This study presents TL-MC-ShuffleNetV2, a lightweight and transferable fault diagnosis framework designed for elevator guideway vibration analysis. To tackle challenges such as limited labeled data and the constraints of real-time deployment, the approach integrates Variational Mode Decomposition (VMD) for multi-scale signal separation and employs a customized 1D ShuffleNetV2 backbone with multi-channel (MC) inputs. Squeeze-and-Excitation (SE) attention modules are embedded throughout the network to enhance channel-wise feature sensitivity. A transfer learning (TL) strategy is adopted, in which the model is initially trained using the Case Western Reserve University (CWRU) bearing dataset and subsequently adapted to the elevator domain by freezing early convolutional layers while fine-tuning higher-level layers. Evaluation results demonstrate that the proposed framework achieves a classification accuracy of 96.4%, alongside significantly reduced inference time and parameter complexity. Comparative and ablation experiments further validate the individual contributions of VMD preprocessing, SE modules, and transfer learning to model performance. Overall, the method exhibits strong adaptability, computational efficiency, and suitability for deployment in smart elevator monitoring systems under Industry 4.0 environments.
Achieving deep decarbonization while maintaining power system operational feasibility represents a critical challenge in regions with a high penetration of renewable energy. In this study, we developed an integrated LEAP–NEMO modeling framework to evaluate long-term decarbonization pathways and power system flexibility in Guangxi Province, China, from 2022 to 2060. Four scenarios were designed, the Business-as-Usual Scenario (BAS), the Policy Strengthening Scenario (PSS), the Technology-Centered Scenario (TCS), and the Integrated Optimization Scenario (IOS), to assess the impacts of policy, technology, and system-level coordination. Our results show that under the IOS, renewable electricity reaches approximately 78% by 2050 and further increases to 88% by 2060, while coal consumption declines by over 60% compared with the baseline. Energy storage deployment under the IOS reaches 25 GW by 2060 (rising to approximately 28 GW under a higher-resolution 48-representative-day validation), approximately five times the level in the BAS (5 GW), significantly enhancing system flexibility and reducing renewable curtailment from 10.2% to 1.8%. Carbon emissions peak in 2027 at 228 Mt CO2, approximately 20% lower than the baseline, and decline to 52 Mt CO2 by 2060, achieving a cumulative reduction of 3.85 Gt CO2. Economically, the IOS pathway reduces total system costs to USD 815 billion, saving approximately USD 95 billion compared with the BAS, with the levelized cost of electricity declining to USD 63/MWh by 2060. These results demonstrate that integrated optimization of generation and storage resources is essential in achieving deep decarbonization while maintaining system deterministic capacity adequacy and economic efficiency.
The rapid expansion of photovoltaic power generation increases the difficulty of maintaining stable performance, because traditional operation and maintenance methods cannot handle the efficiency loss caused by array faults. To improve the intelligent operation and energy performance of photovoltaic stations, this paper proposes a collaborative scheme integrating improved visual detection with intelligent optimization. An enhanced YOLOv7 model, incorporating a Hierarchical Attention Fusion Module and a Decoupled Detection Head, is designed to strengthen fault feature extraction. Furthermore, a dynamic reconfiguration method based on a Grow Optimizer with parameter self-adaptive characteristics is proposed to optimize array energy efficiency. Experimental results demonstrate that the proposed model achieves a detection accuracy of 97.45% in noise-free scenarios and maintains a robustness of over 84.75% under noise interference. Meanwhile, the reconfiguration algorithm achieves a maximum power increase of 78.99% and a system conversion efficiency of 97.24% under open-circuit, short-circuit, and complex shading conditions. These results prove that the method improves fault recognition accuracy and power generation efficiency. It also builds a reliable intelligent operation loop through multi-technology coordination and offers effective support for reducing cost and improving performance in photovoltaic stations.
The purpose of this research is to study and conduct an experimental test for investigating the increase of efficiency of a solar panel when cooling it with a backside cooling system. There were 3 experiment test-rigs by using 3 different refrigerants: water, ice, and dry ice. By containing a type of refrigerant into its identical box that was designed by wrapping insulation around the box while drilling the holes for the box that did not contain any refrigerant. Then covered the box with a solar panel. To determine which refrigerant could be the best helped increase the efficiency of a solar panel. The research team conducted the tests with 2 time periods: a short time for 1 hour operation when collecting experimental results for every 2 minutes and a long time for 6 hours when collecting experimental results every 10 minutes. It is found from the experimental tests that significantly different capacity in power generated for different types of refrigerants could be obviously noticed even for the short time. For the long-time test, ice was the best refrigerant that continuously enhanced the efficiency for the solar panel with 25.2 watts compared with 24.0 watts for no refrigerant cooling, which was 6.25% increase.
28 February 2025 in Edelweiss Applied Science and Technology
People all throughout the world enjoy tea. The traditional detection of tea diseases relies heavily on farm experts, which often takes a lot of time. Computer vision technology and artificial intelligence can automatically detect diseases and control their spread in tea leaves in real time. This study suggested a deep learning-based version of the YOLOv5 algorithm. It then improved the model structure of YOLOv5 and integrated Omni Dimensional Dynamic Convolution (ODConv) and Convolutional Block Attention Module (CBAM) attention mechanisms into YOLOv5. The findings demonstrate that the enhanced YOLOv5's accuracy is better than that of previous approaches and 5.06% greater than it was prior to the enhancement. Using the improved YOLOv5 detection algorithm to achieve a high-precision model for identifying tea pests and diseases and combining it with the Pyside6 library to design an interface recognition system, the development of the target detection recognition page is completed. In conclusion, this work offers a useful deep learning-based technique for the quick and precise diagnosis of tea disorders in the area of autonomous tea disease detection.
28 February 2025 in Edelweiss Applied Science and Technology
Face detection is a crucial aspect of computer vision, often challenged by factors such as varying scales, occlusions, and diverse facial features. In this study, we introduce GCSEM-YOLO, an innovative real-time face detection method built upon the YOLOv8 architecture. This approach incorporates a novel feature extraction module (GCSEM) alongside a specialized small-scale detection head, designed to capture pixel information across multiple levels and enhance the receptive field, thereby improving the accuracy of small face detection. To address the imbalance between easy and difficult samples, we employ an adaptive anchor box filtering algorithm coupled with the new WIoUv3 loss function. Experimental results demonstrate that GCSEM-YOLO achieves outstanding efficiency on the WIDER FACE validation datasets.