With the large-scale grid integration of new energy sources, the complexity of power quality disturbances (PQDs) continues to increase. Traditional time-frequency analysis methods suffer from issues such as modal aliasing and limited resolution when processing non-stationary signals. This paper proposes and investigates a multi-disturbance classification and identification method for power quality that integrates Variational Modal Decomposition (VMD) with the Hilbert Transform (HT). VMD adaptively decomposes complex signals into a series of finite-bandwidth Intrinsic Mode Functions (IMFs), effectively addressing the iterative screening limitations of Empirical Mode Decomposition (EMD). Subsequently, HT extracts the instantaneous amplitude and frequency characteristics of each IMF to construct a high-precision time-frequency model. MATLAB simulation experiments demonstrate that this method achieves extremely high recognition accuracy and noise robustness under various typical and complex disturbance scenarios, such as voltage dips and swells, providing critical technical support for the stable operation of new energy power grids.

An ethylene cracking system comprises multiple parallel furnaces operating in inherent cracking–decoking cycles, making the scheduling of cyclic furnace operations a key determinant of plant profitability. Meanwhile, exogenous fluctuations in product prices driven by market supply–demand conditions and production costs significantly affect the economic performance of scheduling decisions. Therefore, this paper develops a new two-stage Wasserstein distributionally robust scheduling framework. The Wasserstein metric is employed to construct the data-driven ambiguity set that describes all possible probability distributions of uncertainty. Afterwards, the kernel density estimation algorithm is utilized to design the data-driven support set, further mitigating the adverse effects of outliers. By introducing the ambiguity set and support set, a tractable mixed-integer nonlinear programming model is re-derived based on the duality theory. A practical case study shows that the proposed method can effectively hedge against price uncertainty, and the obtained robust scheduling yields the daily profit of 8.639×106 RMB/day.

This article presents a visualization of the (theta, E) manifold that governs P - Q coupling in grid-forming converters (GFM). A simplified large-signal model is derived, using the internal voltage magnitude E and phase angle theta as state variables. The model clarifies the coupling among P , Q , E , and theta , and shows how the thresholds of current and voltage limiters determine both the existence and accessibility of stable equilibrium points. By tracing the operating point trajectories on the E - theta manifold, the P - Q coupling surface reveals the mechanism that leads to transient loss of synchronization during abrupt phase jumps. The analysis further explains how limiters modify the dynamic critical phase angle that separates secure and insecure operating regions, and highlights the dual role of small-signal based P - Q decoupling: it increases the admissible phase jump margin but can prolong post-disturbance recovery (the disturbance considered throughout is a permanent grid voltage phase jump) when limiters are active.

Visual-tactile fusion has emerged as an essential paradigm in pattern recognition, particularly in robotic perception systems where physical interaction, partial observability, and gripper-induced occlusion pose significant challenges, providing complementary sensory cues for accurate and robust object understanding under complex perception conditions. This paper proposes the Wavelet-Enhanced Visual-Tactile Transformer (Wave-VTformer), a unified framework that integrates generative augmentation, multiscale feature encoding, and cross-modal attention alignment for discriminative multimodal representation learning. Specifically, a Visual Generative Adversarial Network (ViGAN) is employed to augment small-scale datasets with high-fidelity and class-consistent synthetic samples. A discrete wavelet decomposition module is incorporated to capture localized structural patterns while attenuating redundant frequency components, thus improving representation compactness. Furthermore, an interactive cross-modal attention decoder is designed to align visual and tactile embeddings in a semantically consistent latent space, enabling effective information exchange across modalities. Extensive experiments on the ObjectFolder 2.0 benchmark demonstrate that Wave-VTformer achieves a Top-1 accuracy of 99.8 %, significantly outperforming state-of-the-art fusion models. Comprehensive ablation and robustness analyses confirm that the joint effect of generative augmentation, wavelet-based encoding, and attention-guided fusion is critical to the model's superior generalization capability. These results highlight the potential of Wave-VTformer as a principled and extensible framework for robotic multimodal perception and fine-grained object recognition.

The fusion of visual and tactile information is essential for enhancing robotic perception, object recognition, and interaction in unstructured environments. However, existing methods often fail to fully leverage the complementary strengths of visual and tactile modalities due to significant heterogeneity and spatial misalignment. This paper proposes a novel framework called Visual-Tactile Graph Convolutional Network (VTGCN), which is based on Graph Convolutional Networks (GCNs) and facilitates effective multimodal fusion through graph-structured representation learning. Specifically, modality-specific subgraphs are constructed by extracting keypoint-based visual nodes using ORB and CNNs, and grid-structured tactile nodes through convolutional feature mapping and spatial partitioning. Intra-modal edges are formed via K-Nearest Neighbors (KNN) for vision and local grid adjacency for touch. To bridge the modality gap, an adaptive cross-modal edge construction strategy is introduced, which dynamically selects inter-modal connections based on feature similarity. The resulting fusion graph supports end-to-end representation learning using stacked GCN layers, enabling accurate and robust recognition in both simulated and real-world scenarios. Experiments on five visual-tactile datasets—including Touch and Go, ObjectFolder 2.0, ObjectFolder Real, MSDO, and a self-collected VTDaily dataset—demonstrate that VTGCN achieves an accuracy of up to 99.59%, outperforming existing CNN and Transformer-based models. The proposed approach effectively models structural heterogeneity and deep inter-modal dependencies, providing a scalable solution for visual-tactile fusion. VTGCN shows promising potential in robotic grasping, human-computer interaction, and augmented reality applications, where precise perception and adaptive multimodal integration are critical.

In the intricate operations of mining rods, precise behavior recognition is paramount for operational safety. Addressing target detection and posture feature extraction challenges, this study proposes a method that integrates attention mechanisms with a Spatial-Temporal Graph Convolutional Network. An efficient channel attention mechanism is introduced during target detection, allocating weights to each channel to adapt to diverse features accurately. Multihead attention modules are incorporated in posture feature extraction, effectively capturing critical behavioral information. Behavior classification is achieved through the SoftMax function. Experimental results demonstrate the method's accuracy of 95.3% and a recall rate of 91.6% on the custom mining dataset. On the NTU-RGB+D public dataset, the method significantly improves accuracy and recognition speed. This research provides an innovative approach to behavior recognition in complex environments, ensuring precise identification of various behaviors in real-world scenarios, safeguarding worker safety, and holding crucial implications for applying behavior recognition technology in industrial fields.

Compared to offline simulation, real-time simulation can more accurately and effectively replicate the dynamic behavior of inverters. This paper presents the development and design of an FPGA-based real-time inverter simulator. The inverter is modeled using the binary LC method, and the model accuracy is improved by applying the enhanced binary LC method and the trapezoidal integration method. A multi-channel synchronized DAC based on FPGA is designed for the real-time simulator, achieving an average input-output signal delay of 45 μs. Unlike automatic model code generation techniques, the inverter’s detailed FPGA model is directly implemented using Verilog Hardware Description Language (HDL). Taking an LC-filtered three-phase inverter as an example, simulations are conducted to verify the accuracy of the designed real-time simulator.

In the pursuit of higher levels of intelligence, embodied intelligences need to integrate information from multiple perceptual channels through a multimodal information fusion mechanism to comprehensively understand the surrounding scene and the manipulated objects. Most of the current multimodal perception research focuses on the fusion of vision-based, and most of them involve only the fusion of two modalities, while the fusion of three or more modalities is seldom explored. Therefore, we propose TVT-Transformer (TVT: Tactile-Visual-Textual): a new framework for joint learning from three modal data: tactile, visual and semantic text. The approach utilizes Attention mechanism to deeply mine and align information from different perceptual modalities, allowing the model to efficiently integrate features from tactile, visual, and semantic text data, enabling a deeper level of information fusion through cross-modal interaction. The framework introduces a novel semantic representation approach to generate standardized semantic descriptions by combining human observation and touch interactions with objects. Next, the semantic descriptions are encoded using a pre-trained Bert model, aligning them with visual and tactile information. Following this, the Query (Q), Key (K), and Value (V) components from tactile, visual, and textual modalities are integrated into a unified Q, K, and V. Subsequently, the Attention mechanism within the Transformer architecture is employed for cross-attention computation, enabling more accurate and efficient cross-modal feature integration and understanding. The textual modality provides semantic support that enhances the effectiveness of information integration and improves the accuracy of object recognition. With tactile sensor data, visual image data, and corresponding semantic text descriptions as input data, the method has been validated for its effectiveness and superiority on both publicly available and self-made datasets, enhances the feature expression capability, achieves significant performance enhancement in multimodal data integration, and improves the accuracy of object recognition. Compared with the classical Transformer, TVT-Transformer is able to effectively fuse visual, tactile, and semantic textual information through a cross-modal Self-Attention mechanism, and exhibits greater adaptability and robustness in processing multimodal information. This study not only provides a new perspective in the field of multimodal information fusion, but also provides a strong technical support for the development of embodied intelligence. The TVT-Transformer framework has a wide range of potential applications, and it is expected to play an important role in the future in the fields of intelligent robotics, human-robot interaction, and assisted decision-making. The resource has been released at https://github.com/huakaichengbei/MSDO/tree/master.

This chapter explores the problem of decoupling distribution networks through an improved distributed parametric long-line decoupling method. The traditional long-line decoupling method has certain limitations when applied to the distribution network, mainly due to the short-line characteristics in the distribution network, which cannot fully utilize the transmission delay decoupling. In response to this problem, this chapter proposes an improved line decoupling method. This improved method introduces compensation capacitance to increase the wave propagation time without changing the line length, thereby realizing the decoupling of lines in the distribution network. And by building a model for simulation analysis, the effectiveness of the method is verified, providing new ideas and methods for solving the decoupling problem of distribution networks.

This paper presents an examination of grid-forming converters (GFM) under low-voltage-ride-through (LVRT) conditions. It emphasizes the influence of inner loop control strategies, and grid topologies on GFM performance. The study introduces a versatile equivalent modeling methodology suitable for different inner loop control strategies. Additionally, it evaluates three LVRT testing devices: the shunt impedance based voltages aggenerator (SIVSG), programmable voltage source with impedance (PVS), and Hardware-in-the-Loop (HIL) system, highlighting their differences in simulating grid fault characteristics. The results indicate that SIVSG and PVS have limitations in mimicking actual grid fault scenarios, potentially leading to GFMs erroneously passing LVRT tests. Conversely, HIL-based testing methods more accurately replicate grid faults, offering a more reliable assessment of GFM performance.

To address the safety risks posed by short-circuit faults in power systems and accurately capture the dynamic characteristics of electrical quantities under fault conditions, this study constructs a classical IEEE 9-bus standard model based on the RTDS real-time digital simulation platform to conduct fault simulation experiments. The model includes three generators, nine buses, three transformers, and six transmission lines, with node types comprising slack buses, voltage-controlled buses, and load buses, reflecting the core components of a real power system. The experiment sets a three-phase short-circuit-to-ground fault at bus8 for 100 ms, focusing on observing the instantaneous waveforms of three-phase voltage and current at the fault point as well as buses 4, 7, and 9. By quantifying fault propagation patterns, the study verifies the accuracy of RTDS simulation, providing data support for the design of power system fault protection schemes and equipment selection.

The underground drilling environment in coal mines is critical and prone to accidents, with common accident types including rib spalling, roof falling, and others. High-quality datasets are essential for developing and validating artificial intelligence (AI) algorithms in coal mine safety monitoring and automation field. Currently, there is no comprehensive benchmark dataset for coal mine industrial scenarios, limiting the research progress of AI algorithms in this industry. For the first time, this study constructed a benchmark dataset (DsDPM 66) specifically for underground coal mine drilling operations, containing 105,096 images obtained from surveillance videos of multiple drilling operation scenes. The dataset has been manually annotated to support computer vision tasks such as object detection and pose estimation. In addition, this study conducted extensive benchmarking experiments on this dataset, applying various advanced AI algorithms including but not limited to YOLOv8 and DETR. The results indicate the proposed dataset highlights areas for improvement in algorithmic models and fills the data gap in the coal mining, providing valuable resources for developing coal mine safety monitoring.

This paper proposes a design scheme and experimental verification based on the communication between the FPGA development board and the industrial computer, which aims to realize the efficient data exchange between the industrial computer and the FPGA. The design uses Xilinx's XDMA IP core, realizes software-level construction and hardware-level on-board testing in the Vivado environment through Verilog language, and then combines the industrial computer with the XDMA driver and the corresponding code written in the Linux system environment for actual testing, realizing high-speed data transmission between the FPGA and the industrial computer. Experimental results show that the design achieves a delay of 100 microseconds in practical applications, which meets the stringent requirements of industrial automation systems for real-time performance. This research provides an effective solution for FPGA-based low-latency communication systems.

Aiming at the shortcomings of traditional power grid simulators in terms of output waveform quality and dynamic response performance, a five-level modular multi-level converter grid simulator based on the nearest level approximation modulation strategy is designed and implemented. The system adopts the MMC cascaded by the half-bridge sub-module as the core power transformation topology, and the key component parameters such as sub-module capacitance and bridge arm inductance are determined through rigorous parameter design. In order to suppress the high-frequency switching harmonics, an LCL filter with a resonant target of 10 kHz is designed, and a stepup transformer is set to achieve a high-voltage AC output of 3600 V. This paper elaborates on the principle of nearest level approximation modulation and its implementation in five-level MMC. The simulation results show that the 10 kW power grid simulator can operate stably under various working conditions such as normal, voltage drop, frequency suddenness, harmonic injection and asymmetric fault, with high output waveform quality and rapid dynamic response, which verifies the effectiveness and superiority of the five-level MMC topology based on the nearest level approximation modulation in power grid simulation applications.

Multimodal information fusion research struggles with aligning heterogeneous modalities and addressing data imbalance, especially when integrating visual, haptic, and text-three modalities offering complementary perceptual and semantic features. Current research focuses on Transformers for unimodal and vision-haptics bimodal tasks, neglecting tri-modal integration. Leveraging text's semantic bridging capacity could address this limitation in cross-sensory learning. We propose VHTformer, a Transformer-based framework designed to unify visual, haptic, and textual modalities via joint query learning. The model leverages hierarchical attention mechanisms: self-attention refines intra-modal features (e.g., extracting texture from haptic signals or contextual semantics from text). Meanwhile, cross-attention aligns spatial-semantic patterns across modalities through learnable joint queries. This enables synergistic fusion of geometric shapes (vision), material properties (haptics), and descriptive attributes (text). Experiments were conducted on three multimodal datasets-ObjectFolder 2.0, Touch and Go, and ObjectFolder Real-covering a total of 100 + object categories with diverse material and shape properties. To mitigate class imbalance and ensure statistical reliability, we adopted stratified 5-fold crossvalidation. In addition, we conducted robustness evaluations under Gaussian noise injection to verify the model's robustness. VHTformer achieves up to 99.55 % recognition accuracy and demonstrates strong robustness, highlighting the value of tri-modal integration for comprehensive object understanding.

T-type neutral clamp (TNPC) three-level inverters have gained widespread attention in medium to high power applications due to their excellent efficiency and performance. However, its complex switching state and real-time control strategy pose challenges to the computational efficiency and accuracy of the simulation model. In this paper, a high-precision real-time simulation method combining algorithmic modeling (ABM) and average switching modeling (ATMT) is proposed. By constructing switch state coding and interlocking logic, the precise abstraction of the switching behavior of TNPC inverter is realized. In the open-loop experiment, the system compares three methods: pure SimPower Systems modeling, ABM+ATMT modeling and ABM+SimPower hybrid modeling, and the results show that the total harmonic distortion (THD) of the output voltage is less than 3.1% in the whole working condition range of modulation ratio 0.6 to 1.0, which is significantly better than the comparison method, while maintaining excellent fundamental amplitude reproduction accuracy. In the closed-loop experiment, the VF control strategy is used to verify that the proposed ABM model and the physical Simscape model have a high degree of consistency between the power grid voltage and the inverter current waveform (fundamental amplitude deviation <0.1%, THD deviation <0.03%). This study provides an effective solution for efficient and high-precision real-time simulation of three-level inverters.

In real-time simulation of power electronic system, the modeling method and numerical integration method are key factors affecting simulation performance. The LC binary method significantly reduces hardware resource utilization and FPGA workload. However, different integration methods have varying error accuracy and convergence. This paper provides a detailed analysis of the impact of different numerical integration methods on the system and proposes an effective and versatile real-time simulation method for multi-switch circuit, demonstrated through the application case of a single-phase inverter based on LC filtering. Comparative tests with Simulink model under a real-time simulation step size of 1μs show that the proposed method achieves a relative error of less than ± 1

In some cases, we need to collect data in real time and need to meet the requirements of multi-channel data acquisition, so based on this, a multi-channel data acquisition system based on Lab Viewis developed. Provide a short introduction regarding PCI-1706U hardware along with LabView software. The PCI1706 data acquisition card was selected as the system hardware, and the framework structure of the software program was established. The system's feasibility and stability were validated through the utilization of the LabView DAQ program of PCI-1706U for gathering eight signals from the signal generator. Meanwhile, The experiments involving multi-channel synchronous acquisition and delayed start acquisition were finished. Experiments show that the system platform realizes the functions of continuous and accurate acquisition, filtering, display, and storage.

Abstract In the context of application in a power grid simulator, we investigated the parameter design issues of several crucial components of the Modular Multilevel Converter (MMC) topology applied to the power grid simulator system. These components include the bridge arm inductance and submodule capacitance. Based on the fundamental circuit of the MMC applied to the power grid simulator system, we introduced an equal discharge time constant H. We analyzed the capacitor voltage fluctuation rate and the equivalent capacitance discharge time constant “H”, discovering that they still exhibit an inversely proportional relationship across different engineering scenarios. Consequently, a general method for determining submodule capacitance parameters was proposed, and in conjunction with previous engineering parameters, recommended values for an equal discharge time constant of 40 ms were provided. Furthermore, we introduced the concepts of double-frequency circulating resonant angular frequency and phase resonant angular frequency, outlining the principles for determining the bridge arm reactance values. Through the analysis of corresponding parameters in previous engineering studies, universally applicable recommended values for the bridge arm reactance parameters were proposed, considering the unit resonant angular frequency as the power grid frequency angular frequency 1.0ω.

Through the application of deep learning and multi-sensor data, fault features can be automatically extracted and valuable information can be integrated to tackle intricate challenges in motor bearing fault diagnosis. Most existing fusion models focus primarily on the original time series signal with information extraction largely restricted to the time domain (without extensions into multiple transformation domains). Also, in most fusion models, the sensor fusion level is kept relatively simple which could lead to the oversight of correlations and complementarities among the information. To enhance the recognition capability of diagnostic network features, in this paper, we propose a novel framework for motor bearing fault diagnosis from the perspectives of multi-transformation domain and multi-source data fusion. Within this framework, feature extraction and fusion from various source data are achieved in the time domain, frequency domain, and time–frequency domain. Distinct independent networks are set up within these domains: one network is designated for overseeing feature fusion, while the others are dedicated to extracting features from individual sensors. To support the extraction of pivotal features across multiple fusion layers in various transformation domains, several fusion nodes are inserted between the layers of the multiple feature extraction networks and the feature summarization network. Furthermore, a channel attention mechanism is introduced as a fusion strategy that serves to pinpoint the significance of different features, thus enhancing the efficiency of feature extraction. Experimental evaluation reveals the efficacy of the proposed model and highlights its noteworthy performance attributes such as scalability and universality.