Pre-trained language models, represented by Bidi-rectional Encoder Representations from Transformers (BERT), show great potential for deep learning-based intrusion detection systems (IDS) due to their strong semantic modeling capability. However, the high cost of training and fine-tuning limits their applicability in large-scale and resource-constrained environments. To address this challenge, we propose a Split-Federated BERT framework with adversarial training for edge intrusion detection. The framework partitions BERT into an Embedding layer deployed at the edge and Transformer and Head layers hosted in the cloud, enabling collaborative training between edge devices and the cloud. At the edge, a conditional generative adversarial network (CGAN) integrated with BERT enhances traffic feature extraction. Guided by BERT, the generator adapts to local traffic distributions, improving sample coverage and feature representation. Edge devices perform local updates to the Embedding layer, while the cloud conducts high-dimensional semantic learning using BERT’s Transformer and Head layers. During federated aggregation, a multi-head attention mechanism is employed in the cloud to differentially weight model updates, ensuring distributional alignment and stable convergence. This design decouples edge-side adversarial enhancement from federated aggregation, reducing both computational and communication overhead. Experimental results on multiple authoritative datasets demonstrate that the proposed method consistently outperforms local deep learning, BERT, federated learning, and split learning baselines in precision, recall, and F1-score, while improving edge computational efficiency and communication cost.

To address the inadequacy of conventional emergency generator tripping (EGT) in adapting to the topological changes in power systems, this paper proposes a novel knowledgeguided transfer reinforcement learning framework. For training efficiency optimization, we architect a knowledge-embedded action guidance module that incorporates invalid action masking and critical action weighting. This innovative approach transforms raw EGT action into optimized control commands, ensuring physical rationality while reducing ineffective exploration and improving decision-making efficiency. To enhance model adaptability to topology variations, we propose a knowledge-informed spatiotemporal transformer (KISTT) driven proximal policy optimization (PPO) architecture. The framework integrates the KISTT module as the PPO’s feature extraction layer, which effectively captures the spatiotemporal correlations by incorporating expert knowledge, enabling accurate perception of topological changes in EGT scenarios. Regarding model degradation to overload operations and topological variations, we propose a hybrid model transfer framework that facilitates adaptive parameter transfer from source to target domain, thereby significantly enhancing convergence speed and control performance in the target domain. Simulation results on the IEEE 39-bus system and Northeast China Power Grid demonstrate that the proposed method outperforms existing approaches in control performance, model adaptability and transferability, providing an innovative solution for EGT in dynamic grid environments.

Dynamic Security Assessment (DSA) is critical for maintaining stability in large-scale smart grids, especially with the growing integration of renewable energy sources and the inherent uncertainties. Traditional model-based analytical methods are increasingly inadequate under these complex conditions. To address these challenges, we propose a pioneering Quantum Federated Learning-based DSA (QFLDSA) method by combining hybrid quantum-classical machine learning and federated learning. QFLDSA offers an effective way to deal with high-dimensional data and uncertainties inherent in the grid. Moreover, QFLDSA leverages the unique capabilities of quantum computing to enhance the processing of differential-algebraic equations that underpin grid stability. This paper demonstrates through extensive simulations that QFLDSA significantly outperforms traditional methods, achieving the highest average F1-score performance at 97.94%, while maintaining 97.67$\pm$0.17% prediction accuracy on both classical and quantum computing devices only with fewer transmitted model parameters (reducing up to$\sim$1000X). These enhancements enable more reliable and rapid deployment of preventive stability control measures across smart grids. Our results underscore QFLDSA’s potential as a robust solution for the dynamic security challenges of modern smart grids, paving the way for future innovations in grid management technology.Note to Practitioners—In the rapidly evolving world of smart cyber-physical grids, ensuring the stability of electric power systems is paramount. Failures in these systems can lead to catastrophic blackouts, affecting countless homes and businesses. Traditional DSA methods to assess and ensure this stability, while effective, are becoming increasingly complex and vulnerable to single points of failure or cyberattacks. Enter the QFLDSA method, a novel approach we introduce in this paper. In simple terms, this method combines the strengths of quantum machine learning and federated learning to analyze data efficiently across a distributed system. Here’s why these matters: 1) Localized Analysis: Instead of relying on a central hub to analyze all data, QFLDSA allows for localized data analysis. This means that if one part of the system fails, it does not bring down the entire grid’s analysis capabilities. It is akin to having multiple control rooms instead of one, ensuring that a problem in one room does not halt the entire operation. 2) Future-Ready: As we move towards a future where quantum computing becomes more prevalent, QFLDSA is designed to work seamlessly with both today’s classical devices and tomorrow’s quantum devices. This ensures that as technology evolves, our method remains relevant and efficient. 3) Proven Performance: We have not just introduced a new method; we have rigorously tested it. Our theoretical proofs and practical tests confirm that QFLDSA offers accurate and efficient data analysis for smart grids. For industry professionals, the takeaway is clear: if looking for a resilient, future-ready, and proven method to ensure the stability of smart grid, QFLDSA offers a compelling solution.

Modern power systems face growing uncertainties and structural variations, challenging conventional emergency voltage control in terms of efficiency, scalability, and adaptability. To address these limitations, this paper proposes a scale-adaptive physics-guided transfer reinforcement learning method designed for automation-ready emergency voltage control. Training efficiency is improved by embedding an expert strategy knowledge base into imitation learning, guiding the deep reinforcement learning (DRL) agent rapidly to converge toward superior control strategies. Scalability is achieved through a hierarchical graph pooling–ensemble graph attention network (HAGPool-GAT), which produces fixed-length, scale-independent features that eliminate the necessity for retraining from scratch when deploying the model across power grids of varying scales. To enhance adaptability across heterogeneous grids, a hybrid transfer learning mechanism integrates experience transfer, model transfer, and imitation transfer. This mechanism dynamically selects strategies for intra- and cross-grid scenarios, mitigating degradation under system variations. This integration of physics-informed modeling, advanced graph representation, and hybrid transfer learning represents a novel automation paradigm for emergency voltage control. The proposed method is validated on IEEE 39-bus, IEEE 57-bus, and the Northwest China power grid, demonstrating superior efficiency, robust scalability, and reliable transferability, with clear potential application in large and evolving power systems.

Misinformation detection in social networks faces challenges due to complex semantics, scarcity of labeled data, and rapidly evolving false narratives. To address these issues, we present large language model (LLM)-augmented contrastive learning (LACL), a novel framework that integrates LLMs with contrastive learning (CL) for robust and accurate misinformation detection. We begin with an LLM-driven social media data augmentation strategy, utilizing prompt orchestration to generate diverse yet semantically consistent misinformation samples. These augmented samples are integrated into a CL-based detector, where the semantic richness and diversity introduced by the LLM enhance the CL’s discriminative feature extraction and predictive capability, thus improving generalization beyond the original training data. To align with CL’s discriminative goal, we develop a contrastive loss-aware joint training and fine-tuning approach where CL’s discriminative feature learning actively constrains the LLM’s hallucinations and guides the quality of augmentation. Through this closed-loop optimization, the CL-based detector progressively absorbs latent semantic knowledge from the LLM, effectively overcoming semantic complexity and reducing erroneous generations. Experimental results on four benchmark datasets (Twitter15, Twitter16, Weibo, and PHEME) demonstrate that LACL outperforms mainstream deep learning methods and surpasses approaches that apply commercial LLMs for detection without task-specific adaptation. These results hold consistently across different backbone LLMs (qwen and llama), highlighting LACL’s enhanced robustness, adaptability to varying language contexts, and superior generalization capability.

While Electric Vehicles (EVs) have the potential to contribute to low-carbon transportation, the burgeoning adoption of EVs and uncoordinated charging can place a significant burden on the power grid. This paper addresses the challenges of the integration of massive EVs into smart grids by introducing a comprehensive framework with heterogeneous flexibility aggregation. First, a Spatio-Temporal Heterogeneous Graph Neural Network (STH-GNN) is proposed to accurately predict EV charging demands by analyzing complex spatio-temporal relationships. Furthermore, a bi-level coordinated charging scheduling framework optimizes grid operations and individual EV charging schedules, relieving grid burden and enhancing efficient energy use based on the STH-GNN prediction. It solves the gaps between prediction errors and real-time actual charging demand. Additionally, the EV flexibility set is modelled to effectively aggregate heterogeneous EV resources, allowing for optimized deployment of large-scale EV assets by constructing the inner approximation of the Minkowski sum of the individual flexibility sets. The proposed method is verified in simulation, utilizing data from a realistic urban setting with diverse EV charging behaviours and grid conditions. Simulations demonstrate the effectiveness of the STH-GNN in accurately forecasting EV charging demands across different times and locations. The bi-level charging scheduling framework successfully manages grid burden while showcasing significant improvements in operational efficiency and cost reduction. Results validate the proposed model’s robustness and scalability, proving its potential applicability in real-world smart grid environments.

This paper presents MobiFormer, a split-federated transfer learning framework with multi-head attention designed for distributed drone Radio Access Network (RAN) slicing. The objective is to optimize slice performance isolation and training costs. Based on a flexible service metric, the problem of maximizing slice performance isolation quality is formulated as a joint optimization of slice windowing and resource allocation. For single drone autonomous operations, we construct an “unconstrained mobility and sustainable fine-tuning” paradigm, enabling drones to adapt previously trained resource slicing models to new environments with the assistance of multiple target-domain terrestrial Base Stations (BSs). This adaptation is facilitated by a Source-free Multi-target-domain Transfer Learning (SMTL) approach, where Transformer-based multi-head attention is employed on drones to integrate fine-tuned models from multiple target-domain BSs. Building on SMTL and continuing its scenario, a Clustered Split Federated Learning (CSFL) approach is developed to support multi-drone collaborative training, where BSs serve as cluster heads to aggregate parameters from member drones. To save energy, part of the onboard models are migrated to BSs while local iterations occur through gradient exchanges. Unlike SMTL, the Transformer is deployed at BSs to enhance the global model's adaptability and generalization. Extensive simulations demonstrate that MobiFormer outperforms benchmark approaches in terms of performance isolation, energy consumption, and online decision-making efficiency in distributed learning settings.

The development of mobile charging facilities for electric vehicles (EVs) has provided significant help in alleviating the pressure on active distribution networks (ADN) and traffic flow. This article proposes using the interaction between mobile charging facilities for EVs and the ADN to improve the power quality while ensuring the utility of mobile charging facility operators. First, the utility function of mobile charging facility operators is established with normal operation and emergency operation modes. The normal operation is to dispatch the mobile charging facilities for EVs requesting to be charged, while maximizing the charging benefits. To ensure the power quality for the ADN, the emergency operation is proposed to realize the power interaction between the mobile charging facilities and power grid. Furthermore, we propose an electricity price incentive mechanism to encourage optimal charging and discharging for mobile charging facilities. During the emergency operation, coordination between the mobile charging facilities and ADN is formulated as a Stackelberg game. We propose a sensitivity-based electricity price regulation algorithm and theoretically prove its equilibrium. Simulation results confirm the effectiveness and superiority of this approach, showing that the mobile charging facility and ADN can achieve a mutually beneficial outcome.

Multi-object detection in autonomous driving faces challenges due to multi-scale entities, diverse streetscapes, and limited computational resources. To address these challenges, we present MT-DyNN, a Multi-Teacher knowledge-distilled Dynamic Neural Network framework for instance-adaptive detection, optimizing detection accuracy and inference cost in autonomous driving. The framework’s student network comprises a customizable multi-branch residual detection network and a lightweight policy network. The former efficiently extracts multi-scale features in parallel without altering receptive fields, while the latter, depending on curriculum learning, captures task-relevant features and dynamically generates routing vectors to guide the activation or deactivation of residual blocks according to image instance complexity. The framework’s teacher network employs a soft-voting strategy to consolidate knowledge from multiple pre-trained teacher models, providing consistent guidance to the student. Within this distillation paradigm, the policy network’s routing search space is gradually refined, and the policy and detection networks are jointly fine-tuned to optimize the alignment between routing decisions and feature extraction. Experimental results on CIFAR and ImageNet demonstrate that compared to early exiting and stochastic depth methods, MT-DyNN achieves higher accuracy at the same inference cost and reduces the cost by 50% and 59% at comparable accuracy levels. The generated routing maintains channel sparsity across diverse scenarios.

Emergency control is essential for ensuring transient stability in power systems after faults. This study addresses the limitations in existing methods by proposing a knowledge-generative pretrained transformer (GPT)-guided generalizable reinforcement learning (RL) approach for intelligent emergency generator tripping. This approach incorporates general electrical principles and knowledge-GPT to assist deep reinforcement learning (DRL). The general electrical principles involve identifying severely disturbed generators and selecting appropriate control actions through dynamic probability. The knowledge-GPT model extracts insights from an expert strategy knowledge base, reshaping the DRL reward structure by comparing the DRL strategy with the knowledge-GPT outputs. This paradigm is designed to leverage electrical laws and domain expertise to guide the DRL training process, thereby enhancing both training efficiency and electrical consistency. To enhance generalization capability under topological changes, message passing neural networks (NNs) are integrated into the DRL architecture, effectively simulating power flow dynamics in transmission lines. The proposed method is validated through simulations on the IEEE 39-bus system and the Northeast power grid of China, demonstrating superior control effectiveness and adaptability compared to existing approaches, offering a more robust solution for emergency control in complex power systems.

The current state-of-the-art approach for battery state-of-health (SOH) estimation typically employs a centralized computing framework, wherein data from local battery management systems (BMSs) is aggregated and trained on a cloud server, due to limited computing resources at the BMS. However, this framework presents various challenges, including frequent data communication, latency, data security, and degraded prediction accuracy. To address these issues, this study proposes a novel adaptive multipersonalized federated learning (FL) algorithm for evaluating the SOH of multiple batteries, aggregating multiple local SOH estimation models into a global model while locally preserving battery data. The algorithm utilizes the difference of importance weights between global and local models to regulate the local loss, incorporates adaptive personalization layers with loss variation, and employs clustering techniques to form multiple global models from distinct local models, leading to a more accurate and tailored prediction. Additionally, an adaptively SOH-related differential privacy protection mechanism is integrated to enhance the protection of local battery data while ensuring robust model performance. An extensive case study has demonstrated that the adaptive multipersonalized FL algorithm outperforms other methods in terms of estimation accuracy and operational risk. Specifically, it achieves a reduction in mean absolute error by 0.14% and 6.01% compared to traditional FL and local training methods, respectively, and exhibits nearly fivefold lower operational risk compared to centralized training.

Deep reinforcement learning (DRL) has become prevalent for decision-making task assignments in mobile crowdsensing (MCS). However, when facing sensing scenarios with varying numbers of workers or task attributes, existing DRL-based task assignment schemes fail to generate matching policies continuously and are susceptible to environmental fluctuations. To overcome these issues, a twin-delayed deep stochastic policy gradient (TDDS) approach is presented for balanced and low-latency MCS task decomposition and parallel subtask allocation. A masked attention mechanism is incorporated into the policy network to enable TDDS to adapt to task-attribute and subtask variations. To enhance environmental adaptability, an off-policy DRL algorithm incorporating experience replay is developed to eliminate sample correlation during training. Gumbel-Softmax sampling is integrated into the twin-delayed deep deterministic policy gradient (TD3) to support discrete action space decisions and a customized reward strategy to reduce task completion delay and balance workloads. Extensive simulation results confirm that the proposed scheme outperforms mainstream DRL baselines in terms of environmental adaptability, task completion delay, and workload balancing.

A slicing-based collaborative task offloading framework for space-air-ground integrated vehicular networks is proposed in this study, which can provide differentiated quality-of-service (QoS) guarantees for task offloading for high-speed vehicles while maximizing the number of completed tasks. A service-oriented radio access network (RAN) slicing framework is presented that supports slicing window adaptation, spectrum and computing resource orchestration, and collaboration among heterogeneous base stations. Based on the queuing model, the collaborative decision-making of RAN slicing and task offloading is modeled as a problem of maximizing the number of long-term task completions, which consists of three subproblems-slicing window division, resource slicing, and task scheduling-which are solved by a multi-access edge computing (MEC)-enabled controller, forming a closed loop with the slicing window as the period. When a new slicing window arrives, the controller determines its duration according to task traffic fluctuations and allocates resources to RAN slices through an optimization method. A double deep Q-learning network (DDQN)-based algorithm is developed for scheduling workflow on small time scales within a slicing window. Simulation results demonstrate that the proposed scheme performs better than existing approaches in terms of adaptability, task completion rate, and control overhead.

With advancements in cellular vehicle-to-everything (C- V2X) and drone manufacturing technologies, integrating drone-small-cells (DSCs) into terrestrial cellular networks is a promising solution to enabling diversified vehicle applications. In this paper, a multi-DSC-assisted dynamic spectrum management framework is presented to maximize the network utility under quality-of-service (QoS) constraints in 5G and beyond cellular vehicular networks. The network utility maximization problem is formulated as mixed-integer nonlinear programming regarding association patterns between vehicles and base stations (BSs) and spectrum partitioning among heterogeneous BSs. For mathe-matical tractability, the joint optimization problem for spectrum partitioning and vehicle- DSC associations is transformed as a biconcave optimization problem. An alternate search algorithm is then designed to determine vehicle association patterns and spec-trum slicing ratios. Our simulation demonstrates that compared with state-of-the-art methods, the proposed scheme achieves a significant performance improvement in network throughput and spectrum utilization.

Incorporating link correlation awareness into wireless network protocols to facilitate data transmission is an important research issue. In this paper, we focus on link correlation based security threat and countermeasure in wireless networks. By taking advantage of the vulnerability of beacon-based link correlation measurement and the blind spot of malicious node detection mechanisms, we design a new type of link correlation spoofing attack (LCSA) to decrease protocol performance by distorting link correlation information while escaping the tracking of any watchdog and trust systems. Typical cases are analyzed to quantify how the LCSA covertly weakens protocol performance. We also propose beacon-trap (BT), a countermeasure embedded in the beacon-based link condition measurement protocol. Using link diversity as a cover, BT sets traps in the beacon sending sequence to ambush malicious nodes that launch LCSAs without extra control overhead. The performance of BT is not affected by changes in the size of a network or the distribution of nodes. Numerical results demonstrate the superiority and effectiveness of BT against LCSAs in terms of malicious node detection success rate and speed under different parameter settings.

Aiming to minimize the operators' cost while preserving user experience, we propose a resource scheduling mechanism for cooperative cloudlets in edge computing with a centralized controller. The interactions between cloudlets and the controller are formulated as a two-stage Stackelberg game to determine the amount of physical resources assigned to each cloudlet during deployment phase and the price of resources shared among cooperated cloudlets during operation phase.

In recent years there has been a growing interest in sparse representation of signals based on overcomplete dictionaries. Selecting few atoms that best match the signal structure, the signal is described by linear combination of these atoms. In this paper, we propose a novel overcomplete dictionary design algorithm for sparse representation of piecewise stationary signals. An effective and easy-to-use overcomplete dictionary is constructed in accordance with the parametric autocorrelation function model of piecewise stationary processes. Furthermore, a sparse decomposition algorithm in terms of nonlinear approximation is designed to obtain sparse representation of piecewise stationary signals, which has lower computational complexity and better practicability than the conventional sparse decomposition algorithms. The experimental results demonstrate that the proposed method avails for higher sparsiry of signal representation and better reconstruction performance than sparse representation of signals based on overcomplete DCT dictionary.