The integration of distributed energy resources via T-taps introduces strong source-load fluctuations that often cause maloperations in conventional two-terminal current differential protection. To address this issue, this paper proposes an adaptive current differential protection scheme utilizing inverse amplitude cross-restraint and phase angle adaptation. An inverse cross-restraint mechanism based on two-terminal current amplitude ratios is introduced to prevent maloperations from internal T-tap fluctuations. Meanwhile, a phase-angle-evolution algorithm rapidly reduces restraint quantities during internal faults, and phase-to-phase differential characteristics are incorporated to further enhance sensitivity to single-phase short-circuits. Simulations confirm reliable operation under heavy T-tap loading, inverter-based resource integrations with diverse control strategies, and single-phase faults at the hundred-ohm level. The proposed method requires no additional voltage-measurement hardware effectively mitigates unmeasured T-tap interference and improves protection sensitivity in active distribution networks.
Railway-based express transportation has become an important component of modern logistics systems, offering an alternative to express deliveries. To enhance coordination between road and rail transportation, this study investigates a strategic service network design and freight flow allocation problem for integrated road-rail express freight, in which shipments can be transported either entirely by road or through a combination of road transport and existing passenger trains equipped with attached parcel carriages. An integrated road-rail express service network is constructed to support flexible mode choice, road-rail transfers, and cross-line transshipment within the railway service network. Based on this network, an integer linear programming model is established to jointly optimize the design of the railway express service network and freight flow allocation, to minimize total system operating cost. To address the computational challenges in large-scale instances, a hybrid solution framework combining adaptive large neighborhood search (ALNS) with a commercial solver is developed. The effectiveness of the methodology is evaluated through numerical experiments of different scales and a case study based on a realistic service-network setting. The results indicate that the integrated road-rail express service network achieves a lower total system cost than a road-only transportation alternative and better accommodates degraded road operating conditions. In addition, the optimization model tends to select long-distance train services for express freight operations.
Fairness in same-day delivery (SDD) becomes increasingly important as customers demand equitable service pledge, yet cost-driven merchants’ preferences for nearby customers create service disparities across regions. Drone fleet resupply presents a fast and traffic-free solution capable of navigating complex terrains, while truck-based delivery complements it by overcoming the drone’s limitations in endurance and operational range. Previous studies mainly address single-drone, single-truck problems, with limited attention to the potential of drone fleet resupply with multi-truck systems to solve fairness issues. This research fills this gap by originally proposing a drone fleet resupply system for stochastic requests, where drones perform multiple trips to replenish any truck as needed, with consideration of drone endurance limitations. This problem is formulated as a sequential decision process (SDP) and solved through a deep Q-learning (DQL) approach to maximize service rates while ensuring fairness. A piecewise linear reward function is designed to improve convergence. Comparative results against benchmark policies demonstrate DQL’s potential to manage highly stochastic logistics operations. Key findings include: (1) Drone fleet resupply enhances regional fairness and overall service rates, demonstrating strong robustness in serving long-distance and time-sensitive deliveries; (2) An appropriate reward function significantly facilitates fairness, while penalties further improve it but are ineffective alone; (3) An optimal drone-truck fleet mix exists to balance service rates and associated costs, while adding more trucks alone substantially improves service rates but incurs larger costs; (4) The proposed approach remains effective across drone capacities, while extended drone endurance yields limited gains in service rate but significantly lowers delivery costs.
For the traction power supply systems based on PWM converters (PWM-TPSS), to resolve the dual dilemmas of a well-defined model lacking a viable solution method and a Procrustean approach in the optimal power flow (OPF) problem considering piecewise nonlinear characteristics, an equivalent reconstruction OPF model is introduced. Based on the basic OPF model considering piecewise nonlinear characteristics such as train braking behavior, auxiliary variables including the main transformer incoming power and the train voltage contribution factors are introduced to perform equivalent transformations on the objective function, AC-DC coupling relationship, and train power equation. This improves the partial differentiability of the OPF model, and the interior point method is adopted for solution. Case study shows that the reconstructed model is equivalent to the basic model; compared with the genetic algorithm and the grey wolf-whale hybrid optimization algorithm, the reconstructed model improves the solution speed by 235 times and 574 times respectively while ensuring optimization accuracy. Comparative results indicate that ignoring the train braking characteristics will lead to an optimization error of 2.11%, demonstrating the necessity of considering it. This paper provides the model support for the efficient coordinated operation of PWM-TPSS.
Rumor detection models must remain robust to breaking events that are unseen during training. Although large language models (LLMs) offer strong generalization potential, their text-centric nature makes them vulnerable to spurious textual patterns in noisy social media discussions and leaves them without explicit awareness of how rumors propagate through conversation threads. To address these limitations, we propose RumorGAL, a graph-anchored LLM framework for domain-generalized rumor detection. The framework first employs a graph neural network (GNN) to derive graph-anchored structural prefixes from the propagation graph, providing structural guidance for the LLM. It then identifies high-quality textual evidence from noisy discussions through an evidence density-guided mining strategy. Finally, a hierarchical reasoning mechanism organizes both the structural prefixes and textual evidence into a unified reasoning process, enabling the LLM to ground its predictions in both propagation patterns and informative content. Across the three transfer settings, RumorGAL achieves robust performance, improving average accuracy over the strongest baseline, zero-shot GPT-4.1, by 5.27 percentage points under the domain generalization setting. When compared with domain adaptation methods, it surpasses the best domain adaptation method in two out of three tasks, with improvements of 3.1 and 0.5 percentage points in F1 score for the rumor class, respectively. These findings suggest that an LLM-as-primary, GNN-as-auxiliary design, which couples structure-aware guidance with evidence-grounded reasoning, is a promising approach to robust and interpretable rumor detection.
