses these challenges by employing a comprehensive approach to analyze PM_2.5 concentration levels within TLEZ. This study utilizes PM_2.5 data collected by taxi fleets, integrating static road network features and dynamic time series features to gain a detailed understanding of pollution distribution patterns across different urban areas. To capture these complex distribution patterns of PM_2.5, a sophisticated deep learning model that combines Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) networks, and Attention Mechanism (AM) is deployed. This model adeptly identifies spatial and temporal variations in PM_2.5 concentrations, allowing for a more accurate and responsive analysis of pollution levels. A multi-objective optimization model is developed to minimize the overall impact on residents' daily lives, which considers both environmental and social factors in the delineation of TLEZ. The optimization model is solved using the Non-dominated Sorting Genetic Algorithm II (NSGA-II), which is a robust evolutionary algorithm that facilitates the identification of Pareto-optimal solutions. These solutions can help define the optimal boundaries for Low, Ultra-Low, and Zero Emission Zones. By establishing a framework for assessing and optimizing these zones, this study provides valuable insights and actionable guidance for policymakers and urban planners.