Yijun Lu

Mr. Yijun Lu

orcid

Research Focus & Metrics

Academic Career Length

2Years (2025-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
21 January 2025 in Electronics
Verified
Journal Article
Open access
PDF file
ZPZhigang Pei
Zhigang Pei
ZZZhiyuan Zhang
Zhiyuan Zhang
JCJiaming Chen
Jiaming Chen
WLWeikang Liu
Weikang Liu
BCBailian Chen
Bailian Chen
YLYijun Lu
Yijun Lu
Electric Vehicle (EV) load forecasting is critical for optimizing resource allocation and ensuring the stability of modern energy systems. However, traditional machine learning models, predominantly based on Multi-Layer Perceptrons (MLPs), encounter substantial challenges in modeling the complex, nonlinear, and dynamic patterns inherent in EV charging data, often leading to overfitting and high computational costs. To overcome these limitations, this study introduces KAN–CNN, a novel hybrid architecture that integrates Kolmogorov–Arnold Networks (KANs) into traditional machine learning frameworks, specifically Convolutional Neural Networks (CNNs). By combining the spatial feature extraction strength of CNNs with the adaptive nonlinearity of KAN, KAN–CNN achieves superior feature representation and modeling flexibility. The key innovations include bottleneck KAN convolutional layers for reducing parameter complexity, Self-Attention Kolmogorov–Arnold Network with Global Nonlinearity (Self-KAGN) Attention to enhance global dependency modeling, and Focal KAGN Modulation for dynamic feature refinement. Furthermore, regularization techniques such as L1/L2 penalties, dropout, and Gaussian noise injection are utilized to enhance the model’s robustness and generalization capability. When applied to EV load forecasting, KAN–CNN demonstrates prediction accuracy comparable to state-of-the-art methods while significantly reducing computational overhead and simplifying parameter tuning. This work bridges the gap between theoretical innovations and practical applications, offering a robust and efficient solution for dynamic energy system challenges.

18 April 2026 in Applied Sciences
Verified
Journal Article
Open access
PDF file
YLYijun Lu
Yijun Lu
YFYingjie Fang
Yingjie Fang
JLJiwu Lu
Jiwu Lu
XYXiang Yuan
Xiang Yuan
Generative artificial intelligence (AI) has demonstrated significant potential in education, yet empirical research on its application in “hardware-driven” interdisciplinary design courses remains scarce. This study employed a randomized crossover experimental design in an IoT Hardware and Design Innovation course at Hunan University. Twelve industrial design undergraduates with no prior IoT background alternated between AI-assisted (ChatGPT-4o) and traditional learning resource conditions across six short-cycle tasks. The crossover design enabled each participant to serve as both experimental and control subjects, yielding 72 observation-level data points. Grounded in Cognitive Load Theory, the study examined three dimensions: process efficacy, difficulty structure, and switching adaptation costs. Results indicated that AI significantly improved perceived task completion efficiency, self-reported goal attainment, and learning experience, yet self-assessed knowledge transfer did not differ significantly between conditions. AI reduced the total number of reported difficulties but altered the difficulty-type distribution: resource-retrieval difficulties decreased while information-verification difficulties increased—a phenomenon we term “difficulty restructuring”. Furthermore, switching from AI back to traditional resources incurred significantly higher adaptation costs than the reverse transition, revealing emerging dependency risks. These findings suggest that generative AI may function more as a “difficulty restructurer” than a “difficulty eliminator” in hardware-driven design education, providing exploratory empirical evidence for incorporating verification literacy into future course design and calling for calibrated scaffold fading that may help mitigate emerging dependency risks.