Search Results (131)

To address the issue of heavy reliance on manual intervention in substation maintenance tasks, this paper proposes an end-to-end hierarchical intelligent inference method for collaborative operation of grid switches. The method constructs a unified knowledge environment that can simultaneously describe the operational characteristics of both the power grid and the substation, and combines Dueling Double Deep Q-Network (D3QN) with Multi-Task Dueling Double Deep Q-Network (MT-D3QN) algorithms for interactive training, achieving hierarchical inference. The upper layer uses bays as the base nodes to reflect the power flow, designing a reward and penalty function under N-1 power flow constraints and ring-current impact constraints, optimizing the load transfer plan for the power outages caused by maintenance tasks. The lower layer uses switches as the base nodes to reflect the main wiring status of the substation, introduces a multi-task learning mechanism for parallel training of bays with the same tasks, designs the reward and penalty function according to the five protection rules, and optimizes the switching operations within the bay. The experimental results show that the trained model can quickly deduce the switching operation sequence for different maintenance tasks. Full article

This paper presents Hy-LIFT (Hybrid LLM-Integrated Fault Diagnosis Toolkit), a multi-stage framework for interpretable and data-efficient fault diagnosis in 5G/6G networks that integrates a high-precision interpretable rule-based engine (IRBE) for known patterns, a semi-supervised classifier (SSC) that leverages scarce labels and abundant unlabeled logs via consistency regularization and pseudo-labeling, and an LLM Augmentation Engine (LAE) that generates operator-ready, context-aware explanations and zero-shot hypotheses for novel faults. Evaluations on a five-class, imbalanced Dataset-A and a simulated production setting with noise and label scarcity show that Hy-LIFT consistently attains higher macro-F1 than rule-only and standalone ML baselines while maintaining strong per-class precision/recall (≈0.85–0.93), including minority classes, indicating robust generalization under class imbalance. IRBE supplies auditable, high-confidence seeds; SSC expands coverage beyond explicit rules without sacrificing precision; and LAE improves operational interpretability and surfaces potential “unknown/novel” faults without altering classifier labels. The paper’s contributions are as follows: (i) a reproducible, interpretable baseline that doubles as a high-quality pseudo-label source; (ii) a principled semi-supervised learning objective tailored to network logs; (iii) an LLM-driven explanation layer with zero-shot capability; and (iv) an open, end-to-end toolkit with scripts to regenerate all figures and tables. Overall, Hy-LIFT narrows the gap between brittle rules and opaque black-box models by combining accuracy, data efficiency, and auditability, offering a practical path toward trustworthy AIOps in next-generation mobile networks. Full article

Objectives: This paper presents an innovative control framework for the classical Cart–Pole problem. Methods: The proposed framework combines Interval Type-2 Fuzzy Logic, the Dueling Double DQN deep reinforcement learning algorithm, and adaptive reward shaping techniques. Specifically, fuzzy logic acts as an a priori knowledge layer that incorporates measurement uncertainty in both angle and angular velocity, allowing the controller to generate adaptive actions dynamically. Simultaneously, the deep Q-network is responsible for learning the optimal policy. To ensure stability, the Double DQN mechanism successfully alleviates the overestimation bias commonly observed in value-based reinforcement learning. An accelerated convergence mechanism is achieved through a multi-component reward shaping function that prioritizes angle stability and survival. Results: Given the training results, the method stabilizes rapidly; it achieves a 100% success rate by episode 20 and maintains consistent high rewards (650–700) throughout training. While Standard DQN and other baselines take 100+ episodes to become reliable, our method converges in about 20 episodes (4–5 times faster). It is observed that in comparison with advanced baselines like C51 or PER, the proposed method is about 15–20% better in final performance. We also found that PPO and QR-DQN surprisingly struggle on this task, highlighting the need for stability mechanisms. Conclusions: The proposed approach provides a practical solution that balances exploration with safety through the integration of fuzzy logic and deep reinforcement learning. This rapid convergence is particularly important for real-world applications where data collection is expensive, achieving stable performance much faster than existing methods without requiring complex theoretical guarantees. Full article

When trained on long-tailed distributions, deep neural networks often suffer performance degradation and model bias due to the dominance of head classes. Existing reweighting and sampling strategies have significant limitations, such as reliance on fixed heuristics and inability to adapt to dynamic sample difficulty and class imbalance. Additionally, they fail to integrate sample-level granularity with class-level balance, further inadequately addressing the global imbalance issue. Motivated by these challenges, we introduce the Progressive Hierarchical Adaptation for Sample-Efficient rebalancing (PHASE) training framework, which employs a double-layer tuning paradigm to optimize performance under long-tailed distributions. Specifically, the double-layer tuning paradigm adopted by PHASE runs as follows: (1) an early-stage difficulty-aware mechanism targets those difficult-to-classify samples to guide representation learning; and (2) a later-stage multi-scale reweighting strategy integrates class distribution statistics with sample characteristics. This method ensures fine-grained adaptability and global balance, and thus outperforming those static or localized techniques. Extensive experiments on CIFAR10-LT, CIFAR100-LT, and ImageNet-LT datasets demonstrate that PHASE can significantly improve the accuracy of tail classes without degenerating head class performance, and acquire state-of-the-art classification results. PHASE provides a novel paradigm for long-tailed image recognition. Full article

Cloud Top Height (CTH), defined as the altitude of the highest cloud layer above mean sea level, is a crucial geophysical parameter for quantifying cloud radiative effects, analyzing severe weather systems, and improving climate models. To enhance the accuracy of CTH retrieval from Fengyun-4A (FY-4A) satellite data, this study proposes a multi-stage deep learning framework that progressively refines cloud parameter estimation. The method utilizes cloud information from the FY-4A/AGRI (Advanced Geosynchronous Radiation Imager) Level 1 calibrated scanning imager radiance data product to construct a multi-source data fusion neural network model. The model inputs combine multi-channel radiance data with cloud parameters, including Cloud Top Temperature (CTT) and Cloud Top Pressure (CTP). We used the CTH measurement data from the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) satellite as a reference to verify the model output. Results demonstrate that the proposed multi-stage model significantly improves retrieval accuracy. Compared to the official FY-4A CTH product, the Mean Absolute Error (MAE) was reduced by 49.12% to 2.03 km, and the Pearson Correlation Coefficient (PCC) reached 0.85. To test the applicability of the model under complex weather conditions, we applied it to the CTH inversion of the double typhoon event on 10 August 2019. The model successfully characterized the spatial distribution of CTH within the typhoon regions. The results are consistent with the National Satellite Meteorological Centre (NSMC) reports and clearly reveal the different intensity evolutions of the two typhoons. This research provides an effective solution for high-precision retrieval of high-level cloud CTH at a large scale, using geostationary meteorological satellite remote sensing data. Full article

Intelligent connected vehicles (ICVs) face challenges in handling intensive onboard computational tasks due to limited computing capacity. Vehicular edge computing networks (VECNs) offer a promising solution by enabling ICVs to offload tasks to mobile edge computing (MEC), alleviating computational load. As transportation systems are dynamic, vehicular tasks and MEC capacities vary over time, making efficient task offloading and resource allocation crucial. We explored a vehicle–road collaborative edge computing network and formulated the task offloading scheduling and resource allocation problem to minimize the sum of time and energy costs. To address the mixed nature of discrete and continuous decision variables and reduce computational complexity, we propose a hybrid hierarchical deep reinforcement learning (HHDRL) algorithm, structured in two layers. The upper layer of HHDRL enhances the double deep Q-network (DDQN) with a self-attention mechanism to improve feature correlation learning and generates discrete actions (communication decisions), while the lower layer employs deep deterministic policy gradient (DDPG) to produce continuous actions (power control, task offloading, and resource allocation decision). This hybrid design enables efficient decomposition of complex action spaces and improves adaptability in dynamic environments. Results from numerical simulations reveal that HHDRL achieves a significant reduction in total computational cost relative to current benchmark algorithms. Furthermore, the robustness of HHDRL to varying environmental conditions was confirmed by uniformly designing random numbers within a specified range for certain simulation parameters. Full article

The Vehicular Internet of Things (V-IoT) has emerged as a cornerstone of next-generation intelligent transportation systems (ITSs), enabling applications ranging from safety-critical collision avoidance and cooperative awareness to infotainment and fleet management. These heterogeneous services impose stringent quality-of-service (QoS) demands for latency, reliability, and fairness while competing for limited and dynamically varying spectrum resources. Conventional schedulers, such as round-robin or static priority queues, lack adaptability, whereas deep reinforcement learning (DRL) solutions, though powerful, remain computationally intensive and unsuitable for real-time roadside unit (RSU) deployment. This paper proposes a lightweight and interpretable reinforcement learning (RL)-based spectrum management framework for Vehicular Internet of Things (V-IoT) networks. Two enhanced Q-Learning variants are introduced: a Value-Prioritized Action Double Q-Learning with Constraints (VPADQ-C) algorithm that enforces reliability and blocking constraints through a Constrained Markov Decision Process (CMDP) with online primal–dual optimization, and a contextual Q-Learning with Upper Confidence Bound (Q-UCB) method that integrates uncertainty-aware exploration and a Success-Rate Prior (SRP) to accelerate convergence. A Risk-Aware Heuristic baseline is also designed as a transparent, low-complexity benchmark to illustrate the interpretability–performance trade-off between rule-based and learning-driven approaches. A comprehensive simulation framework incorporating heterogeneous traffic classes, physical-layer fading, and energy-consumption dynamics is developed to evaluate throughput, delay, blocking probability, fairness, and energy efficiency. The results demonstrate that the proposed methods consistently outperform conventional Q-Learning and Double Q-Learning methods. VPADQ-C achieves the highest energy efficiency (≈$8.425\times {10}^7$ bits/J) and reduces interruption probability by over $60\%$, while Q-UCB achieves the fastest convergence (within ≈190 episodes), lowest blocking probability (≈0.0135), and lowest mean delay (≈0.351 ms). Both schemes maintain fairness near $0.364$, preserve throughput around 28 Mbps, and exhibit sublinear training-time scaling with $\mathcal{O}\left(1\right)$ per-update complexity and $\mathcal{O}\left(N^2\right)$ overall runtime growth. Scalability analysis confirms that the proposed frameworks sustain URLLC-grade latency (<0.2 ms) and reliability under dense vehicular loads, validating their suitability for real-time, large-scale V-IoT deployments. Full article

Mapping lithology in areas with dense vegetation remains a major challenge for remote sensing, as plant cover tends to obscure the spectral signatures of underlying rock formations. This study tackles that issue by comparing the performance of three custom-built lightweight deep learning models in the mixed-vegetation terrain of the surroundings of the Vălioara Valley, Romania. We used time-series data from Sentinel-2 and elevation data from the SRTM, with preprocessing techniques such as the Principal Component Analysis (PCA) and the Forced Invariance Method (FIM) to reduce the spectral interference caused by vegetation. Predictions were made with a Multi-Layer Perceptron (MLP), a Convolutional Neural Network (CNN), and a Vision Transformer (ViT). In addition to measuring the classification accuracy, we assessed how the different models handled vegetation coverage. We also explored how vegetation density (NDVI) correlated with the classification results. Tests show that the Vision Transformer outperforms the other models by 6%, offering a stronger resilience to vegetation interference, while FIM doubled the model confidence in specific (locally rare) lithologies and decorrelated vegetation in multiple measures. These findings highlight both the potential of ViTs for remote sensing in complex environments and the importance of applying vegetation suppression techniques like FIM to improve geological interpretation from satellite data. Full article

Electrical networks face operational challenges from power quality-affecting disturbances. Since disturbance signatures directly affect classifier performance, optimized feature selection becomes critical for accurate power quality assessment. The pursuit of robust feature extraction inevitably constrains the dimensionality of the discriminative feature set, but the complexity of the recognition model will be increased and the recognition speed will be reduced if the feature vector dimension is too high. Building upon the aforementioned requirements, in this paper, we propose a feature extraction framework that combines improved symplectic geometric mode decomposition, refined generalized multiscale quantum entropy, and refined generalized multiscale reverse dispersion entropy. Firstly, based on the intrinsic properties of power quality disturbance (PQD) signals, the embedding dimension of symplectic geometric mode decomposition and the adaptive mode component screening method are improved, and the PQD signal undergoes tri-band decomposition via improved symplectic geometric mode decomposition (ISGMD), yielding distinct high-frequency, medium-frequency, and low-frequency components. Secondly, utilizing the enhanced symplectic geometric mode decomposition as a foundation, the perturbation features are extracted by the combination of refined generalized multiscale quantum entropy and refined generalized multiscale reverse dispersion entropy to construct high-precision and low-dimensional feature vectors. Finally, a double-layer composite power quality disturbance model is constructed by a deep extreme learning machine algorithm to identify power quality disturbance signals. After analysis and comparison, the proposed method is found to be effective even in a strong noise environment with a single interference, and the average recognition accuracy across different noise environments is 97.3%. Under the complex conditions involving multiple types of mixed perturbations, the average recognition accuracy is maintained above 96%. Compared with the existing CNN + LSTM method, the recognition accuracy of the proposed method is improved by 3.7%. In addition, its recognition accuracy in scenarios with small data samples is significantly better than that of traditional methods, such as single CNN models and LSTM models. The experimental results show that the proposed strategy can accurately classify and identify various power quality interferences and that it is better than traditional methods in terms of classification accuracy and robustness. The experimental results of the simulation and measured data show that the combined feature extraction methodology reliably extracts discriminative feature vectors from PQD. The double-layer combined classification model can further enhance the model’s recognition capabilities. This method has high accuracy and certain noise resistance. In the 30 dB white noise environment, the average classification accuracy of the model is 99.10% for the simulation database containing 63 PQD types. Meanwhile, for the test data based on a hardware platform, the average accuracy is 99.03%, and the approach’s dependability is further evidenced by rigorous validation experiments. Full article

This paper proposes a new accelerated fixed-point algorithm based on a double-inertial extrapolation technique for solving structured variational inclusion and convex bilevel optimization problems. The underlying framework leverages fixed-point theory and operator splitting methods to address inclusion problems of the form $0\in (A+B)(x)$, where A is a cocoercive operator and B is a maximally monotone operator defined on a real Hilbert space. The algorithm incorporates two inertial terms and a relaxation step via a contractive mapping, resulting in improved convergence properties and numerical stability. Under mild conditions of step sizes and inertial parameters, we establish strong convergence of the proposed algorithm to a point in the solution set that satisfies a variational inequality with respect to a contractive mapping. Beyond theoretical development, we demonstrate the practical effectiveness of the proposed algorithm by applying it to data classification tasks using Deep Extreme Learning Machines (DELMs). In particular, the training processes of Two-Hidden-Layer ELM (TELM) models is reformulated as convex regularized optimization problems, enabling robust learning without requiring direct matrix inversions. Experimental results on benchmark and real-world medical datasets, including breast cancer and hypertension prediction, confirm the superior performance of our approach in terms of evaluation metrics and convergence. This work unifies and extends existing inertial-type forward–backward schemes, offering a versatile and theoretically grounded optimization tool for both fundamental research and practical applications in machine learning and data science. Full article

Accurate state-of-charge (SOC) estimation is crucial in smart-building energy management systems, where rooftop photovoltaics and lithium-ion energy storage systems must be coordinated to align renewable generation with real-time demand. This paper introduces a novel, modular hybrid framework for SOC estimation, which synergistically combines the predictive power of artificial neural networks (ANNs), the logical consistency of finite state automata (FSA), and an adaptive dynamic supervisor layer. Three distinct ANN architectures—feedforward neural network (FFNN), long short-term memory (LSTM), and 1D convolutional neural network (1D-CNN)—are employed to extract comprehensive temporal and spatial features from raw data. The inherent challenge of ANNs producing physically irrational SOC values is handled by processing their raw predictions through an FSA module, which constrains physical validity by applying feasible transitions and domain constraints based on battery operational states. To further enhance the adaptability and robustness of the framework, two advanced supervisor mechanisms are developed for model selection during estimation. A lightweight rule-based supervisor picks a model transparently using recent performance scores and quick signal heuristics, whereas a more advanced double deep Q-network (DQN) reinforcement-learning supervisor continuously learns from reward feedback to adaptively choose the model that minimizes SOC error under changing conditions. This RL agent dynamically selects the most suitable ANN+FSA model, significantly improving performance under varying and unpredictable operational conditions. Comprehensive experimental validation demonstrates that the hybrid approach consistently outperforms raw ANN predictions and conventional extended Kalman filter (EKF)-based methods. Notably, the RL-based supervisor exhibits good adaptability and achieves lower error results in challenging high-variance scenarios. Full article

The switching characteristics of GaN HEMT devices exhibit a very complex dynamic nonlinear behavior and multi-physics coupling characteristics, and traditional switching current models based on physical mechanisms have significant limitations. This article adopts a hybrid architecture of convolutional neural network and long short-term memory network (CNN-LSTM). In the 1D-CNN layer, the one-dimensional convolutional neural network can automatically learn and extract local transient features of time series data by sliding convolution operations on time series data through its convolution kernel, making these local transient features present a specific form in the local time window. In the double-layer LSTM layer, the neural network model captures the transient characteristics of switch current through the gating mechanism and state transfer. The hybrid architecture of the constructed model has significant advantages in accuracy, with metrics such as root mean square error (RMSE) and mean absolute error (MAE) significantly reduced, compared to traditional switch current models, solving the problem of insufficient accuracy in traditional models. The neural network model has good fitting performance at both room and high temperatures, with an average coefficient close to 1. The new neural network hybrid architecture has short running time and low computational resource consumption, meeting the needs of practical applications. Full article

Various approaches have been explored to address the path planning problem for mobile robots. However, it remains a significant challenge, particularly in environments where a multi-tasking mobile robot operates alongside stochastically moving humans. This paper focuses on path planning for a mobile robot executing multiple pickup and delivery tasks in an environment shared with humans. To plan a safe path and achieve high task success rate, a Reinforcement Learning (RL)-based double layer controller is proposed in which a double-layer learning algorithm is developed. The high-level layer integrates a Finite-State Automaton (FSA) with RL to perform global strategy learning and task-level decision-making. The low-level layer handles local path planning by incorporating a Markov Decision Process (MDP) that accounts for environmental uncertainties. We verify the proposed double layer algorithm under different configurations and evaluate its performance based on several metrics, including task success rate, reward, etc. The proposed method outperforms conventional RL in terms of reward (+63.1%) and task success rate (+113.0%). The simulation results demonstrate the effectiveness of the proposed algorithm in solving path planning problem with stochastic human uncertainties. Full article

This study aims to investigate the effectiveness of two machine learning methods for the prediction of the mechanical and fracture properties of Cenosphere-reinforced lightweight thermoset polyurethane composites. To evaluate the effectiveness of the models, datasets from our experimental study of composites made of five different volume fractions (0% to 40%) of Cenospheres (hollow Aluminum Silicate particles) in increments of 10% are fabricated. Experiments are conducted to determine the effect of the volume fraction of Cenospheres on Young’s modulus (both in tension and compression), percentage elongation at break, tensile strength, specific tensile strength, and fracture toughness of the composites. Two machine learning models, shallow artificial neural network (ANN) and the non-linear deep neural network (DNN), are employed to predict the above properties. A parametric study was performed for each model and optimized parameters were identified and later used to predict the properties beyond 40% volume fraction of Cenospheres. The predictions of non-linear DNN demonstrated less slope than shallow ANN and, for mass density, the non-linear DNN had unexpected predictions of increasing mass density with the addition of lighter Cenospheres. Hence, a double-hidden-layer DNN is used to predict the mass density beyond 40%, which provides the expected behavior. Full article

Driven by the global “double carbon” goal, the volatility of renewable energy poses a challenge to the stability of power systems. Traditional methods have difficulty dealing with high-dimensional nonlinear data, and the single deep learning model has the limitations of spatiotemporal feature decoupling and being a “black box”. Aiming at the problem of insufficient accuracy and interpretability of power load forecasting in a renewable energy grid connected scenario, this study proposes an interpretable spatiotemporal feature fusion model based on an attention mechanism. Through CNN layered extraction of multi-dimensional space–time features such as meteorology and electricity price, BiLSTM bi-directional modeling time series rely on capturing the evolution rules of load series before and after, and the improved self-attention mechanism dynamically focuses on key features. Combined with the SHAP quantitative feature contribution and feature deletion experiment, a complete chain of “feature extraction time series modeling weight allocation interpretation and verification” is constructed. The experimental results show that the determination coefficient R² of the model on the Australian electricity market data set reaches 0.9935, which is 84.6% and 59.8% higher than that of the LSTM and GRU models, respectively. The prediction error (RMSE = 105.5079) is 9.7% lower than that of TCN-LSTM model and 52.1% compared to the GNN (220.6049). Cross scenario validation shows that the generalization performance is excellent (R² ≥ 0.9849). The interpretability analysis reveals that electricity price (average absolute value of SHAP 716.7761) is the core influencing factor, and its lack leads to a 0.76% decline in R². The research breaks through the limitation of time–space decoupling and the unexplainable bottleneck of traditional models, provides a transparent basis for power dispatching, and has an important reference value for the construction of new power systems. Full article