Integrating a Heterogeneous Driver Choice Model into an Equilibrium Traffic Assignment Model for Urban Networks with Fast Charging Stations

The relevance of this study stems from the need for accurate demand forecasting for electric vehicle fast-charging infrastructure, a critical task for urban transport planning. Existing modeling approaches often rely on simplified cost functions, ignoring key psychological factors and significant heterogeneity in driver preferences, which leads to inaccurate results. The goal of this paper is to develop and substantiate a comprehensive methodology for integrating a detailed heterogeneous driver choice model into a computationally efficient equilibrium model of an urban transport network. The research methodology is based on the synthesis of two theoretical components: a modified classical Frank- Wolfe assignment model, adapted for networks with charging infrastructure, and a discrete choice behavioral model using latent classes. The proposed methodology involves a sequential multi-stage transformation procedure. It begins with the specification and estimation of the behavioral model from survey data, followed by behavioral filtering to identify an «active group» of drivers potentially willing to charge, and culminates in the construction and adaptation of behaviorally consistent cost functions for each user class. The main results of the study consist in the creation of a complete algorithm and toolkit that transforms probabilistic estimates of individual preferences into deterministic macromodel parameters. This allows network equilibrium models to account for factors such as range anxiety, sensitivity to queuing time, and the attractiveness of charging station attributes. The scientific novelty lies in the development of principles for specifying class-specific cost functions, which reduces a complex behavioral problem to a multi-class version of the Frank-Wolfe algorithm while preserving key information about the heterogeneity of driver preferences. The practical significance is that the proposed approach provides transport planners with a tool for the direct calibration of cost functions using empirical survey data, eliminating the need for complex heuristic parameter tuning. This opens up opportunities for more accurate scenario analysis and optimization of charging infrastructure development. Directions for further research include adapting the proposed method for stochastic equilibrium models and its verification using real-world traffic flow data.