With the increasing penetration of distributed photovoltaic (PV) generation in distribution grids, PV access locations significantly affect voltage stability and network losses. In this paper, we proposed a critical node identification and PV access planning method based on fuzzy cognitive maps (FCMs). An FCM-based distribution grid model was established by incorporating time-varying load characteristics and PV output fluctuations, and the weight matrix was learned from historical operating data using a data-driven optimization approach. A node importance index derived from out-degree centrality was then proposed to identify critical nodes in the distribution grid. Simulation studies on the IEEE 33-node and 118-node standard test systems under different PV penetration levels demonstrated that connecting PV to non-critical nodes effectively mitigates voltage fluctuations and reduces system power losses compared with critical-node-based access strategies. The proposed method provides an interpretable and effective decision-support tool for distributed PV integration in distribution grids.
Real-time detection of tiny insulator defects is critical for power grid reliability. However, it remains a major challenge for Unmanned Aerial Vehicle (UAV) based inspection systems due to the trade-off between model accuracy and computational efficiency. To address this challenge, we proposed a novel lightweight object detector that incorporates a Frequency-Aware Enrichment Module (FAEM) into a You Only Look Once (YOLO) architecture for tiny insulator defect detection. FAEM introduces a learnable frequency-domain filtering pipeline operating on the Fourier magnitude spectrum while preserving phase information. It selectively amplifies high-frequency textural signatures of small defects (e.g., cracks and pollution flashovers) often missed by conventional detectors (YOLOv5n, YOLOv8s, and LiteYOLO-ID (2024)). The proposed approach was evaluated on three public datasets: IDID-Plus (~1,600 images), PLIDD (~4,927 images), and SFID (~320 images). The IDID-Plus and PLIDD datasets were used for training, validation, and testing with a 70/15/15 split, while SFID was employed exclusively for benchmarking robustness under degraded visibility. The performance was measured using Mean Average Precision (mAP) derived from the area under the precision recall curve. All experiments were conducted in a reproducible Google Colab Pro environment using an NVIDIA Tesla T4 GPU and PyTorch 2.6.0. The experimental results demonstrated that FAEM-YOLO achieved 90.7% mAP@0.5 and 50.2% mAP@0.5:0.95 on IDID-Plus, significantly outperforming LiteYOLO-ID (2024). On PLIDD, the model attained 73.9% mAP@0.5 and 40.5% mAP@0.5:0.95, surpassing the YOLOv8n baseline by 3.9 points on the stricter metric while using 20% fewer parameters (4.2M vs. 5.3M). Furthermore, FAEM-YOLO recorded 99.2% mAP@0.5 and 84.4% mAP@0.5:0.95 on the SFID dataset. These results underscore that frequency-domain enrichment generalizes well on different dataset; hence, it is an efficient strategy for visual inspection in UAV-based power line monitoring.
This work has proposed a novel robust fixed-time tracking control scheme for quadrotor unmanned aerial vehicles subject to unknown external disturbances and model uncertainties. Accounting for the inherent underactuated dynamics of the quadrotor, a hierarchical control architecture was constructed that decouples position and attitude regulation into two separate control loops, namely, an outer loop dedicated to translational motion and an inner loop responsible for rotational dynamics. The proposed framework integrates a fixed-time backstepping design with a command filtering technique to mitigate the explosion-of-complexity problem that typically arises in conventional backstepping approaches, while simultaneously guaranteeing fixed-time convergence of the tracking errors. To effectively address the effects of model uncertainties and external perturbations acting on the system, an adaptive fixed-time disturbance observer was introduced and integrated into the control architecture. This observer is capable of accurately estimating the unknown disturbances and uncertainties, achieving estimation convergence within a predetermined time that remains independent of the initial conditions of the system. Rigorous stability analysis based on Lyapunov theory was carried out to establish the fixed-time convergence of the tracking errors to a bounded neighborhood of the origin. Comprehensive numerical simulations were performed on a quadrotor system, and the results obtained were compared against several existing control methods. The comparative analysis clearly demonstrated that the proposed controller achieves superior tracking accuracy, faster convergence, and improved disturbance rejection performance compared to the alternate approaches considered.
The asymmetrical nature of the modern cyber threat landscape allows advanced persistent threats (APTs) to innovate tactics at a significantly faster rate than defensive frameworks can document them. While the MITRE ATT&CK® framework provides a standardized taxonomy of known behaviors, it essentially functions as a retrospective database - a ―dictionary of the past‖ which fails to anticipate future ―zero-day‖ tactics, techniques, and procedures (TTPs). This paper introduces Evo-TTP, a comprehensive framework for the predictive generation of novel and robust tactics, techniques, and procedures via big data mining and adversarial learning. By leveraging the massive structured data from the MITRE ATT&CK® Enterprise Matrix v18.0, Evo-TTP treats the prediction of future threats as a high-dimensional pattern completion problem. Our methodology addresses two primary failures in current generative AI applications to big data: mode collapse, which refers to hallucinating biologically or technically impossible scenarios, and algorithmic brittleness, which is characterized by its susceptibility to adversarial perturbations. We employ a tripartite approach: (1) applying semantic pattern mining on the v18.0 dataset to create a baseline knowledge graph that reveals hidden correlations; (2) utilizing synthetic novelty expansion with a teacher-student architecture, using Llama-3.1-405B as the teacher and Llama-3.1-8B as the student model, to overcome data scarcity; and (3) conducting training in adversarial group relative policy optimization (GRPO). This training regime maximizes a composite reward function by balancing novelty, technical feasibility, and resilience against adversarial noise. Validated against the 2025 benchmarks and vetted according to SafeGen-X principles, Evo-TTP demonstrates a 23.1% increase in utility and an 18.2% improvement in robustness to adversarial attacks when compared with standard fine-tuning methods. This research positions generative AI not only as a text processor but also as a critical instrument in big data for uncovering the hidden evolutionary mechanics of cyberwarfare.
In the copper smelting Flue Gas Sulfuric Acid, accurate prediction of the converter inlet temperature is of great engineering significance for ensuring safe operation, reducing energy consumption, and improving system stability. To address the strong coupling, long time delay, and pronounced nonlinearity inherent in this process, we proposed a data-driven soft-sensing method designed for real-world industrial applications. The proposed approach was validated in an actual “two-conversion–two-absorption” industrial system. First, the Partial Autocorrelation Function was employed to analyze Distributed Control System data, identifying the most relevant input variables and time windows to provide a high-correlation feature foundation for model construction. Then, a Gated Recurrent Unit neural network was utilized to model temporal dependencies, effectively capturing complex dynamic relationships while suppressing noise interference. Furthermore, an Attention Mechanism was incorporated to enhance the model’s focus on key temporal features, improving adaptability to process fluctuations and enhancing the interpretability of prediction results. Finally, the proposed model was implemented and verified in an industrial Flue Gas Sulfuric Acid process. The experimental results demonstrated that the method maintains stable predictive performance and accurately captures the dynamic variation of the first-layer converter inlet temperature. It effectively assists operators in adjusting process parameters in real time, thereby improving system safety, continuity, and operational reliability. The findings verify the engineering feasibility and practical applicability of the proposed method, providing strong technical support for the intelligent operation of copper smelting sulfuric acid production systems.
Smart grid construction is a goal of the State Grid Corporation of China (SGCC). Transformer equipment, as the foundation of substation power supply and a critical component of the national power grid, requires stable operation. This paper proposes an early warning method for transformer equipment based on robot infrared temperature prediction, which mainly includes three parts: robot infrared temperature measurement, GWO-SVR temperature prediction based on historical equipment data, and determination of defect types according to heating defect classification standards for transformer equipment. This paper leverages the strong generalization ability of support vector regression (SVR), which maintains good prediction performance on small-sample data. To address limitations in the traditional SVR model caused by penalty factors and kernel function parameters, the Greey Wolf Optimizer (GWO) is introduced to optimize parameters and achieve optimal selection. The resulting GWO-SVR model is used for equipment temperature prediction, and defect types are identified in advance using dual-level grading standards from the state and the Taizhou Power Bureau, enabling early warning of transformer equipment. Compared with the GRNN, BP neural network, SVR model, LSTM, and Attention-LSTM, the GWO-SVR achieves better performance in substation temperature prediction, as evaluated by mean absolute error (MAE) and mean square error (MSE) metrics. In addition, GWO-SVR enables heat fault early warning based on temperature prediction and classification standards of heating defects for transformer equipment. This paper contributes to the unique application context of State Grid intelligence with previously unaddressed challenges.
This paper proposes an adaptive fixed-frequency hysteresis band control method for a unified power quality conditioner (UPQC). First, a detailed analysis of the mathematical model and control for UPQC is performed. The series inverter is controlled to compensate for harmonic distortions in the source voltage, thereby maintaining a balanced voltage waveform applied to the load with harmonic distortion within permissible limits. Simultaneously, the shunt inverter is controlled to compensate for the current harmonics generated by nonlinear loads. The reference voltage for compensation is determined based on the distorted source voltage, while the reference harmonic current is derived using the ip –iq harmonic detection method. The harmonic mitigation performance of the UPQC is evaluated under two control scenarios. In Scenario 1, a conventional proportional integral (PI) controller is used for DC-link voltage regulation, and traditional hysteresis band controllers are used for both the series and shunt inverters. In Scenario 2, a fuzzy-PI controller is applied for DC-link voltage regulation, along with an adaptive fixed-frequency hysteresis band controller for both the series and shunt inverters. Simulation and experimental results under ideal, sag, swell, unbalanced, and distorted source-voltage conditions demonstrate that Scenario 2 is more effective than Scenario 1 in reducing the total harmonic distortion (THD) of the supply current and the load voltage and in stabilizing the DC-link voltage.
Accurately tracking planned output is a core requirement for ensuring the stable grid connection of hybrid wind-photovoltaic-storage systems. However, the strong randomness of wind and solar power output not only affects the tracking accuracy but also leads to frequent charge and discharge cycles of the energy storage system, thereby shortening its service life. To address this, we propose a control method for the energy storage system that accounts for wind and solar uncertainty, with the dual objectives of improving the planned output tracking performance and extending the energy storage system lifespan. By optimally controlling the charge and discharge of the energy storage system, this method enables the output of the hybrid wind-photovoltaic-storage system to track the planned output. The approach models wind and solar power joint distribution using Copula functions and applies K-means clustering to extract representative scenarios, reducing computational complexity. An interval control mechanism defines acceptable deviation ranges around planned output, minimizing unnecessary energy storage system activity during minor fluctuations. The core contribution is a generalized mutual entropy–proximal policy optimization control strategy, which quantifies non-Gaussian tracking errors using generalized mutual entropy while leveraging proximal policy optimization for adaptive learning and improved robustness. Real-world data validation demonstrated significant improvements: 31.6% reduction in average tracking error, 12.6% increase in output compliance within acceptable ranges, 33.4% reduction in maximum tracking error, and approximately 35% fewer ESS charge/discharge cycles. These results confirmed the method’s effectiveness in addressing renewable generation uncertainty while improving tracking performance and extending energy storage system service life.