Young-Joo Kim’s research while affiliated with Yonsei University and other places

The increasing demand for electricity and the environmental challenges associated with traditional fossil fuel-based power generation have accelerated the global transition to renewable energy sources. While renewable energy offers significant advantages, including low carbon emissions and sustainability, its inherent variability and intermittency create challenges for grid stability and energy management. This study contributes to addressing these challenges by developing an AI-driven power consumption forecasting system. The core of the proposed system is a multi-cluster long short-term memory model (MC-LSTM), which combines k-means clustering with LSTM neural networks to enhance forecasting accuracy. The MC-LSTM model achieved an overall prediction accuracy of 97.93%, enabling dynamic, real-time demand-side energy management. Furthermore, to validate its effectiveness, the system integrates vehicle-to-grid technology and reused energy storage systems as external energy sources. A real-world demonstration was conducted in a commercial building on Jeju Island, where the AI-driven system successfully reduced total energy consumption by 21.3% through optimized peak shaving and load balancing. The proposed system provides a practical framework for enhancing grid stability, optimizing energy distribution, and reducing dependence on centralized power systems.

Accurate and fast characterization of nanostructures using spectroscopic ellipsometry (SE) is required in both industrial and research fields. However, conventional methods used in SE data analysis often face challenges in balancing accuracy and speed, especially for the in situ monitoring on complex nanostructures. Additionally, optical constants are so crucial for accurately predicting structural parameters since SE data were strongly related to them. This study proposes a three-step algorithm designed for fast and accurate extraction of structural parameters from SE measurements. The method utilizes three neural networks, each trained on simulation data, to obtain optical constants and progressively refine the prediction on structural parameters at each step. When tested on both simulation and measurement data on the fabricated 1D SiO2 nanograting specimen, the algorithm demonstrated both high accuracy and fast analysis speed, with average mean absolute error (MAE) of 0.103 nm and analysis speed of 132 ms. Also, the proposed algorithm shows more flexibility in accounting for any change of optical constants to serve as a more efficient solution in the real-time monitoring.

Despite the widespread applications of high-pressure plasma in semiconductor and display industry, such as deposition and ashing process, the use of cutoff probes for diagnosing high-pressure plasma was rarely studied. In this study, we investigated a method of measuring electron density in a cutoff probe using the resonance peak in a high-pressure plasma environment. This method is validated through both electromagnetic wave simulations and experimental methodologies. Our findings reveal that the proposed method demonstrates discrepancies of less than 1.47% compared to the input plasma frequency in the results of electromagnetic wave simulations at a gas pressure of 10 mTorr, while at 2.5 Torr, it exhibited a maximum discrepancy of 13.3% when selecting resonance frequencies lower than the electron–neutral collision frequency. This discrepancy at high pressure is reduced to within 1.92% by selecting a resonance frequency higher than the electron–neutral collision frequency. Also, the feasibility of these electron density measurements has been confirmed under conditions of high gas pressure where the cutoff frequency is not measurable, as evidenced by both simulation and experimental results. Our research on the diagnostic methods in high-pressure plasmas could significantly enhance the measurement and interpretation of plasma parameters in various industrial processes.

In the rapidly advancing realm of mobile technology, under-display camera (UDC) systems have emerged as a promising solution for achieving seamless full-screen displays. Despite their innovative potential, UDC systems face significant challenges, including low light transmittance and pronounced diffraction effects that degrade image quality. This study aims to address these issues by examining degradation phenomena through optical simulation and employing a deep neural network model incorporating hybrid frequency–spatial domain learning. To effectively train the model, we generated a substantial synthetic dataset that virtually simulates the unique image degradation characteristics of UDC systems, utilizing the angular spectrum method for optical simulation. This approach enabled the creation of a diverse and comprehensive dataset of virtual degraded images by accurately replicating the degradation process from pristine images. The augmented virtual data were combined with actual degraded images as training data, compensating for the limitations of real data availability. Through our proposed methods, we achieved a marked improvement in image quality, with the average structural similarity index measure (SSIM) value increasing from 0.8047 to 0.9608 and the peak signal-to-noise ratio (PSNR) improving from 26.383 dB to 36.046 dB on an experimentally degraded image dataset. These results highlight the potential of our integrated optics and AI-based methodology in addressing image restoration challenges within UDC systems and advancing the quality of display technology in smartphones.

Lung adenocarcinoma (LUAD) exhibits significant molecular heterogeneity; however, previous studies have not fully explored its classification into distinct molecular subtypes. Here, we identified LUAD-significant chromosomal instability (CIN) phenotype genes (n = 24) using a TCGA-LUAD cohort (n = 592) and evaluated their ability to predict pathologic grade. Unsupervised clustering and principal component analysis revealed that LUAD patients could be classified into CIN phenotype-related subtypes (GroupLow, GroupModerate, and GroupHigh), each exhibiting distinct transcriptomic patterns. Notably, the GroupHigh showed significantly poor overall survival [OS; hazard ratio (HR) = 1.43, p-value < 10⁻³] and disease-free survival (DFS; HR = 1.27, p-value < 10⁻³). Univariate and multivariate analysis confirmed that its expression status was an independent prognostic predictor (p-value < 10⁻³, HR = 2.18, 95% C.I = 1.26–3.76) of the clinical outcomes, outperforming pathologic grade (p-value < 10⁻³, HR = 1.2, 95% C.I = 1.08–1.33). Moreover, analysis of surfactant metabolism-related genes revealed higher expression in the GroupLow, which was associated with a favorable prognosis. By integrating multiple independent cohorts (n = 779), we validated these findings and confirmed that CIN phenotype gene status serves as a critical prognostic marker in LUAD. Furthermore, genomic profiling showed that the GroupHigh exhibited frequent mutations in key genes such as KEAP1, LYST, SETD2, and TP53, with oncogenes in this group preferentially showing copy number gains. Our study highlights the significance of CIN phenotype gene status as a predictor of LUAD prognosis and its association with transcriptomic and genomic alterations, paving the way for further clinical validation and potential therapeutic interventions.