Work

Interpretable Station-Level Charging Congestion Pressure Assessment and Multi-Horizon Early Warning for Electric-Vehicle Charging Infrastructure

Year: 2026

Type: article

Abstract: The rapid growth of electric-vehicle charging demand has increased the need for reliable station-level congestion monitoring and early warning. Existing studies mainly predict charging demand, load, occupancy, or availability, whereas charging congestion pressure is usually shaped by multiple operational factors. This study proposes an interpretable station-level charging congestion pressure assessment and multi-horizon early-warning framework. A Charging Congestion Pressure Index (CCPI) is constructed by integrating occupancy, arrival pressure, charging or occupation duration, service volume, and price–time context into a unified station–hour pressure representation. Based on temporally aligned current, lagged, and rolling features, future high-pressure states are predicted at 1 h, 3 h, and 6 h horizons. Using 1423 charging stations and 6,181,512 station–hour observations from September 2022 to February 2023, this study evaluates whether the proposed station–hour pressure representation can support multi-horizon high-pressure warning under temporal and station-level validation settings. Results show that current pressure is a strong short-term persistence baseline, while learning-based models provide larger F1-score gains at longer horizons. Extreme Gradient Boosting (XGBoost) achieved F1 gains of +0.022, +0.040, and +0.068 over the persistence baseline at the 1 h, 3 h, and 6 h horizons, respectively. Ablation, temporal validation, station holdout validation, and block-bootstrap tests further support the stability of the proposed framework. These findings indicate that interpretable pressure-index construction and temporally consistent multi-horizon warning can provide an engineering decision-support basis for charging-infrastructure operation, station-level congestion monitoring, and proactive resource management.

Source: World Electric Vehicle Journal

DOI: 10.3390/wevj17090443

Author: Kai Shi

Institution: Wuhan University of Technology

Language: English

Topic: Electric Vehicles and Infrastructure

Subfield: Electrical and Electronic Engineering

Field: Engineering

Domain: Physical Sciences

SDGs: Industry, innovation and infrastructure, Sustainable cities and communities

Open Access status: gold

Estimated APC paid: $1,689