Power Capability Research Articles

Study of square-wave AC electrophoretic deposition of lithium titanate anodes for Li-ion batteries

Electrophoretic deposition (EPD), while being a scientifically interesting technique due to its inherent versatility, has only limited success in industrial applications. This is partly due to the need to employ expensive and potentially hazardous organic solvents in traditional suspensions. Recently, a significant effort has been focused on the use of aqueous systems and AC-EPD approaches. In this article we present a study and characterization of square-wave electrophoretic deposition at different frequencies of the lithium titanate water-based suspension for application in lithium-ion batteries. We have found that the mass of the deposit linearly increases with time. This reveals that depletion zones are less prone to form in an alternating field, hence enabling formation of homogeneous films and diminishing a growth in bath resistance as compared to DC mode. The EPD performed at critical frequency of 40kHz results in higher content of carbon co-deposited with LTO, improved capacity and power capability and of Li/LTO cells.

A Survey on Liver Cancer Detection Using Hyperfusion of CNN and SVM in Machine Learning

Since liver cancer ranks among of the most aggressive renditions of the disease, improving patient outcomes requires early identification. We propose an inventive tactic to liver cancer detection by integrating CNN and SVM. CNNs, known for their powerful feature extraction capabilities, are particularly effective in analysing complex medical images. SVMs, on the other hand, are efficient classifiers that can separate data points in high-dimensional spaces with accuracy. By merging the feature extraction strength of CNN with the classification efficiency of SVM, the proposed model aims to enhance liver cancer detection accuracy and robustness. The experimental results reveal that the fused CNN-SVM model significantly surpasses the performance of standalone CNN and SVM models, achieving a high detection accuracy of 95.2%. This hybrid method offers a promising direction for improving the precision of computer-aided diagnosis systems, contributing to more effective and reliable liver cancer detection methods that can assist healthcare professionals in making timely decisions.

High‐Isolation Multidimensional Holography Multiplexing in Non‐Interleaved Metasurfaces

AbstractMetasurfaces have revolutionized holography by enabling high‐density channel multiplexing, positioning them as promising candidates for applications in full‐color holographic imaging, optical data storage, and encryption. However, conventional non‐interleaved metasurfaces face limitations in the number of controllable dimensions, restricting the channels available in each dimension. Here, a novel high‐dimensional multiplexing scheme is introduced that significantly enhances both the channel multiplexing capacity and isolation in non‐interleaved metasurfaces. By integrating the Rayleigh‐Sommerfeld diffraction (RSD) formula into the Gerchberg‐Saxton (GS) algorithm and incorporating gradient descent optimization, the approach achieves precise phase profile control across multiple dimensions—extending the controllable parameters to include wavelength, polarization state, and spatial distance. This method enables simultaneous multiplexing of three polarization channels, three wavelength channels, and two focal plane distances, producing a total of 18 distinct, well‐isolated holographic channels with minimal crosstalk. These results showcase the powerful channel multiplexing and isolation capabilities of this non‐interleaved metasurface design, underscoring its potential to advance optical color imaging, secure data storage, and high‐capacity information transmission, thereby contributing to the development of next‐generation intelligent optical devices.

Brain-inspired multimodal motion and fine-grained action recognition

IntroductionTraditional action recognition methods predominantly rely on a single modality, such as vision or motion, which presents significant limitations when dealing with fine-grained action recognition. These methods struggle particularly with video data containing complex combinations of actions and subtle motion variations.MethodsTypically, they depend on handcrafted feature extractors or simple convolutional neural network (CNN) architectures, which makes effective multimodal fusion challenging. This study introduces a novel architecture called FGM-CLIP (Fine-Grained Motion CLIP) to enhance fine-grained action recognition. FGM-CLIP leverages the powerful capabilities of Contrastive Language-Image Pretraining (CLIP), integrating a fine-grained motion encoder and a multimodal fusion layer to achieve precise end-to-end action recognition. By jointly optimizing visual and motion features, the model captures subtle action variations, resulting in higher classification accuracy in complex video data.Results and discussionExperimental results demonstrate that FGM-CLIP significantly outperforms existing methods on multiple fine-grained action recognition datasets. Its multimodal fusion strategy notably improves the model's robustness and accuracy, particularly for videos with intricate action patterns.

Optimal Configuration Strategy of Soft Open Point in Flexible Distribution Network Considering Reactive Power Sources

The intelligent soft open point (SOP) has powerful power flow regulation capabilities in the distribution network. If applied to the distribution network, it can flexibly cope with the output uncertainty of unmanageable distributed energy sources. However, considering the investment, operation, and maintenance costs, as well as the assistance of reactive power equipment, the site selection and capacity determination of SOP have become an urgent problem to be solved. This article proposes the optimal configuration strategy of SOP in a flexible interconnected distribution network, taking into account the features of distributed generation and reactive power sources. Firstly, based on the unpredictability of DG output, this paper uses improved sensitivity analysis to determine the optimal SOP installation location. Subsequently, with the optimization objective of minimizing the annual cost of the distribution network, this paper considers the characteristics of DGs, CBs, and OLCTs and uses second-order cone programming to optimize and solve SOP capacity under constraints such as trends. Finally, in the enhanced IEEE 33-node distribution system model, the effects of different scenarios on node voltage, reactive power components, and SOP location and capacity are compared.

Investigation of a transformer-based hybrid artificial neural networks for climate data prediction and analysis

IntroductionClimate change isone of the major challenges facing the world today, causing frequent extreme weather events that significantly impact human production, life, and the ecological environment. Traditional climate prediction models largely rely on the simulation of physical processes. While they have achieved some success, these models still face issues such as complexity, high computational cost, and insufficient handling of multivariable nonlinear relationships.MethodsIn light of this, this paper proposes a hybrid deep learning model based on Transformer-Convolutional Neural Network (CNN)-Long Short-Term Memory (LSTM) to improve the accuracy of climate predictions. Firstly, the Transformer model is introduced to capture the complex patterns in cimate data time series through its powerful sequence modeling capabilities. Secondly, CNN is utilized to extract local features and capture short-term changes. Lastly, LSTM is adept at handling long-term dependencies, ensuring the model can remember and utilize information over extended time spans.Results and DiscussionExperiments conducted on temperature data from Guangdong Province in China validate the performance of the proposed model. Compared to four different climate prediction decomposition methods, the proposed hybrid model with the Transformer method performs the best. The resuts also show that the Transformer-CNN-LSTM hybrid model outperforms other hybrid models on five evaluation metrics, indicating that the proposed model provides more accurate predictions and more stable fitting results.

Remote Deep-Ultraviolet Laser Ablation in Connection with Electrospray Ionization-Atmospheric Pressure Chemical Ionization (rDUVLAESCI): A Novel Dual Ionization Source for Molecular Mass Spectrometry.

A novel remote deep ultraviolet laser ablation inlet connected to a dual electrospray ionization-atmospheric pressure chemical ionization (rDUVLAESCI) source is presented. This system allows for the simultaneous and spatial acquisition of mass spectrometry (MS) data for organic molecules with diverse polarities and molecular weights. Deep 193 nm UV laser ablation was used to sample analytes from dried spots for molecular MS analysis precisely. Furthermore, molecular MS imaging (MSI) with a variable laser spot size down to 3 μm was demonstrated. The complementary ionization modes generated mass spectra with sufficient analyte signals, detecting a broad range of molecules from polar compounds like caffeine and PEG 600, to nonpolar analytes, such as anthracene and wax esters, all within a single analytical run. Detection limits were found in the tens of attomoles per ablated/desorbed pixel. The powerful capabilities of the fully automated rDUVLAESCI dual source were demonstrated by visualizing the spatial distribution of new psychoactive substances on latent fingerprints. MSI for both sebum components and psychoactive substances revealed a connection between the chemical evidence and biometrical information. The rDUVLAESCI-MSI enabled the unambiguous identification of individuals, even using partially overlapped latent fingerprints. This unique rDUVLAESCI approach, with its remote laser ablation unit, improved spatial resolution and analyte coverage, particularly for nonpolar compounds.

Research on Downhole Two-Phase Flow Parameter Measurement Method Based on Deep Neural Networks

With the rapid development of the oil and gas industry, accurate measurement of downhole two-phase flow parameters has become particularly critical. However, traditional measurement methods face numerous challenges in complex downhole environments and cannot meet the modern industry's demands for high precision and real-time requirements. The flow behavior of downhole two-phase flows (such as oil-gas, water-gas, etc.) directly affects the production efficiency and safety of oil and gas wells. Measuring key parameters (such as flow rate, gas content, and flow pattern, etc.) is crucial for optimizing production processes, improving recovery rates, and ensuring wellbore safety. Traditional measurement methods, such as mechanical flowmeters and capacitance sensors, are easily affected by various factors such as temperature, pressure, and composition changes in complex downhole environments, leading to inaccurate measurement results. In addition, traditional methods usually require a long response time and cannot meet the modern industry's dual demands for real-time and high precision. Under these circumstances, researching a new, high-precision downhole two-phase flow parameter measurement method has become urgent. With the rapid development of artificial intelligence and machine learning technologies, deep neural networks (DNN) offer a potential solution due to their powerful data processing and non-linear modeling capabilities. Applying deep neural networks to the measurement of downhole two-phase flow parameters can not only overcome the limitations of traditional methods but also significantly improve the accuracy and real-time performance of measurements, providing strong technical support for the oil and gas industry. Therefore, this paper proposes a downhole two-phase flow parameter measurement method that integrates deep neural networks, focusing on the research of multi-phase flow parameter measurement technology based on deep learning and its application in the production process of oil and gas wells. By constructing and optimizing models such as Generative Adversarial Networks (GAN), Bidirectional Long Short-Term Memory Networks (BI-LSTM), and Graph Convolutional Networks (GCN), this paper achieves efficient identification and parameter measurement of downhole two-phase flow pattern characteristics. Experimental results show that the integrated model can effectively predict key parameters of downhole two-phase flow under different working conditions, significantly improving measurement accuracy and robustness. This study not only provides a new solution for the measurement of downhole two-phase flow parameters but also has important significance for improving the automation level of measurement technology, reducing manual intervention, and reducing production costs. It further provides new ideas and methods for solving complex fluid flow problems in the energy field, with broad application prospects and profound practical value.

ADA-NAF: Semi-Supervised Anomaly Detection Based on the Neural Attention Forest

In this study, we present a novel model called ADA-NAF (Anomaly Detection Autoencoder with the Neural Attention Forest) for semi-supervised anomaly detection that uniquely integrates the Neural Attention Forest (NAF) architecture which has been developed to combine a random forest classifier with a neural network computing attention weights to aggregate decision tree predictions. The key idea behind ADA-NAF is the incorporation of NAF into an autoencoder structure, where it implements functions of a compressor as well as a reconstructor of input vectors. Our approach introduces several technical advances. First, a proposed end-to-end training methodology over normal data minimizes the reconstruction errors while learning and optimizing neural attention weights to focus on hidden features. Second, a novel encoding mechanism leverages NAF’s hierarchical structure to capture complex data patterns. Third, an adaptive anomaly scoring framework combines the reconstruction errors with the attention-based feature importance. Through extensive experimentation across diverse datasets, ADA-NAF demonstrates superior performance compared to state-of-the-art methods. The model shows particular strength in handling high-dimensional data and capturing subtle anomalies that traditional methods often do not detect. Our results validate the ADA-NAF’s effectiveness and versatility as a robust solution for real-world anomaly detection challenges with promising applications in cybersecurity, industrial monitoring, and healthcare diagnostics. This work advances the field by introducing a novel architecture that combines the interpretability of attention mechanisms with the powerful feature learning capabilities of autoencoders.

An Embodied Intelligence System for Coal Mine Safety Assessment Based on Multi-Level Large Language Models.

Artificial intelligence (AI), particularly through advanced large language model (LLM) technologies, is reshaping coal mine safety assessment methods with its powerful cognitive capabilities. Given the dynamic, multi-source, and heterogeneous characteristics of data in typical mining scenarios, traditional manual assessment methods are limited in their information processing capacity and cost-effectiveness. This study addresses these challenges by proposing an embodied intelligent system for mine safety assessment based on multi-level large language models (LLMs) for multi-source sensor data. The system employs a multi-layer architecture implemented through multiple LLMs, enabling not only rapid and effective processing of multi-source sensor data but also enhanced environmental perception through physical interactions. By leveraging the tool invocation and reasoning capabilities of LLM in conjunction with a coal mine safety knowledge base, the system achieves logical inference, anomalous data detection, and potential safety risk prediction. Furthermore, its memory functionality ensures the learning and utilization of historical experiences, providing a solid foundation for continuous assessment processes. This study established a comprehensive experimental framework integrating numerical simulation, scenario simulation, and real-world testing to evaluate the system through embodied intelligence. Experimental results demonstrate that the system effectively processes sensor data and exhibits rapid, efficient safety assessment capabilities during embodied interactions, offering an innovative solution for coal mine safety.

Mining Silent Biosynthetic Gene Clusters for Natural Products in Filamentous Fungi.

Filamentous fungi are of great interest due to their powerful metabolic capabilities and potentials to produce abundant various secondary metabolites as natural products (NPs), some of which have been developed into pharmaceuticals. Furthermore, high-throughput genome sequencing has revealed tremendous cryptic NPs underexplored. Based on the development of in silico genome mining, various techniques have been introduced to rationally modify filamentous fungi,awakening the silent biosynthetic gene clusters (BGCs) and visualizing the NPs originally cryptic. This review summarizes the key strategies and research advances in the activation of cryptic BGCs in filamentous fungi, including endogenous non-targeted activation, single-target activation, and heterologous expression. It also provides a critical evaluation of these strategies and offers perspectives on the current state of research.

A novel group-based framework for nature-inspired optimization algorithms with adaptive movement behavior

This paper proposes two novel group-based frameworks that can be implemented into almost any nature-inspired optimization algorithm. The proposed Group-Based (GB) and Cross Group-Based (XGB) framework implements a strategy which modifies the attraction and movement behaviors of base nature-inspired optimization algorithms and a mechanism that creates a continuing variance within population groupings, while attempting to maintain levels of computational simplicity that have helped nature-inspired optimization algorithms gain notoriety within the field of feature selection. Through this functionality, the proposed framework seeks to increase search diversity within the population swarm to address issues such as premature convergence, and oscillations within the swarm. The proposed frameworks have shown promising results when implemented into the Bat algorithm (BA), Firefly algorithm (FA), and Particle Swarm Optimization algorithm (PSO), all of which are popular when applied to the field of feature selection, and have been shown to perform well in a variety of domains, gaining notoriety due to their powerful search capabilities.

Remote sensing image destriping with two-stage image decomposition network

ABSTRACT Stripe noise often affects remote sensing images, leading to the degradation of imaging quality and impacting subsequent image processing. Deep learning methods have made remarkable advancements in removing stripes from remote sensing images with their powerful feature extraction capability. However, it is noteworthy that these methods are still insufficient in analysing structural characteristics in the direction of stripes and multi-scale contextual information. To deal with the above issues, we propose a remote sensing image destriping with a two-stage image decomposition network, named TSIDNet, which utilizes the stripe structural characteristics to effectively remove stripe noise while retaining better details. Firstly, stripes are directional, and the structural information of the stripes is mainly concentrated in the image after the differential decomposition of the direction along the stripes. Therefore, a differential decomposition stripe extraction subnetwork (DDSESN) is constructed to generate latent images. Within this subnetwork, we further design a multi-scale cross-fusion residual block (MCRB) and cross-scale fusion attention block (CSFAB) to gradually expand the network’s receptive field, which is conducive to extract and remove stripes more completely. Secondly, the features obtained by adding the latent clear image with details extracted from the stripe’s reverse direction are input into a DWT decomposition detail enhancement subnetwork (DDDESN), which utilizes the discrete wavelet transform (DWT) and the dense residual network for detail enhancement. Demonstrated through experiments, the proposed TSIDNet enables the removal of image stripes while preserving details, and surpassing the comparative methods in qualitative as well as quantitative evaluations. The code will be provided after acceptance.

Hierarchical Transfer Learning with Transformers to Improve Semantic Segmentation in Remote Sensing Land Use

Land use classification remains a significant challenge in remote sensing semantic segmentation. While convolutional neural networks (CNNs) are widely used, their inherent limitations, such as restricted receptive fields, hinder their widespread application in remote sensing. Additionally, the scarcity of labeled remote sensing data and domain shift issues adversely impact deep learning model performance. This study proposes a hierarchical transfer learning framework for fine-category semantic segmentation tasks, leveraging the powerful global relationship modeling capabilities of Transformer models to classify land use in Dongpo District, Meishan City, in mainland China. Our framework represents multilevel transfer learning, progressing from non-remote sensing classification to coarse classification, then to the refined classification of remote sensing. We compared the performance of Transformer models with representative baseline CNNs like U-Net and DeepLab V3+. Results show that the Swin-Unet model outperforms the other models used in this study. It achieved the highest test mean intersection over union (MIoU) of 0.837 and 0.810 for residential and transportation in level 1 (coarse) classification, respectively, and 0.545 for irrigated land in level 2 (fine-grained) classification. Transfer learning from pre-trained models significantly enhanced semantic segmentation accuracy compared to random parameter initialization (ranging from 0.4% to 17.7%), with up to a 17.7% improvement in test MIoU for the public land category. The hierarchical transfer learning framework further improved segmentation accuracy for corresponding level 2 categories, leveraging pre-trained level 1 models. Our study shows the applicability of Transformer-based transfer learning in remote sensing land use classification.

Intelligent Finance: Emerging Applications and Challenges of Machine Learning in Asset Pricing

This paper explores in depth the impact of big data and machine learning on asset pricing. By addressing the limitations of traditional asset pricing models, it analyzes the challenges and opportunities brought by big data to the financial markets. Traditional models, due to their linear assumptions, struggle to handle the complexity of high-dimensional data. However, machine learning techniques, particularly deep learning, overcome this issue with their powerful data processing capabilities, thereby enhancing the models external validation capacity. This paper summarizes various methods of applying modern machine learning in asset pricing, including feature engineering and end-to-end deep learning, and discusses the adaptive mechanisms of AI technologies to the complexity of asset pricing. Through case analysis, the article also emphasizes the significant effects of AI in improving predictive accuracy and market efficiency. This paper provides a new perspective for asset pricing research and offers practical pathways for the intelligence and dynamic adaptability of financial markets.

Analysis and Prediction of Grouting Reinforcement Performance of Broken Rock Considering Joint Morphology Characteristics

In tunnel engineering, joint shear slip caused by external disturbances is a key factor contributing to landslides, instability of surrounding rock masses, and related hazards. Therefore, accurately characterizing the macromechanical properties of joints is essential for ensuring engineering safety. Given the significant influence of rock joint morphology on mechanical behavior, this study employs the frequency spectrum fractal dimension (D) and the frequency domain amplitude integral (Rq) as quantitative descriptors of joint morphology. Using Fourier transform techniques, a reconstruction method is developed to model joints with arbitrary shape characteristics. The numerical model is calibrated through 3D printing and direct shear tests. Systematic parameter analysis validates the selected quantitative indices as effective descriptors of joint morphology. Furthermore, multiple machine learning algorithms are employed to construct a robust predictive model. Machine learning, recognized as a rapidly advancing field, plays a pivotal role in data-driven engineering applications due to its powerful analytical capabilities. In this study, six algorithms—Random Forest (RF), Support Vector Regression (SVR), BP Neural Network, GA-BP Neural Network, Genetic Programming (GP), and ANN-based MCD—are evaluated using 300 samples. The performance of each algorithm is assessed through comparative analysis of their predictive accuracy based on correlation coefficients. The results demonstrate that all six algorithms achieve satisfactory predictive performance. Notably, the Random Forest (RF) algorithm excels in rapid and accurate predictions when handling similar training data, while the ANN-based MCD algorithm consistently delivers stable and precise results across diverse datasets.

Programmable packet-optical network security and monitoring using DPUs with embedded GPUs [Invited

Data processing units (DPUs) with embedded graphics processing units (GPUs) have the potential to revolutionize optical network functionalities at the edge. These advanced units can significantly enhance the performance and capabilities of optical networks by integrating powerful processing capabilities directly at the network edge, where data is generated and consumed. We explore the use cases for DPUs in optical data monitoring with local artificial intelligence (AI) processing and embedded security. This paradigm shift aims to enable more efficient data handling, reduced latency, and improved overall network performance by leveraging local AI processing capabilities embedded within DPUs. In this paper, we show how DPUs can analyze vast amounts of optical data in real-time, implementing advanced data analysis algorithms and security protocols directly on the DPUs to provide robust monitoring and protection for the optical networks. Results indicate that DPUs with embedded GPUs can significantly improve the detection and response times to network anomalies, performance issues, and security threats.

Multi-agent Deep Reinforcement Learning-based Key Generation for Graph Layer Security

Recently, the emergence of Internet of Things (IoT) devices has posed a challenge for securing information and avoiding attacks. Most of the cryptography solutions are based on physical layer security (PLS), whose idea is to fully exploit the properties of wireless channel state information (CSI) for generating symmetric keys between two communication nodes. However, accurate channel estimation is vulnerable for attackers and relies on powerful signal processing capability, which is not suitable for low-power IoT devices. In this paper, we expect to apply graph layer security (GLS) to exploit the common features of physical dynamics detected by IoT sensors placed in networked systems to generate keys for data encryption and decryption, which we believe is a new frontier to security for both industry and academic research. We propose a distributed key generation algorithm based on multi-agent deep reinforcement learning (MADRL) approach, which enables communication nodes to cooperatively generate symmetric keys based on their locally detected physical dynamics (e.g., water/gas/oil/electrical pressure/flow/voltage) with low computational complexity and without information exchange. In order to demonstrate the feasibility, we conduct and evaluate our key generation algorithm in both a simulated and real water distribution network. The experimental results show that the proposed algorithm has considerable performance in terms of randomness, bit agreement rate (BAR), etc.

Application of Convolutional Neural Networks and Recurrent Neural Networks in Food Safety.

This review explores the application of convolutional neural networks (CNNs) and recurrent neural networks (RNNs) in food safety detection and risk prediction. This paper highlights the advantages of CNNs in image processing and feature recognition, as well as the powerful capabilities of RNNs (especially their variant LSTM) in time series data modeling. This paper also makes a comparative analysis in many aspects: Firstly, the advantages and disadvantages of traditional food safety detection and risk prediction methods are compared with deep learning technologies such as CNNs and RNNs. Secondly, the similarities and differences between CNNs and fully connected neural networks in processing image data are analyzed. Furthermore, the advantages and disadvantages of RNNs and traditional statistical modeling methods in processing time series data are discussed. Finally, the application directions of CNNs in food safety detection and RNNs in food safety risk prediction are compared. This paper also discusses combining these deep learning models with technologies such as the Internet of Things (IoT), blockchain, and federated learning to improve the accuracy and efficiency of food safety detection and risk warning. Finally, this paper mentions the limitations of RNNs and CNNs in the field of food safety, as well as the challenges in the interpretability of the model, and suggests the use of interpretable artificial intelligence (XAI) technology to improve the transparency of the model.

Intelligent technologies in traffic flow control: a review

With the acceleration of global urbanization, traffic congestion has become a major bottleneck in the development of modern cities, leading to serious economic losses and a decline in the quality of life. To address this challenge, traffic flow prediction techniques have received much attention. This paper reviews the application of three key technologies, including 5G, artificial intelligence (AI), and vehicular networking (V2X) in intelligent traffic management. 5G technology provides a communication foundation with high speed rates, low latency, and large-scale device connectivity, which effectively supports real-time data transmission and communication needs. AI technology, on the other hand, with its powerful data processing capability, is able to make high-precision traffic predictions in dynamic traffic environments. However, the effectiveness of AI models is highly dependent on the support of high-quality data.V2X technology greatly improves road safety and traffic mobility by realizing real-time information exchange between vehicles and infrastructure. However, relying on a single technology alone cannot comprehensively solve complex transportation problems. For this reason, this paper proposes a technology fusion scheme of 5G, AI, and V2X, aiming to optimize the intelligence level of the traffic management system through the complementary advantages of each technology. The study shows that the synergistic application of the technologies can not only effectively alleviate traffic congestion but also improve the overall safety of roads, providing a feasible solution for the future intelligent transportation system in cities.