Search Results (274)

Reconfigurable battery modules can dynamically adjust the connection topology among battery cells, significantly improving the energy utilization efficiency of battery energy storage systems. However, existing state estimation methods focus primarily on individual battery cells. Frequent topology changes cause traditional State of Charge (SOC) estimation algorithms to accumulate large errors due to mismatches in equivalent capacity and internal resistance, making them ineffective for reconfigurable battery modules. To address this limitation, this paper proposes a Gated Recurrent Unit–Transformer architecture for precise SOC estimation in reconfigurable battery modules. The model uses a Gated Recurrent Unit to capture the temporal continuity of electrochemical evolution and employs the Transformer’s self-attention mechanism to analyze discrete topology changes. Experimental results show excellent estimation accuracy across different initial SOC levels, with a mean absolute error as low as 0.085% at full charge and 0.035% at 50% SOC. The architecture demonstrates strong topology self-identification capability and maintains high robustness even under abrupt voltage steps caused by reconfiguration. This method provides accurate and reliable state estimation for large-scale two-layer reconfigurable battery systems while reducing control complexity and improving operational efficiency. Full article

Accurately capturing spatiotemporal dependencies and enabling effective decision support are core challenges in Intelligent Transportation Systems (ITS). Existing research often treats traffic prediction and path planning as isolated tasks. Moreover, mainstream prediction models struggle with long-term periodic patterns, while Reinforcement Learning (RL)-based planning often suffers from inefficient exploration in sparse topologies. To address these issues, this paper proposes a unified framework combining a spectral-temporal Graph Neural Network (GNN) and bi-directional decoding RL. Specifically, a time-frequency dual-stream adaptive learning module is introduced for prediction. Fast Fourier Transform (FFT) and Gated Recurrent Unit (GRU) are employed to capture global frequency periodicities and local temporal dynamics, respectively. Their adaptive fusion effectively mitigates the long-sequence information forgetting problem. For path planning, the task is formulated as sequence generation. A graph-aware attention encoder with adjacency masking is designed, and heuristic feature embeddings are incorporated to guide efficient exploration. Furthermore, a bi-directional autoregressive decoding strategy enhances robustness against topological bottlenecks. On PEMSD4 and PEMSD8, the proposed predictor achieves MAE/RMSE/MAPE values of 18.211/30.433/12.006 and 13.587/23.566/8.955, respectively. Path-planning simulations on the PEMSD4-derived sparse topology further demonstrate stable bi-directional RL optimization, faster convergence with heuristic guidance, and a sparsity-aware encoder that reduces redundant attention interactions in sparse road networks. These results validate the effectiveness of the proposed “predict-then-plan” paradigm. Full article

Social media sentiment analysis demands systems that are simultaneously accurate, scalable, and interpretable. Lexicon-based methods offer transparency but ignore context, while pre-trained language models (PLMs) capture contextual semantics yet encode sentiment only implicitly. Existing integration strategies inject lexicon signals at the input, attention, or feature layer—all outside the recurrent gating mechanism that controls how affective evidence accumulates over a sequence. We propose the SAGU-Net, a framework built around the Sentiment-Aware Gated Unit (SAGU), a gated recurrent unit (GRU) variant with a dedicated sentiment gate conditioned on external lexicon signals. A complementary Context-Adaptive Sentiment Scoring (CASS) module transforms static polarity scalars into context-dependent vectors via learned projections over PLM representations, bridging the gap between discrete lexicon scores and continuous embeddings. The sentiment gate activations provide token-level explainability without post hoc attribution. On a 12,700-sample Chinese social media corpus of intellectual property co-branding reviews (Fleiss’ $\kappa =0.82$) and two public benchmarks, the SAGU-Net achieves 93.62% accuracy and 93.21% Macro-F1, outperforming nine baselines and matching or exceeding LoRA-fine-tuned large language models (GPT-5, Claude Sonnet 4.6, DeepSeek V3.2, Qwen3.5) while requiring three to four orders of magnitude fewer parameters. Ablation confirms the sentiment gate as the single most impactful component. Full article

Forecasting scrap steel prices is crucial for the economic sustainability of recycling operations, yet it remains challenging due to inherent volatility and non-stationary behavior. In this study, we develop and evaluate a delta-targeted Hybrid forecasting pipeline for short horizons of 1, 3, and 7 days. We benchmark classical baselines (Naive, Seasonal Autoregressive Integrated Moving Average (SARIMA), and Exponential Smoothing (ETS)) against recurrent deep learning models (Simple Recurrent Neural Network (RNN), Gated Recurrent Unit (GRU), and Long Short-Term Memory (LSTM)) and recent neural forecasting baselines, including Decomposition-Linear (DLinear), Convolutional Kolmogorov–Arnold Network (C-KAN), and Neural Basis Expansion Analysis for Time Series (N-BEATS), using real-world daily scrap steel price data. The results indicate that delta-targeting generally yields more stable predictive performance than direct raw-price forecasting as the prediction horizon increases. For example, at the 7-day horizon, the predictive fit improves from approximately $R^2\approx 0.87$ for raw-price LSTM to around $R^2\approx 0.90$ for delta-trained recurrent models. At the same horizon, a delta-based RNN achieves the lowest Mean Absolute Percentage Error (MAPE) among the evaluated models (approximately $1.39\%$), while the proposed Hybrid model remains competitive across all tested horizons and maintains a goodness-of-fit of approximately $R^2\approx 0.90$ without uniformly minimizing point error relative to the best-performing recurrent baseline. Attention profiling and permutation-based feature importance analyses indicate that the model places relatively higher weight on calendar-related inputs, consistent with the presence of weekly patterns in the data; these results should be interpreted as sensitivity diagnostics rather than causal evidence. Overall, the findings suggest that delta-transformed targets provide a more suitable prediction space than raw-price targets for short-horizon scrap steel forecasting, while the Hybrid design offers a balanced combination of predictive performance and diagnostic interpretability for operational decision support. Full article

To address the drawbacks of traditional convolutional neural network-based rolling bearing fault diagnosis techniques, including poor feature extraction, low diagnostic accuracy, and poor generalization capability, a dual-channel rolling bearing fault diagnosis model based on VMD-BiGRU and GADF-ResNet-CBAM was proposed. Variational mode decomposition (VMD) was used to first break down and reconstruct the original vibration signal. The rebuilt signal was then input into a bidirectional gated recurrent unit (BiGRU) network in order to extract temporal information. Second, the Gramian angular difference field (GADF) transformed the one-dimensional vibration signal into a two-dimensional picture. This image was then fed into a residual network that was merged with the convolutional block attention module (CBAM) in order to extract spatial characteristics. After concatenating and fusing the data from the two channels, Softmax was finally employed at the output layer to classify different types of faults. The Case Western Reserve University (CWRU) bearing dataset and a self-collected independent dataset from the Xinjiang University experimental rig were utilized for validation. The model achieved diagnosis accuracies of 99.39% and 99.58%, respectively. These results demonstrate the robustness and practical applicability of the proposed method on data acquired from distinct hardware sources and experimental environments, outperforming alternative approaches. Full article

Lithium-ion battery thermal runaway has become a key safety hazard restricting the development of electric vehicles. Early precursor signals of thermal runaway are characterized by multi-scale features, weak signal strength and spatial coupling, posing significant challenges for traditional methods in achieving accurate early warning. To solve this problem, a multi-scale spatiotemporal attention network (MSTA-Net) is proposed for battery thermal runaway early warning. First, a systematic feature engineering process is designed, including signal denoising, normalization processing and multi-level feature construction, to fully extract discriminative information from voltage and temperature signals. Then, the MSTA-Net architecture is constructed, which includes three parallel feature extraction branches: local fine perception branch based on 1D depthwise separable convolution to capture transient anomalies, a temporal evolution modeling branch based on bidirectional gated recurrent units to learn long-term trends, and a global spatial dependence branch based on a graph attention network to model the spatial propagation of thermal runaway. Finally, an adaptive fusion gate is designed to dynamically fuse the features of each branch according to the input context. The experimental results on the self-built battery thermal runaway dataset show that the proposed MSTA-Net achieves a recall rate of 98.7%, an average early warning time of 115 s and a false alarm rate of 0 times/h. Compared with traditional machine learning and deep learning models such as Random Forest, LSTM and Transformer, the model has significant advantages in early warning accuracy, timeliness and robustness. Ablation experiments verify the effectiveness of each component of the MSTA-Net. The proposed method can provide reliable early warning of thermal runaway only by using the existing voltage and temperature sensors of the battery management system, which has important engineering application value. Full article

Solving geometry problems is a critical challenge in education, for it demands the integration of textual semantic descriptions, mathematical formula logic and spatial graphical information, as well as rigorous geometric theorem application and stepwise logical deduction. These are core capabilities that underpin the realization of personalized intelligent tutoring and efficient educational resource allocation. Traditional geometry problem solving methods often suffer from deficiencies in accuracy and the fusion of text, formula and image features. Hence, this paper proposes a method of solving geometry problems based on a text–formula–image (TFI) multimodal parsing and fusion model. The TFI parser employs a self-attention multilayer Transformer to enhance the extraction of logical relations among geometric text expressions. Meanwhile, it parses formulas into tree structures to overcome the loss of formula structural features, which utilizes symbolic embedding and tree-structured encoding to preserve hierarchical logical information and yields unified formula representations via a multi-granularity fusion module. The TFI parser also leverages a Feature Pyramid Network (FPN) for the accurate detection of geometric and non-geometric instances, resolves the issues of blurred segmentation for slender geometric elements and the inaccurate localization of small-sized symbols through mask averaging and RoIAlign, and generates high-dimensional image features using DenseNet-121. The TFI multimodal fusion model integrates a contrastive learning mechanism and constructs fused feature representations by stacking self-attention and cross-attention layers. This design effectively narrows the semantic gap between text, formula, and image features, addressing the inadequacy of traditional fusion approaches in deep cross-modal feature alignment. An attention-augmented Gated Recurrent Unit (GRU) network processes the fused TFI features to produce target operation trees and geometry solutions, ensuring interpretable and precise reasoning performance. The proposed method is evaluated on the PGDP5K and GeoEval datasets, and it achieves an average accuracy of 59.63% in geometry problem solving, which validates its effectiveness. This paradigm offers a viable technical approach for uniformly modeling complex educational tasks, including geometry problem solving and timetable scheduling. Full article

Rapid and accurate gas identification and concentration prediction are of critical importance for industrial safety, medical diagnostics, and environmental monitoring. However, signal distortion in complex environments and feature loss during data processing often degrade prediction accuracy and response speed. To address these challenges, this study proposes a dual-model framework for electronic nose systems. A gas classification model transforms time-series sensor data into two-dimensional feature maps using a composite Gramian Angular Field representation and end-to-end classification using a convolutional neural network (CNN). A gas concentration prediction model integrates a multi-branch attention mechanism, a CNN, and a bidirectional gated recurrent unit to capture spatial–temporal dependencies. A cascaded identification–prediction scheme is further developed to mitigate data distribution heterogeneity and enhance model robustness. The proposed method supports both single-label and multi-label tasks and exhibits strong adaptability under complex conditions, including low concentrations, varying humidity, and gas mixtures. Validation on public and laboratory-collected datasets demonstrates that, using only initial response-stage data, the classification model achieves 100% identification accuracy, while the prediction model attains R² > 0.99 for the majority of target gases. These results confirm that the proposed framework provides an efficient and robust solution for rapid qualitative identification and quantitative prediction in electronic nose systems. Full article

Wearable sensor-based human activity recognition (HAR) has gained growing significance in healthcare monitoring and assisted living systems. Although considerable advances have been made in classifying dynamic movements, stationary activities—such as sleeping, driving, and watching TV—remain difficult to distinguish owing to their weak sensor signatures and limited discriminative cues. This paper presents the multi-scale hybrid attention temporal network (MHAT-Net), a deep learning framework whose key architectural novelty lies in the parallel (non-sequential) dual-pathway temporal modeling: a BiGRU branch and a transformer encoder branch operate simultaneously on the same spatially encoded representation, combined via a learnable attention-based fusion module. This design targets the underexplored problem of distinguishing stationary activities from weak inertial sensor signatures. The architecture is built upon three integrated components: (1) a multi-branch CNN with kernel sizes three, five, and seven combined with channel attention for adaptive spatial feature extraction across multiple temporal scales; (2) parallel bidirectional gated recurrent unit (BiGRU) and transformer encoder pathways for jointly capturing short-range sequential patterns and long-range temporal correlations; and (3) an attention-driven fusion module that adaptively weights the outputs of both temporal branches. The model was assessed on a publicly available benchmark comprising three motionless activity categories collected from 25 participants via smartphone sensors. In 5-fold cross-validation, MHAT-Net attained 97.42% (±4.69%) accuracy with accelerometer data and 92.31% (±0.31%) with gyroscope data, substantially exceeding the accuracies of five baseline architectures: CNN, LSTM, BiLSTM, GRU, and BiGRU. Ablation experiments identified multi-scale spatial feature extraction as the most influential module (2.21–2.47% contribution), followed by the hybrid temporal modeling components. Cross-modality analysis confirmed that accelerometer signals yielded richer discriminative content for stationary activities, while MHAT-Net sustained consistent performance across both sensor types. The proposed integration of multi-scale spatial encoding, hybrid temporal modeling, and multi-level attention gives MHAT-Net the ability to reliably detect subtle activity-specific patterns, establishing a new benchmark in wearable sensor-based recognition for comprehensive daily behavior monitoring. Full article

With the rapid growth of the global shipping industry and the increasing availability of Automatic Identification System (AIS) data, accurate vessel trajectory prediction has become crucial for ensuring navigational safety and optimizing maritime traffic management. This paper presents a systematic review of recent advances in deep learning-based methods for vessel trajectory prediction. We provide a comprehensive analysis of mainstream models, including Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs) such as Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) networks, Sequence-to-Sequence (Seq2Seq) models, and the Transformer architecture. Their performance is compared in terms of spatio-temporal data processing capability, prediction accuracy, and computational efficiency. Furthermore, this review examines practical applications of these methods in scenarios such as collision avoidance and route optimization. Despite notable progress, several challenges remain, including data quality issues, real-time prediction capability, and model interpretability. Future research directions may focus on multi-source data fusion and the development of lightweight model designs to further improve prediction performance. This survey aims to serve as a valuable reference for researchers and contribute to ongoing innovation in vessel trajectory prediction technology. Full article

Unmanned aerial vehicles (UAVs) operating in safety-critical missions require effective anomaly detection methods to identify propulsion-system faults before they cause catastrophic failures. However, current vibration-based diagnostic models typically rely on datasets representing only discrete, isolated fault states, and do not capture the continuous structural degradation that occurs during real flight operations. To address this gap, this study proposes a severity-ordered vibration data augmentation framework for anomaly detection in rotary-wing UAV propulsion systems. Controlled experiments were conducted under healthy, tape-induced imbalance, scratch, and cut propeller conditions using stepped throttle excitation from 10% to 100% in 10% increments, with 40 s per level. A severity-ordered arrangement strategy based on throttle level and a robust peak-to-peak severity metric generated approximately 7.5 h of augmented vibration data per axis, representing a continuous degradation trajectory. Three-axis continuous wavelet transform (CWT) scalograms of size $48\times 96\times 3$ were used to train an unsupervised anomaly detection framework. Comparative experiments with Isolation Forest, One-Class SVM, and LSTM–AE demonstrated that the proposed Convolutional Neural Network (CNN)–Bidirectional Gated Recurrent Unit (BiGRU)–State-Space Model (SSM)–Autoencoder (AE) architecture achieved the best performance, reaching 0.9959 precision, 0.4428 recall, 0.6131 F1-score, and 0.9284 Area Under the Receiver Operating Characteristic Curve (AUROC). The ablation study further showed that incorporating temporal modeling and state-space dynamics improves detection robustness compared with CNN–AE and CNN–BiGRU–AE baselines. These results show that combining severity-ordered augmentation with deep temporal learning improves progressive propulsion anomaly detection in UAV vibration monitoring. This work introduces a methodology that connects rotor dynamics principles with deep learning, providing a continuous degradation manifold that improves early-stage detection and condition monitoring of UAV propulsion systems. Full article

Accurate photovoltaic (PV) power forecasting is essential for the efficient operation of residential energy systems and microgrids, as reliable short-term predictions enable improved energy scheduling, demand management, and operational planning in distributed energy environments. In this study, one-hour-ahead forecasting of residential PV power generation is investigated using real-world data collected from multiple households within an Irish energy community. Several deep-learning architectures, including long short-term memory (LSTM), gated recurrent unit (GRU), convolutional neural networks (CNN), CNN–LSTM hybrid networks, and attention-based LSTM models, are evaluated and compared with a seasonal autoregressive integrated moving average (SARIMA) statistical model. A sliding-window approach is employed to transform the PV time series into a supervised learning problem. To ensure statistical robustness, deep-learning models are evaluated using a multi-run framework, and results are reported as mean ± standard deviation based on MAE, RMSE, MAPE, and R² metrics across multiple households. The results indicate that deep-learning models achieve consistently strong forecasting performance, with GRU frequently providing the most reliable predictions across several households. For instance, in House 5, GRU achieved an RMSE of 142.02 ± 1.87 W and an R² of 0.694 ± 0.008, while in Houses 11 and 13 it attained R² values of 0.837 ± 0.002 and 0.835 0.08, respectively. However, performance varied across households, reflecting the influence of data variability and generation patterns on model effectiveness. In comparison, the SARIMA model demonstrated competitive performance and, in certain cases, outperformed deep-learning models. For example, in House 4, it achieved the lowest RMSE of 90.68 W and the highest R² of 0.709. Overall, these findings highlight that while deep-learning models offer greater adaptability and stability, statistical models remain effective for more regular PV generation patterns. Consequently, the study emphasizes the importance of evaluating forecasting models under realistic household-level conditions and demonstrates that both deep-learning and statistical approaches can provide short-term PV forecasting. Full article

Stability and sustainability are the operational bottom lines of modern power grids. However, the inherent volatility and non-stationarity of wind energy, particularly in complex terrains, severely threaten power grid stability. To address this challenge, we propose an end-to-end architecture named the Adaptive Multi-scale Routing Wind Power forecasting (AMR-Wind) framework. The framework is principally composed of three sequential modules: an Adaptive Frequency Disentanglement Module (AFDM), an inverted Transformer (iTransformer), and a Scale-Routing Gated Recurrent Unit (SRGRU). The AFDM utilizes a differentiable filter bank to dynamically disentangle complex spectral signatures and mitigate mode mixing. The iTransformer is employed to effectively capture the complex multivariate dependencies between these disentangled modes and exogenous meteorological features. The SRGRU utilizes hierarchical temporal routing to synchronize localized high-frequency ramp events with macroscopic evolutionary trends. Comprehensive evaluations across four diverse wind farms demonstrate that AMR-Wind reduces the RMSE by an average of 8.4% and improves the $R^2$ by at least 1.0% compared to state-of-the-art baselines. Ablation studies further confirm the modules’ strong synergistic effects, yielding a 7.6% reduction in forecasting errors. This framework reduces the error in wind energy prediction, providing a reliable tool for the stability and sustainability of the power grid. Full article

Accurate demand forecasting and adaptive load balancing are critical for maintaining stability and efficiency in modern smart power grids. This study proposes a hybrid deep learning (DL) framework, termed Transformer-Generative Adversarial Network-Gated Recurrent Unit (Transformer-GAN-GRU), which integrates global attention-based temporal modeling, generative data augmentation, and sequential refinement into a unified architecture. The proposed framework captures both long- and short-term dependencies while improving representation of imbalanced demand patterns. The model is evaluated on three heterogeneous benchmark datasets, namely Pecan Street, the reliability test system-grid modernization laboratory consortium (RTS-GMLC), and the reference energy disaggregation dataset (REDD). Experimental results demonstrate that the proposed model consistently outperforms state-of-the-art baselines, achieving a maximum accuracy (Acc) of 99.49%, a recall of 99.67%, and an area under the curve (AUC) of 99.83%. In addition to high predictive performance, the framework exhibits strong stability, fast convergence, and low inference latency, confirming its suitability for real-time deployment in smart grid environments. Full article

Surface electromyography (sEMG) enables non-invasive measurement of muscle activity for applications such as human–machine interaction, rehabilitation, and prosthesis control. However, high noise levels, inter-subject variability, and the complex nature of muscle activation hinder robust gesture classification. This study proposes a multistream hybrid deep-learning architecture for the FORS-EMG dataset to address these challenges. The model integrates Temporal Convolutional Networks (TCN), depthwise separable convolutions, bidirectional Long Short-Term Memory (LSTM)–Gated Recurrent Unit (GRU) layers, and a Transformer encoder to capture complementary temporal and spectral patterns, and an ArcFace-based classifier to enhance class separability. We evaluate the approach under three protocols: subject-wise, random split without augmentation, and random split with augmentation. In the augmented random-split setting, the model attains 96.4% accuracy, surpassing previously reported values. In the subject-wise setting, accuracy is 74%, revealing limited cross-user generalization. The results demonstrate the method’s high performance and highlight the impact of data-partition strategies for real-world sEMG-based gesture recognition. Full article