Introduction: With the rapid popularization of electric vehicles, their charging load influences the stable operation of the power grid. An accurate prediction of EV charging station load is crucial for optimal resource allocation in power systems. The objective of this study is to address the issue of insufficient accuracy in existing prediction methods, this paper proposes a hybrid prediction model based on Bidirectional Long Short-Term Memory and Adaptive Boosting, aiming to improve the accuracy and stability of medium and long-term EV charging station load forecasting. Methods: The study employs a three-step approach: (1) The pearson correlation analysis was utilized to evaluate multi-dimensional influencing factors and reduce dataset dimensionality; (2) implementation of a BiLSTM neural network for temporal feature extraction and preliminary prediction; and (3) application of the Adaboost algorithm to construct a weighted combination of strong classifiers. The model’s effectiveness was validated through comprehensive simulation tests using real-world charging station data. Results: The proposed Pearson feature selection-based BiLSTM-Adaboost model outperforms traditional benchmark models (LSTM and SVM), effectively reduces data redundancy through feature selection, achieves better performance in key indicators (MSE, RMSE, and MAPE), and demonstrates strong generalization ability and robustness while maintaining high accuracy. Discussion: Experimental results demonstrate that the proposed method effectively extracts key features of charging loads, achieving superior prediction accuracy and generalization ability compared to traditional methods. This provides a reliable decision-making tool for power grid operation, effectively supporting the resilience planning needs of urban power grids under continuously increasing EV penetration rates. But further research is needed to address robustness under extreme weather conditions. Conclusion: This study provides an effective load forecasting methodology for power systems to address the challenges of large-scale electric vehicle integration. Future research will explore more robust feature engineering methods and deep learning architectures, such as combining other more advanced time series prediction models and improving optimization algorithms to enhance model adaptability and generalization capability for complex data patterns.

The rapid growth of vehicle ownership has led to increasingly congested road networks, which significantly reduces the energy efficiency of electric vehicles (EVs) and intensifies user range anxiety. To address these challenges, a dynamic EV route planning process is proposed by integrating traffic flow (TF) prediction, charging service modelling, and time-varying path optimization within a unified framework. First, future TF is predicted using a data-driven forecasting module based on the iTransformer model, which captures multivariate temporal dependencies across road links and provides accurate inputs for downstream decision-making. Based on the predicted traffic states, a time-dependent queuing process is formulated to estimate charging station waiting times by modelling the dynamic interaction between vehicle arrivals and service capacity. These components are then embedded into a time-varying shortest path optimization process that explicitly considers mid-journey charging constraints, with the objective of minimizing total travel time and economic cost. The proposed framework establishes a closed-loop decision-making process that couples traffic evolution, charging service dynamics, and routing behaviour. Extensive comparative experiments against classical Time-Dependent Shortest Path (TDSP) methods under different network scales, together with a real-world case study, demonstrate that the proposed approach achieves higher computational efficiency and improved routing performance under dynamic conditions. The results indicate that the proposed process-oriented method provides an effective and practical solution for EV routing in intelligent transportation systems characterized by time-varying traffic and service processes.

The increasing electrification of urban transportation has formulated a tightly coupled energy-mobility nexus. Under extreme disaster events or grid faults, rapidly restoring power supply capacity and re-dispatching shared electric vehicle (EV) fleets are critical for enhancing system resilience. Existing co-optimization methods face the curse of dimensionality when dealing with high-dimensional discrete grid reconfigurations and continuous spatio-temporal EV queuing dynamics. While multi-agent deep reinforcement learning (MADRL) offers real-time responsiveness, it inherently struggles to satisfy strict physical constraints, frequently generating infeasible and unsafe actions. To bridge this gap, this paper proposes a heuristic-guided safe multi-agent reinforcement learning (Safe-MADRL) framework for the resilient dispatch of the energy-mobility nexus. Instead of relying solely on black-box neural networks, the framework structurally embeds physical models and heuristic solvers into the learning loop. A quantum particle swarm optimization (QPSO) algorithm acts as a heuristic action refiner to ensure that grid topology actions strictly comply with non-linear power flow and voltage constraints. Simultaneously, a mixed-integer linear programming (MILP) model coupled with a single-queue multi-server (SQMS) model serves as a safety projection layer. This layer mathematically guarantees EV battery energy continuity and accurately quantifies spatio-temporal queuing delays at charging stations. Case studies on a coupled IEEE 33-node distribution system and a regional transportation network demonstrate that the proposed Safe-MADRL framework achieves zero physical violations during training and significantly outperforms traditional mathematical optimization and pure learning-based methods in computational efficiency, system power loss reduction, and overall operational economy.