https://doi.org/10.5370/KIEE.2026.75.11.2353

박현욱(Hyeonuk Park) ; 송성윤(Sungyoon Song)

This paper proposes an ultra-fast frequency deviation prediction model for power systems based on Physics-Informed Neural Networks (PINN) and systematically compares three learning configurations distinguished by the source of physics inputs and label availability: supervised learning with dynamic-simulation inputs, supervised learning with static power-flow inputs and unsupervised learning with static power-flow inputs. The proposed model embeds the swing equation as a physics constraint in the loss function, enabling simultaneous data-driven learning and physical consistency. The supervised model utilizes the mechanical-electrical power imbalance obtainable only through dynamic simulation, while the static power-flow based supervised model employs a generation-load imbalance derivable from static power flow calculations with a learnable scale correction parameter. The unsupervised model attempts to learn solely from the physics loss without labeled data. Experiments on the KPG-193 bus system, a reduced Korean power grid model, use 180,768 samples from 2,152 operating conditions generated via PSSE 33. The supervised model achieves an R-squared of 0.9989 and a plus-minus five percent accuracy of 95.22 percent, while the static power-flow based supervised model attains 0.9973 and 90.97 percent using only static power flow inputs. The unsupervised model minimizes the physics loss yet fails to identify the absolute scale parameter, and five-percent weak supervision reaches only an R-squared of 0.51, confirming that labels are the sole identification signal for the absolute magnitude. Ensemble learning with Monte Carlo (MC) Dropout quantifies the prediction uncertainty as the dispersion of ensemble predictions, which serves as a relative reliability indicator and the basis of the 95 percent confidence intervals. The proposed models achieve prediction times under one millisecond, representing over a thousand-fold speedup compared to conventional dynamic simulations.

https://doi.org/10.5370/KIEE.2026.75.11.2364

김현욱(Hyeon Ook Kim) ; 심재웅(Jae Woong Shim)

This paper compares the operational effects of DC/DC and AC/DC soft open points (SOPs), focusing on the difference between internal DC power redistribution and external AC power support, in an AC/DC hybrid distribution network based on a modified IEEE 33-bus unipolar DC system. The performance is evaluated under normal operation, increased and decreased EV charging demand, and an NREL-based time-varying charging profile using voltage profiles, line losses, voltage deviation index (VDI), load balancing index (LBI), and EV charging curtailment. The results show that the AC/DC SOP provides the strongest DC voltage support and the largest reduction in DC feeder line losses under normal and increased charging conditions by exchanging active power with the external AC grid, while also eliminating charging curtailment under increased charging demand. However, under reduced charging demand, SOP operation can increase line losses and line loading depending on the power-flow direction. These results indicate that SOP operating strategies should account for system loading conditions and power-transfer direction.

https://doi.org/10.5370/KIEE.2026.75.11.2373

정재훈(Jaehun Jung) ; 유연태(Yeuntae Yoo)

Grid-forming inverters (GFMs) can contribute to enhancing the voltage and frequency stability of power systems. However, small-signal instability accompanied by low-frequency oscillations may occur depending on grid strength and control parameters. This paper develops a state-space model and a dq-axis admittance model for a GFM inverter employing virtual synchronous generator (VSG) control. Based on these models, the small-signal stability of the multi-input multi-output (MIMO) system is analyzed using the generalized Nyquist criterion. The accuracy of the developed model is validated by comparing the admittance derived from the state-space model with that extracted from PSCAD simulations. It is also confirmed that the oscillation frequency of the dominant mode calculated from the state-space model agrees with that observed in the PSCAD active power response. In addition, eigenvalue trajectories associated with variations in the short-circuit ratio (SCR) and damping coefficient are analyzed, demonstrating that the stability margin tends to decrease as grid strength increases. To mitigate the deterioration of stability under strong-grid conditions, virtual impedance is applied, and the admittance characteristics, generalized Nyquist plots, and active power responses before and after its application are compared. The results demonstrate that virtual impedance improves the oscillation damping performance and small-signal stability of the GFM-VSG system under strong-grid conditions.

https://doi.org/10.5370/KIEE.2026.75.11.2384

고성혁(Seong-Hyeok Ko) ; 김윤수(Yun-Su Kim) ; 김대진(Dae-Jin Kim)

Virtual power plant (VPP) operation commonly follows a predict-then-optimize framework, in which forecasts of generation, demand, and prices feed downstream optimization. However, minimizing statistical forecast errors overlooks cost asymmetry across error directions and periods, and the nonlinear response of decisions. This has motivated decision-focused learning, which incorporates downstream cost, profit, and risk into training or uncertainty representation. This paper reviews VPP and adjacent studies from three perspectives. Market participation benefits from identifying profit-sensitive periods and variables; resource operation requires attention to physical and time-coupled constraints and operational feasibility; and uncertainty handling improves when forecasts are aligned with downstream outcomes. Forecasts should therefore be evaluated using statistical metrics together with economic, operational, and risk implications. We further identify when such approaches are advantageous.

https://doi.org/10.5370/KIEE.2026.75.11.2393

문정민(Jeong-Min Mun) ; 양현호(Hyun-Ho Yang) ; 김정욱(Jeong-Uk Kim)

This study evaluates the accuracy of model parameter derivation using a phasor measurement unit (PMU)-based permanently installed measurement system by comparing its results with those obtained from conventional test equipment at Muju Pumped-Storage Unit 1 (343 MVA). Seven characterization tests were conducted simultaneously using both systems, and a standard eight-step parameter estimation algorithm was applied independently to each dataset. Raw measurements exhibited minor differences (base field current 0.023%, saturation coefficients up to 0.81%). GENSAE and ESST4B model parameters were identical to three decimal places, indicating that the observed measurement differences of ≤1% did not affect the derived parameters at a single operating point of 71.2 MW (20.76% of rated capacity). PSS2A and HYGOV parameters were validated without re-derivation. A three-level comparison ? measurement error (≤1%), equipment replacement effect (4?12%), and design-field discrepancy (8?50%) ? shows that PMU measurement differences are approximately one order of magnitude smaller than parameter changes requiring re-determination. Within the scope of this single-unit, single-point demonstration, these findings provide preliminary evidence supporting PMU-based test exemption under Article 37 of the Korean Grid Reliability Standards.

https://doi.org/10.5370/KIEE.2026.75.11.2400

원종빈(Jong-Bin Won) ; 조주희(Ju-Hee Cho) ; 김수용(Su-Yong Kim) ; 남스라이자브 어트겅잘갈(NAMSRAIJAV OTGONJARGAL) ; 정성인(Sung-In Jeong)

This paper evaluates the multiphysics characteristics of various stator winding layouts for a dual-winding permanent magnet synchronous motor (PMSM) intended for electric power steering (EPS) applications. Dual-winding PMSMs offer enhanced fault tolerance; however, practical EPS applications require the simultaneous satisfaction of thermal stability, vibration response, packaging, and manufacturability constraints, in addition to electromagnetic performance. In this study, five winding layouts for an EPS motor are compared under identical design conditions in terms of electromagnetic, thermal, structural vibration, and manufacturability characteristics under both normal operating conditions and an open-circuit fault in one winding set. The analysis results show that the winding layout has only a limited effect on performance under normal operating conditions, whereas it becomes a major design factor under open-circuit fault conditions, significantly affecting torque ripple, localized temperature rise, housing vibration, and implementation complexity. Therefore, the design of fault-tolerant EPS motors requires an integrated evaluation that considers multiphysics characteristics and manufacturability rather than relying solely on a single electromagnetic performance metric. The selected layout is subsequently evaluated using a prototype under normal dual-inverter operation and a single-inverter gate-block condition at a 1 N·m operating point. The principal contribution of this work is an integrated winding-selection procedure that links electromagnetic, transient thermal, structural vibration, and manufacturability/packaging results for five existing winding layouts.

https://doi.org/10.5370/KIEE.2026.75.11.2411

신명호(Myoung-Ho Shin)

This paper proposes a method for analyzing the characteristics of an induction generator considering variations in magnetizing inductance when a three-phase squirrel cage induction motor is used as an induction generator. Magnetizing inductance decreases with increasing magnetizing current due to magnetic saturation. In this study, magnetizing inductance is modeled as a quadratic function that decreases as magnetizing current increases. The characteristics of an induction generator are analyzed with changes in slip while taking into account changes in magnetizing inductance. The results are then compared with those obtained using the conventional analysis method, which assumes a constant magnetizing inductance. By comparing the two sets of results, the differences arising from the consideration of magnetizing inductance variation are investigated, and the errors caused by neglecting inductance variation are evaluated.

https://doi.org/10.5370/KIEE.2026.75.11.2417

장어진(Eo-Jin Jang) ; 오현택(Hyeon-Taek Oh) ; 선종석(Jong-Seok Seon) ; 명수현(Su-hyeon Myeong) ; 이진환(Jin-hwan Lee) ; 여한결(Han-Kyeol Yeo)

This paper analyzes the characteristics of a variable flux axial flux permanent magnet synchronous motor according to rotor magnet arrangement. Both models are evaluated using 3D finite-element analysis under ?160Apk demagnetization and +200Apk remagnetization. The series model exhibits a demagnetization rate 10.71 percentage points higher than the hybrid model, whereas the hybrid model achieves remanent flux density and torque recovery rates that are 26.78 and 40.05 percentage points higher, respectively. These results show that the series arrangement offers better demagnetization control, while the hybrid arrangement provides superior remagnetization and electromagnetic performance recovery.

https://doi.org/10.5370/KIEE.2026.75.11.2428

이동헌(Dong-Heon Lee) ; 최영훈(Young-Hoon Choi) ; 윤재중(Jae-Jung Yun)

This paper presents the design, control, and experimental verification of a 6.6-kW three-phase interleaved Totem-Pole power factor correction (PFC) converter for on-board charger (OBC) applications. Three parallel PFC phases are operated with PWM carrier phase shifts of 0°, 120°, and 240° to distribute the input current and reduce the synthesized input-current ripple. SiC MOSFETs are employed for high-frequency switching operation. A digital control system consisting of output-voltage and input-current control loops, a Second-Order Generalized Integrator Phase-Locked Loop (SOGI-PLL), a notch filter, and duty feed-forward was implemented for stable DC-link voltage regulation and input-current control. A 6.6-kW prototype operating with a 220-Vrms input and a 400 V DC-link voltage was developed. Compared with the two-phase interleaved configuration under the same Boost inductance and switching frequency conditions, the three-phase interleaved configuration reduced the maximum input-current ripple from 3.4 A to 2.28 A, corresponding to a reduction of approximately 32.9%, while increasing the ripple frequency from 140 kHz to 210 kHz. The three- phase interleaved configuration exhibited an input-current THD up to 3.871 percentage points lower than that of the two-phase configuration while maintaining a comparable power factor. At rated load, the developed converter achieved an input-current THD of 5.438%, a power factor of 0.996, and an efficiency of 97.978%.

https://doi.org/10.5370/KIEE.2026.75.11.2437

안진형(Jin-Hyung Ahn) ; 김용헌(Yong-Heon Kim) ; 곽상신(Sang-Shin Kwak)

This paper proposes an artificial-intelligence-based diagnostic algorithm that combines residual feature learning with a Ghost module to reliably identify weak AC series arc faults at the output of a three-phase pulse-width-modulated (PWM) inverter. The current waveform and spectral content of an output-side AC series arc vary with the output fundamental frequency, switching frequency, load inductance, and arc intensity. A weak arc produces substantially smaller waveform distortions and spectral changes than a strong arc, making robust detection difficult using conventional signal-processing methods based on a single fixed threshold. In this study, strong and weak arcs are distinguished according to arc power and the electrode-gap opening condition, and the model is evaluated using output AC-current data acquired over a wide range of operating conditions. With fast Fourier transform (FFT) magnitude spectra as the input, the proposed Ghost-ResNet achieved a diagnostic accuracy of 96.89% on the test set, outperforming all comparison models. It also achieved high precision and recall for the AC-arc class, including weak-arc samples. These results demonstrate that the proposed model is effective for detecting output-side AC series arcs in three-phase PWM inverters, including weak arcs and arcs occurring under varying operating conditions.

https://doi.org/10.5370/KIEE.2026.75.11.2445

임종필(Jong-Pil Lim) ; 박민솔(Min-Sol Park) ; 나진용(Jin-Yong Na) ; 박흥석(Hung-Sok Park) ; 정채균(Chae-Kyun Jung)

Submarine cable ampacity is governed by the heat dissipation characteristics of the surrounding seabed. Conventional ampacity calculation methods, such as IEC 60287, evaluate external heat transfer mainly using soil thermal resistivity and ambient temperature, whereas the influence of natural convection in saturated seabed soils has received relatively little attention. Because permeability controls pore-water flow in porous media, it is expected to affect both heat transfer and cable ampacity. In this study, the influence of seabed soil permeability on the ampacity of submarine cables was investigated using a finite element thermo-fluid coupled model based on the Brinkman momentum equation and the energy equation. An AC 154 kV XLPE 3-Core 500 mm² submarine cable buried in saturated seabed soil was analyzed under various permeability conditions, and the numerical results were compared with those obtained using IEC 60287. The analysis showed that ampacity remained nearly unchanged in the low-permeability region, whereas it increased markedly once permeability reached the convection-dominant range, with a maximum value of 864 A. In addition, the dominant environmental factor affecting ampacity changed with permeability. Soil thermal resistivity had a greater influence under low-permeability conditions, while ambient temperature became more influential as natural convection intensified in high-permeability soils. These results indicate that permeability alters the heat-transfer mechanism around submarine cables and should be considered together with soil thermal resistivity and ambient temperature when evaluating cable ampacity in saturated seabed environments.

https://doi.org/10.5370/KIEE.2026.75.11.2453

강동훈(Dong-Hoon Kang) ; 정관훈(Kwanhun Jeong) ; 김인섭(Insup Kim) ; 김연수(Yeon-Su Kim) ; 정태환(Tae-Hwan Jeong) ; 장유나(You-na Jang) ; 임종식(Jong-sik Lim) ; 안달(Dal Ahn)

This paper presents the design, fabrication, and measurement of a 50W-class GaN HEMT power amplifier for X-band radar and satellite communication systems. To overcome single-device output limitations and achieve high integration, a 12-way parallel combining structure based on an Integrated FET process was proposed. The amplifier features integrated 4-way and 12-way divider/combiner stages within the chip to minimize loss. Phase and amplitude imbalances were resolved, achieving <0.4 dB amplitude error and <1° phase error. Alumina and Taconic RF-35 substrates were used for the RF and bias circuits, respectively, to ensure stable DC supply and high isolation. Evaluated at 10 GHz with a 25 us pulse width and 10% duty cycle, the fabricated amplifier demonstrated 47.78 dBm (approx. 60 W) output power, 44.11% PAE, 16.72 dB small-signal gain, and 12.76 dB large-signal gain, satisfying all design goals. This topology is applicable to next-generation high-power RF systems.

https://doi.org/10.5370/KIEE.2026.75.11.2460

정주희(Juhee Jung) ; 임종식(Jongsik Lim) ; 문형준(Hyungzun Mun) ; 조영호(Young-Ho Cho) ; 한상민(Sang-Min Han)

This paper describes the design of a dual-function microwave power divider architecture that allows the phase of the two outputs to be selected as either in-phase or anti-phase. Inserting a 180o phase converting structure into one of output terminals of a normal power divider causes the anti-phase between the two output terminals. To selectively achieve the dual functions, RF SPDT (single-pole double-throw) switches are ultilzed in this study. By selecting the in-phase or anti-phase path with the SPDT switch, the power divider architecture can be designed with either in-phase or anti-phase outputs. The proposed dual-function power divider architecture consists of a normal Wilkinson power divider, two SPDT switches, and a transmission line with a anti-phase converting structure. The validity of the proposed structure is verified by its measured characteristics with the phase errors within ±2o for the both in-phase and anti-phase power splitting operations.

https://doi.org/10.5370/KIEE.2026.75.11.2469

전하윤(Hayoon Jeon) ; 김한솔(Han Sol Kim)

This study focuses on a sampled-data observer design method for a Takagi-Sugeno (T-S) fuzzy system using a sampling-time- dependent approach. The sampled-data observer dynamics are given by combining a conventional observer gain and a novel observer gain that dynamically changes within each sampling interval. To exploit the information on aperiodic sampling intervals, sampling- dependent matrices are incorporated into a Lyapunov-Krasovskii functional (LKF), from which the sufficient conditions for stabilization of the estimation error dynamics and disturbance attenuation are derived in the form of linear matrix inequalities (LMIs). In addition, the mismatch between the system and the observer is regarded as a disturbance and attenuated via an performance criterion, whereas the mismatch between the continuous-time and sampled membership grades of the observer is explicitly characterized by a prescribed deviation bound and incorporated into the stability conditions using slack matrix variables. Finally, a comparative simulation example with a conventional sampled-data observer demonstrates the effectiveness, reduced conservatism, and improved estimation performance of the proposed method.

https://doi.org/10.5370/KIEE.2026.75.11.2475

김민석(Minseok Kim) ; 조영완(Youngwan Cho)

Crowd counting estimates the number of people in an image while reliably representing individual locations. Density map-based methods are robust in dense scenes but cannot directly output individual locations, whereas the point-based P2PNet predicts head locations as point proposals and learns one-to-one assignments with ground-truth (GT) points via Hungarian matching. However, our proposal-level analysis reveals a duplicate proposal problem: multiple proposals can remain around a single GT point even when the counting error is small, so count-level metrics alone cannot evaluate structural prediction quality. The proposed ValueUp-P2PNet preserves the location and confidence prediction structure of P2PNet while predicting a uniqueness score reflecting each proposal's representativeness. Using matching results as supervision, it raises the uniqueness of representative proposals and lowers that of nearby duplicates. A Gaussian-weighted duplicate loss alleviates threshold discontinuity, and an adaptive loss-ratio curriculum adjusts the auxiliary loss contribution with training progress. On ShanghaiTech Part A, UCF-CC-50, and UCF-QNRF, ValueUp-P2PNet achieves comparable or lower MAE than the original P2PNet under the same setting while consistently increasing the matched-proposal ratio (Matched/GT) and reducing the duplicate-proposal ratio (Dup/Pred) on all three datasets. An ablation study shows that each loss term contributes differently to counting performance and to proposal-set structural quality.

https://doi.org/10.5370/KIEE.2026.75.11.2488

서기성(Kisung Seo)

The precipitation and snowfall in the eastern coastal region of the Korean Peninsula are high-risk meteorological phenomena often prone to heavy rain and heavy snow due to local geographic characteristics, making predictive research highly critical. However, most existing meteorological studies are limited to case studies, and the few predictive studies attempted are primarily restricted to classification based on given simple factors. Based on established meteorological findings that easterly winds are closely related to precipitation and snowfall in the eastern coastal region, this paper conducts a study to simultaneously predict precipitation and snowfall using meteorological factors of vertical layers tightly linked to easterly wind phenomena. In this paper, we construct a modified LSTM-based prediction model using ERA5 data that encompasses multiple vertical layers. We propose the application of vertical and horizontal data structures tailored to these specific meteorological characteristics, along with the processing of top-bottom relationships in vertical layer data by utilizing the sequential structure of the LSTM. Additionally, we propose a two-stage LSTM architecture designed to process low-dimensional features and short-term sequential dependencies in the first stage, while better learning high-dimensional long-term dependencies and non-linear patterns in the second stage. As a case study, we evaluated the performance of the proposed model at major locations in the Gangwon and Gyeongsang coastal regions, demonstrating superior performance to LSTM-FC in terms of precision, recall, and accuracy.

https://doi.org/10.5370/KIEE.2026.75.11.2495

김진현(Jin Hyun Kim)

Chain-of-Causation (CoC) explanations from vision?language driving models may omit when an obligation ends. We link obligations to release conditions and condition-specific evidence, distinguish confirmed contradictions from insufficient evidence, and report the first issue event. UPPAAL observers process authored symbolic frames. In 40 legacy pairs, a memoryless checker missed cross-event errors (20/40 versus 40/40). On 144 authored scenarios, a single-obligation checker matched 96/144 expected decisions; after failure analysis, the condition-linked model matched 144/144 on the same cases, tying a finite-state machine. In 81 authored boundary cases with supplied links, retaining state matched all expected decisions; resetting state judged 41 of 44 contradictions compatible. Bounded enumeration of 35,042,928 event sequences, used to reduce circularity from case selection, matched a separate formula evaluator. Four reviewers disagreed on 27 natural scenes (Fleiss’ κ = 0.027); without independent reference labels, natural accuracy is not reported. These results demonstrate feasibility on authored inputs, not vehicle safety.

https://doi.org/10.5370/KIEE.2026.75.11.2504

김진현(Jin Hyun Kim)

Driving VLMs use Chain-of-Causation (CoC) to link scenes to behavior, but CoC may omit when a trajectory is executable. We propose Safety-Constrained CoC, appending a natural-language guard with activation, invariant, bound, prohibition, release, fallback, and termination clauses. Given a scene, CoC, guard, and trajectory candidates, the model selects one candidate; an external deterministic verifier compares its measurements with the guard. Our paired-contract benchmark changes only the guard while fixing the scene, objective, and candidates. In three-seed temporal tests with LoRA-tuned Qwen3-VL-2B-Instruct, the violation rate decreased from 0.319 without a guard to 0.000, while compliant completion and paired accuracy reached 1.000. For unseen time values, candidate accuracy, violation rate, and paired accuracy were 0.635, 0.351, and 0.306. The results support guard-conditioned synthetic candidate selection, not continuous trajectory generation or vehicle-level safety.

https://doi.org/10.5370/KIEE.2026.75.11.2513

변지예(Ji-Ye Byun) ; 최지호(Ji-Ho Choi) ; 박병건(Byeong-Geon Park) ; 황돈하(Don-Ha Hwang) ; 윤성민(Tae-Gyu Kim) ; 김태규(Sung-Min Yoon) ; 김지원(Ji-Won Kim)

A modular multilevel converter (MMC) consists of a large number of submodules, each containing an embedded controller whose firmware needs to be updated as needed. Conventional wired programming methods, such as JTAG-based individual programming, become increasingly inefficient as the number of submodules grows, resulting in increased maintenance time and wiring complexity. In large-scale MMC systems, individually connecting each submodule controller via wired interfaces not only prolongs the update procedure but also increases the risk of human error during maintenance. This paper proposes a Wi-Fi-based wireless firmware download system for MMC submodule controllers based on a dual-core DSP(Digital Signal Processor). The proposed system adopts a hierarchical communication structure consisting of a host PC, a fixed wireless access point (AP) installed inside the converter panel, Wi-Fi modules, and DSP controllers. Firmware images are transmitted wirelessly via TCP/IP, and a group-based sequential download method is applied to distribute network load and ensure communication stability under large-scale deployment. A custom bootloader and flash kernel are implemented on the DSP to perform reliable flash erasing, programming, and verification with checksum-based integrity checking at each stage. The system supports independent firmware download for dual-core CPU1 and CPU2, and provides real-time status monitoring including connection state, firmware version, and download progress through a graphical HMI(Human-Machine Interface). Experimental results confirm that the proposed system enables reliable and efficient wireless firmware updates, significantly reducing maintenance procedures compared to conventional wired methods.

https://doi.org/10.5370/KIEE.2026.75.11.2520

이동엽(Dongyeop Lee) ; 임대식(Daesik Lim) ; 김기현(Kihyun Kim)

This study develops task-specific arc-flash incident-energy analysis and protection criteria using 41 accidents in power facilities from 2018 to 2023. Each case was reconstructed in PSCAD/EMTDC with system impedance, fault type, working distance, equipment geometry, and protective-device characteristics. Incident energy was calculated using IEEE Std 1584-2018 and a high-voltage arc model. The results ranged from 1.37 to 127.1 J/cm² for 14 low-voltage cases and from 12.7 to 613.6 J/cm² for 27 high-voltage cases. Low-voltage metering work was dominated by delayed upstream-fuse clearing, whereas high-voltage transformer, switchgear, VLF, and MOF tasks were affected by long clearing times or short working distances. Only the NFPA 70E incident- energy analysis method was used for PPE selection; the PPE category table method was not applied. PPE arc ratings must equal or exceed residual incident energy after engineering controls, and de-energized work is prioritized when inputs are uncertain or residual energy exceeds the available protection system.

https://doi.org/10.5370/KIEE.2026.75.11.2526

전혜련(Hye-Ryeon Jeon) ; 김호민(Ho-min Kim) ; 김유나(You-Na Kim) ; 최상재(Sangjae Choi) ; 우필성(Pil-Sung Woo)

Large-Scale energy storage systems (ESSs) require quantitative methods for evaluating cell voltage inconsistency using operational data. In a battery rack composed of numerous series-connected cells, differences in state of charge, capacity, internal resistance, and degradation can result in different voltage behaviors even under the same operating conditions. This study proposes a quantitative method for evaluating relative cell voltage inconsistency based on Robust Principal Component Analysis (RPCA), subspace projection, and Mahalanobis Distance (MD). RPCA was applied to decompose the cell voltage matrix into a low rank component representing the common voltage variation shared among cells and a sparse component representing localized deviations. Singular Value Decomposition was then applied to the low-rank component to extract the dominant temporal subspace, and the original voltage trajectory of each cell was projected onto this subspace to obtain representative charging and discharging features. The two features were combined into a two-dimensional feature vector, and MD was calculated to quantify the relative deviation of each cell from the overall cell population. The proposed method was applied to four battery racks, each consisting of 238 cells, over three operating cycles under identical state-of-charge ranges and constant-power conditions. The results showed that most cells exhibited similar MD levels, whereas several cells repeatedly showed relatively high MD values across different cycles. Comparison of their original voltage trajectories confirmed persistent differences in charging and discharging behavior relative to cells with low MD values. In addition, many high-MD cells were located at the same positions within different modules, indicating that structural, measurement, wiring, or balancing effects should also be considered when interpreting the results. Therefore, the proposed MD-based index is more appropriately interpreted as a screening indicator for identifying cells with relatively distinct voltage behavior rather than as a direct indicator of cell degradation or failure.

https://doi.org/10.5370/KIEE.2026.75.11.2532

김유나(You-Na Kim) ; 김호민(Ho-min Kim) ; 전혜련(Hye-Ryeon Jeon) ; 최상재(Sangjae Choi) ; 우필성(Pil-Sung Woo)

This study proposes a digital twin model and simulation system for safety management of energy storage systems (ESS). The system consists of physical, data, platform, and service layers integrating field facilities, an integrated DBMS, digital twin simulation, and services. A field-synchronized digital twin is constructed using LiDAR-based 3D modeling and a battery equivalent circuit model (ECM) trained with operational data. The system is implemented in four stages: representation, monitoring, simulation, and federation. It supports lifetime prediction, fault simulation, grid-connected analysis, and integrated control simulation. These functions provide battery state of health (SOH) estimation, analysis of voltage and temperature faults, assessment of Power Conversion System (PCS) fault impacts, and energy-management strategies. By integrating ESS models, operational data, and simulation functions, the framework enables real-time monitoring and predictive decision support, providing a basis for preventive ESS safety management.

https://doi.org/10.5370/KIEE.2026.75.11.2541

강병준(YByeong-Jun Kang) ; 성동일(Dong-Il Sung) ; 박소현(So-Hyeon Park) ; 서현호(Hyeon-Ho Seo)

Neutral sections in 25 kV AC railways require trains to coast with the main circuit breaker open, which temporarily interrupts traction. Regenerative-energy use also varies with the power-receptive capacity of the feeding section and the upstream grid. The operational and energy effects of a bidirectional continuous feeding system based on a modular multilevel converter (MMC) were quantified using a one-dimensional train-motion model and an energy-balance calculation. A 300 km/h train was evaluated for effective de-energized lengths of 200?800 m and gradients of 0?15‰. At 400 m, speed loss increased from 1.28 km/h on level track to 3.23 km/h at 12‰ and reached 7.50 km/h at 800 m and 15‰. The calculated delay to initial-speed recovery remained below 0.4 s. When regenerative receptivity increased from 0.32 to 0.88, the reduction in normalized net supply energy ranged from 5.8% to 18.6% as the recoverable braking-energy ratio varied from 0.10 to 0.30. The reduction was 12.0% at a ratio of 0.20. Field validation requires synchronized measurements of train operation and feeder power to replace the assumed vehicle parameters and receptivities.

https://doi.org/10.5370/KIEE.2026.75.11.2556

신승권(Seung-Kwon Shin) ; 안승호(Seung-Ho Ahn) ; 이정아(Jung-Ah Lee) ; 손국현(Guk-hyun Son)

Building energy simulation relies on static design-stage inputs and diverges from actual operation after commissioning. This study designs a railway-station digital twin system in five layers, coupling measurements with simulation without re-exporting the model by combining pyenergyplus.api and eppy, and implements it for railway station. The system assembles an EnergyPlus model from a generation script and a parameter catalogue, modifies envelope, equipment and schedules at the object level with eppy, and runs EnergyPlus 26.1.0 in-process, completing an annual simulation of the ten-zone station model in 3.8 to 6.1 seconds. The model was calibrated to monthly metered electricity within ASHRAE Guideline 14 monthly criteria, and CO2 measured by the on-site sensor network was inverted through a mass balance to update ventilation and occupancy inputs. The service layer serves parametric queries via FastAPI, Celery, a result cache, an XGBoost surrogate and Bayesian optimization, and a web dashboard provides monitoring and scenario exploration. The core API and worker processes ran 27 days from 5 August 2026 without restart. Parametric analysis and measurement coupling are achieved in one environment without the re-export required by the earlier FMU approach, while per-timestep measurement injection and closed-loop updating remain at the design stage with interfaces implemented.

https://doi.org/10.5370/KIEE.2026.75.11.2562

전혜위(Hye-Wi Jeon) ; 이동진(Dong-Jin Lee)

In unattended metros, dwell-time adjustment rebalances headways without changing movement authority, but policies evaluated on historical records may use information that is unavailable at the decision time. This study proposes a retrospective evaluation procedure that replays automatic train supervision (ATS) departure events and restricts the target-headway and follower-departure inputs to records available at each decision time, while the arrival of the controlled train is reconstructed retrospectively. The procedure was applied to 212 days of Incheon Line 2 records to compare a no-adjustment baseline, proportional rules, and proximal policy optimization (PPO) policies under identical replay conditions, constraint filters, and train-test splits. The restricted inputs referenced no later record, and the no-adjustment replay reproduced the recorded departures. PPO slightly reduced headway error relative to the baseline but increased delay and lowered plan adherence, and no advantage over the rule was established. Cumulative shifts caused early departures that stepwise constraint checks missed; blocking them preserved the headway reduction. The procedure exposes such trade-offs and verification limits before field testing.

https://doi.org/10.5370/KIEE.2026.75.11.2574

최주희(Juhee Choi)

This paper proposes checkpoint-based state preservation for energy-harvesting-based railway wayside monitoring nodes experiencing intermittent power due to fluctuations in harvested energy. To evaluate the energy and timing costs of checkpointing, experiments were conducted using an MCU with on-chip FRAM under various state sizes and clock frequencies. Based on these results, a constraint-based design formulation is suggested considering required state size, available checkpoint energy, and remaining execution time. An integrated checkpoint feasibility metric is formulated to determine whether each state-size and clock-frequency combination satisfies both energy and timing constraints. The metric enables feasible operating conditions to be identified according to the available energy and time costs. It also provides a systematic way to evaluate the minimum clock frequency that satisfies the checkpoint requirements for a given state size.

https://doi.org/10.5370/KIEE.2026.75.11.2583

우태민(Taemin Woo) ; 김형철(Hyungchul Kim) ; 박종영(Jong-Young Park) ; 김재원(Jaewon Kim) ; 정호성(Hosung Jung)

This paper presents a technique in which a learning model, after training directly on data, discriminates the fault zone and estimates the fault location in the branch section of an AC power supply system. However, the conventional method has a limitation in that its performance degrades as the fault point approaches the branch point. The proposed technique feeds a multi-decision-tree model with the current distribution ratio ? corrected for measurement-magnification error ? together with a direction vector defined relative to the branch point, allowing the model to determine the fault location on its own. The training and evaluation dataset was constructed by interpolating fault data simulated with PSCAD/EMTDC. Compared with the conventional method, the proposed technique improved the discrimination accuracy by 18.9 percentage points and the average location error by 13m within 50m of the branch point, and by 23.2 percentage points and 32m, respectively, in the 50?100m range, showing substantial performance gains in both regions.

https://doi.org/10.5370/KIEE.2026.75.11.2594

김상암(Sang-Ahm Kim) ; 김경희(Kyung-hee Kim) ; 송은주(Eun-Ju Song)

Railway accidents are low-frequency, high-consequence events. Railway operators must balance safety investment costs and accident risk under limited budgets and resources. This study formulates railway accident risk as a function of occurrence frequency and consequence severity to identify optimal combinations of safety measures. A Bow-Tie model is developed to construct accident scenarios and evaluate outcomes resulting from combinations of system successes and failures. The optimization considers various safety measures, including automatic train control (ATC), and the presence or absence of a driver. The results show that an optimal combination of safety measures can minimize costs while satisfying the allowable risk constraint.

https://doi.org/10.5370/KIEE.2026.75.11.2606

임현빈(Hyun-Bin Lim) ; 홍석민(Seok-Min Hong) ; 오효석(Hyo-Seok Oh) ; 이태현(Tae-Hyun Lee) ; 황지원(Ji-Won Hwang) ; 임재원(Jae-Won Lim) ; 정노건(No-Geon Jung)

As humanoid robot applications expand, there is a growing need for test environments that enable repeated and reproducible evaluation of walking performance. This study establishes minimum design requirements for an evaluation treadmill applicable to various humanoid robots. Public specifications of nine humanoid robots, including mass, payload, walking speed, and operating time, were analyzed to determine the required test load, belt speed, and operating time. Based on these requirements, the drive system and frame were designed, and finite element analysis was conducted under static and dynamic loading conditions. The results showed that the frame stresses remained below the material yield strength, with a maximum deflection of approximately 0.036 mm, confirming structural integrity. The proposed requirements provide a practical reference for the design of evaluation treadmills for humanoid robots.

https://doi.org/10.5370/KIEE.2026.75.11.2616

김정권(Jung Kwon Kim) ; 이우진(Woo Jin Lee)

Metering Out Fit (MOF), though essential for accurate energy billing, is customer-owned and often poorly maintained, raising the risk of undetected low-voltage faults that undermine metering reliability. This study proposes an integrated AI framework to detect such latent anomalies using large-scale data from advanced metering infrastructure. To overcome the scarcity of labeled fault data, an unsupervised phase first identified statistical outliers and established a diagnostic baseline, used to build a refined labeled dataset for training supervised ensemble classifiers, including Random Forest, LightGBM, and XGBoost. XGBoost achieved the most robust performance, balancing precision and recall. Field validation on selected cases showed that predicted voltage transformation ratio errors were consistent with actual physical degradation, suggesting the framework's potential as a scalable, automated solution for asset management, though large-scale field validation remains for future work.

https://doi.org/10.5370/KIEE.2026.75.11.2627

구자영(Ja-Young Koo) ; 곽준호(Jun-Ho Kwak) ; 김희동(Hee-Dong Kim)

When a high-voltage (HV) induction motor is disassembled for stator winding maintenance, moisture ingress frequently occurs on the winding surfaces due to rapid ambient humidity changes. Although operational practices involve hot-air drying followed by insulation resistance and diagnostic tests up to the rated AC voltage, a detailed comparative analysis regarding the degradation stage is required. This paper presents a comparative analysis of insulation diagnostic parameters for two distinct stator windings under both dry and moisture-absorbed conditions : an aged 18-year-old HV motor (336 kW, 4.0 kV) and a newly manufactured HV motor (1,000 kW, 6.6 kV). AC current, dissipation factor (tanδ), capacitance, and partial discharge (PD) magnitudes were measured with increasing applied voltage. The experimental results reveal that while both motors exhibit elevated values under moisture-absorbed conditions compared to their dry baselines, the aged motor shows significantly higher sensitivity and distinct PD phase patterns. These findings demonstrate that the combined analysis of tanδ and PD characteristics can effectively distinguish between transient environmental moisture absorption and permanent insulation degradation, providing critical indicators for HV motor asset management.

https://doi.org/10.5370/KIEE.2026.75.11.2633

윤준수(Junsu Yoon) ; 심재훈(Jaehoon Shim)

Real-time state-of-health (SOH) estimation requires reliable performance with low computational burden. Bayesian model averaging (BMA) provides point estimates and predictive uncertainty, but multiple health indicators and candidate models increase computational cost in resource-constrained embedded systems. In this study, an existing BMA-based SOH estimation method was implemented on a Nucleo-H755ZI-Q board, and the effects of input dimensionality reduction on SOH estimation error, predictive uncertainty, and embedded execution time were analyzed. Nine input-dimensionality conditions were constructed by sequentially excluding health indicators with lower absolute Pearson correlation coefficients. When six input features were used, the mean execution time was 7.266 ms, approximately 19.9 % lower than the 9.070 ms obtained with nine input features, while the RMSE remained comparable to that of the seven-input condition at 0.030. A further reduction from six to five input features shortened the mean execution time by only approximately 1.5 %, whereas the RMSE increased by 46.7 % to 0.044. When four or fewer input features were used, the execution time remained near 6.6 ms, and no substantial additional reduction was observed. These results indicate that six inputs provide a balanced trade-off between SOH estimation performance and execution time under the considered dataset, feature-removal order, and embedded hardware conditions, rather than a generally optimal input dimension.

https://doi.org/10.5370/KIEE.2026.75.11.2643

조나림(Na-Rim Jo) ; 김원호(Won-Ho Kim)

This paper presents the design and experimental validation of a slotless axial flux permanent magnet motor with printed circuit board (PCB) windings, targeting 40 W output and a 30 % reduction in active volume. Pole count, coil count, and number of turns were evaluated under PCB manufacturing constraints. End turn dimensions, overhang configurations, and PCB layer count were examined to balance output power, losses, conductor content, and magnetic air gap length. Corner rounding and multiple vias were applied to lap and wave winding patterns to reduce resistance. The final design reduced the active volume from 57.4 to 40.1 cm³. A prototype was fabricated, and measured phase resistance and no-load back electromotive force were compared with simulations. Load tests confirmed approximately 0.24 N·m torque and over 40 W output. These results provide guidance for compact PCB-winding motors considering manufacturing constraints and electromagnetic characteristics.

https://doi.org/10.5370/KIEE.2026.75.11.2652

고승훈(Seung-Hoon Ko) ; 김원호(Won-Ho Kim)

This paper presents a design process for a 4 kW class interior permanent magnet synchronous motor (IPMSM) for semiconductor vacuum pumps. Based on a conventional induction motor, inverter voltage, current density, and slot fill factor limits are established. Four pole-slot combinations (6/9, 8/12, 10/15, and 12/18) are evaluated at a fixed pole-to-slot ratio of 2:3. For each combination, the electric-to-magnetic loading ratio is adjusted to minimize current density. The stator outer diameter and stack length are then systematically adjusted within the design constraints to determine minimum-volume candidates. Considering electromagnetic performance and inverter limitations, the 8-pole/12-slot design is selected and refined through rotor and stator geometry optimization to reduce cogging torque and torque ripple. Finite element analysis shows a shaft output of 4.46 kW and an efficiency of 96.42 %, with a 60.1 % reduction in stator-envelope volume relative to the reference induction motor.

https://doi.org/10.5370/KIEE.2026.75.11.2660

유광택(Kwang Taig Yu) ; 정범진(Beom Jin Chung)

As services that require high-resolution metering data?time-of-use (TOU) tariffs, residential demand response (DR), virtual power plants (VPP), and municipal carbon monitoring?expand, a reliable data operator for the behind-the-meter (BTM) area of apartment complexes has become essential. This paper proposes an integrated energy data platform based on a metering point operator (MPO) that collects multi-utility data (electricity, water, gas, hot water, and heating) of existing apartment complexes at 15-minute intervals. The platform adopts a two-tier transport security architecture?Transport Layer Security (TLS) between the apartment server (A-EMS) and the central server (MIS), and Virtual Private Network (VPN) tunneling between the MIS and external entities such as the KEPCO meter data management system (MDMS), municipalities, and DR aggregators?together with a pseudonymization roadmap from universally unique identifier (UUID) tokens to keyed hashing (HMAC-SHA-256). The institutional position of the MPO is examined with reference to the Smart Energy Code (SEC) of Great Britain and the Korean Smart Energy Code (K-SEC) under preparation. A one-month field verification across 33 apartment complexes (40,170 households) yielded an end-to-end data integrity ratio of 98.86%, with stage-wise losses of 0.47% (in-complex collection), 0.20% (TLS transmission), and 0.47% (validation). Participation in the 2025 Seoul residential DR program produced settlement revenue of KRW 4.8 million for 10,497 households, confirming that the data-service revenue structure operates in practice. The results provide a practical reference for institutionalizing the MPO role and BTM data utilization under K-SEC.