Mobile QR Code QR CODE : The Transactions of the Korean Institute of Electrical Engineers

탐라해상풍력발전단지가 제주 전력계통에 기여한 경제성 및 환경성 평가에 관한 연구 A Study on the Evaluation of the Economics and Environmental Contribution to the Jeju Power System from Tamla Offshore Wind Farm

https://doi.org/10.5370/KIEE.2020.69.7.955

정세민(Semin Jeong) ; 오웅진(Ungjin Oh) ; 이연찬(Yeonchan Lee) ; 임진택(Jintaek Lim) ; 이기백(Kibaek Lee) ; 최재석(Jaeseok Choi)

This paper proposes an evaluation methodology of the economics and environmental contribution of Tamla offshore wind farm which is owned by KOEN(Korea South-East Power Co, Ltd.). The probabilistic simulation system called “PRAWIN,” which has already been developed by this research team, is used in this study. An alternative new methodology in order to evaluate contribution of WTGs(Wind Turbine Generators) in view point of probabilistic production cost (economics) and CO2 emissions (environment) is developed in this paper. Furthermore, the wind speed contribution functions in view point of probabilistic production cost (economics) and CO2 emissions (environment) are newly formulated and proposed in this paper. The proposed contribution functions describe variation coefficients (sensitivities) of the economics (probabilistic production cost) and environment (CO2 emissions) according to changing wind speed of a wind farm. Using the new contribution functions proposed in this paper, Jeju power system sized model case study including Tamla offshore wind farm is demonstrated. The total production cost, average production cost unit and CO2 emissions of Tamla offshore wind farm are simulated and assessed in the case study. Using the proposed contribution functions, also, the contributions of wind speed of Tamla offshore wind farm in view point of economics (probabilistic production cost) and environment (CO2 emissions) are evaluated in this case study. Finally, It is demonstrated that probabilistic production energies, production costs and CO2 emissions of individual power generator are obtained in the case study

동적 무효전력 예비력 지수를 이용한 풍력 발전 단지의 전압 안정성 향상 방안 Strategy for Enhancing Voltage Stability in Wind Power Plants Using Dynamic Reactive Power Reserves Index

https://doi.org/10.5370/KIEE.2020.69.7.964

심준보(JunBo Sim) ; 최윤혁(Yun-Hyuk Choi)

To increase the penetration of wind turbine (WT) generators in the power system, grid-integration standards have been needed for the stable integration of a large-scale system connected wind turbine generators. Especially, the low voltage ride through (LVRT) function has been emphasized, as it relates to the voltage-reactive power control of the wind turbine generator. In addition, it is important to accurately determine the voltage stability of the system for fast voltage recovery in the system contingency. Therefore, this paper proposes the improved method using dynamic reactive power reserves (DRPR) among wind turbine generators in order to determine the system stability and satisfy the LVRT condition in large-scale system connected wind turbine generators. To prove the effectiveness of the proposed method, the simulation results demonstrate dynamic voltage recovery according to the LVRT conditions.

저압범위 확대에 따른 태양광발전시스템 직류배선 설계방법 Design Method for DC Wiring of PV Generation System by Expansion of Low-Voltage Range

https://doi.org/10.5370/KIEE.2020.69.7.970

현동근(Dong-Geun Hyun) ; 김동규(Dong-Kyu Kim) ; 이현명(Hyeon-Myeong Lee) ; 김재언(Jae-Eon Kim)

The Korea Electro-technical Code (KEC) established in 2018 extended the DC low-voltage range under 750 V to 1,500 V. Accordingly, it is expected that there will be some changes in the domestic photovoltaic industry. In Korea, most of PV generation systems have been operated under 750 V for cost and regulatory reasons. If KEC is implemented from 2021, PV generation system is expected to improve power generation efficiency by constructing under 1,500 V. However, in order to apply the new low- voltage range in PV generation system, it seems to be prepared more concretely. Especially, KEC currently deals with regulations for low-voltage cables and PV generation system, but the impact on PV generation system and existing cables by applying the expansion of DC 1,500 V is not considered. And the rated voltage 1 kV cable currently used is not appropriate to PV generation system expected to be constructed under 1,500 V from 2021. In addition, due to the characteristics of PV generation system, it is necessary to regulate the use of cable for PV generation system rather than using general cables at 1,500 V. Therefore, this paper proposes a revision plan of KEC applied in 2021 and a design method to prepare DC wiring for PV generation system applicable to DC 1,500 V.

최적화 하이퍼 파라미터의 학습자 기반 배깅 모델을 활용한 태양광 출력 예측 Optimized-XGBoost Learner Based Bagging model for Photovoltaic Power Forecasting

https://doi.org/10.5370/KIEE.2020.69.7.978

최성현(Sung-hyeon Choi) ; 허진(Jin Hur)

As the world is aware of the problem of greenhouse gas emissions, the trend of generating energy source has been changing from conventional fossil fuels to sustainable energy such as solar and wind. In order to reduce greenhouse gas emissions, the ratio of renewable energy sources should be increased. However, renewable energy sources highly depend on weather conditions and it has intermittent generation characteristics, thus embedding uncertainty and variability. As a result, it can cause variability and uncertainty in the power system, and that is why it is essential to have accurate forecasting technology of renewable energy to address this problem. We proposed a bagging model which is using an ensemble model as a base learner and what we set for the base learner is a XGBoost. Results showed that ensemble learner-based bagging models averagely have lower error compared to the bagging model using single model learner. Through the use of accurate forecasting technology, we will be able to reduce uncertainties in the power system and expect improved system reliability.

기상인자가 단기 전력수요예측에 미치는 영향 분석 Analysis of the Effect of Weather Factors for Short-Term Load Forecasting

https://doi.org/10.5370/KIEE.2020.69.7.985

권보성(Bo-Sung Kwon) ; 박래준(Rae-Jun Park) ; 송경빈(Kyung-Bin Song)

With the spread of renewable energy, the accuracy of load forecasting has been getting worse. The main cause of error in load forecasting is that the effect of behind-the-meter(BTM) generation is not well considered. In order to improve the accuracy of short-term load forecasting(STLF), the effects of weather factors on load forecasting should be systematically analyzed. A load forecasting model based on deep neural networks is used for analysis. There are several weather factors which are temperature, humidity, wind speed, solar radiation, cloud cover, precipitation, and precipitation probability, etc. Main purpose of the study is finding a combination of weather factors that have a good effect on improving STLF. The load forecast is performed from 2016 to 2018 to analyze forecast errors by using various weather factor combinations in case studies. The test results show that the case of using temperature, solar radiation, and precipitation as input data for weather is the most accurate among the nine case studies in STLF. The mean absolute percentage error(MAPE) at that case is 1.46% for the case studies from 2016 to 2018

불량데이터의 고속 검출을 위한 데이터 활용 단계 상태추정 연구 Two-Step State Estimation Using PMU data for Fast Detection of Bad Data in Power Systems

https://doi.org/10.5370/KIEE.2020.69.7.993

김병호(Byoung-Ho Kim) ; 김홍래(Hongrae Kim)

As the dependence on electrical energy increases, the power system is becoming lager. Since the size of power system is gradually increasing, the calculation of various power system analysis applications used to check the current state of the power system is complicated and may consume a lot of calculation time. In addition, due to the aging of the power system, it is difficult to maintain the precision of the measuring instrument, and when a large number of bad data are included in the measurement set, a lot of time may be spent detecting and identifying them. The factor that determines the computation time of state estimation is related to the size of the system. When the measurement includes a number of bad data, most of the computation time in state estimation is consumed to process the bad data. If the power system can be divided into small scales to reduce the size of the system, the calculation time and the bad data detection of state estimation can be improved. When the state estimation is performed on the partitioned system, the estimated voltage magnitudes are same as the result of the original whole system. When the state estimation is performed on the partitioned small-scale systems, the voltage angles are calculated as angle differences with respect to the slack buses for each small-scale systems. If PMUs are installed for each small-scale systems, it is possible to recognize the difference in phase angle for each partitioned system. It is possible to perform a two-step state estimation, which consists of performing the state estimation of a partitioned system and correcting the phase angle of partitioned system using PMU measurement. In this paper, we propose a method to perform state estimation in two steps on a partitioned system using PMU measurement to shorten the calculation time of the state estimation and the detection time of the bad data for the large-scaled system.

태양광전원이 연계된 배전계통에서 방향성 지락과전류계전기의 오동작 방지 알고리즘 A Prevention Algorithm for Malfunction of Directional Over Current Ground Relay in Distribution System with PV System

https://doi.org/10.5370/KIEE.2020.69.7.1001

박양권(Yang-Kwon Park) ; 이후동(Hu-Dong Lee) ; 태동현(Dong-Hyun Tae) ; 노대석(Dae-Seok Rho)

In distribution system with PV system, malfunction cases of the DOCGR have been reported by an misuse of RCA(reference characteristics angle) and OPA(operation angle) at installation of CT and caused the problems including interruption and unstable power supply. Accordingly, practical countermeasures to prevent a malfunction of DOCGR in the field are being required. Therefore, this paper presents a malfunction mechanism of DOCGR in distribution system with PV system and proposes of prevention algorithm for malfunction of DOCGR to be operated only in case of SLG(single line-to-ground) fault at PV system side. Furthermore, in order to perform a demonstration of malfunction in DOCGR, this paper presents a modeling of test devices for DOCGR which is composed of power sources, protection devices and PV system based on PSCAD/EMTDC S/W, and also implements a 2.5kW scaled test devices for characteristics of DOCGR. From the simulation and test results of malfunction characteristics of DOCGR based on the proposed modeling and test devices, it is confirmed that the malfunction of DOCGR has been occurred in case of SLG faults at both power source and PV system side if direction of CT is installed to reverse(L-K). In addition, it is found that the proposed algorithm is practical and useful tools to prevent malfunction of DOCGR without modification of wiring and position of CT when the DOCGR is operated by changing the proper setting values of RCA and OPA. And also, it can contribute to user education for preventing malfunction of DOCGR in real distribution system.

수중 드론용 일체형 추진기의 프로펠러 구조를 고려한 전동기의 설계 및 특성 분석 Design and Characteristic Analysis of Motor considering Propeller Structure of Integrated Propulsor for Underwater Drone

https://doi.org/10.5370/KIEE.2020.69.7.1010

김성안(Sung-An Kim)

This paper presents the design and characteristic analysis of a motor considering the propeller structure of an integrated propulsor for a underwater drone. In order to improve the hydraulic efficiency of the integrated propulsor, it is essential to simplify the structure of propulsor, minimize the motor size and improve the performances of motor. This paper proposes a fixing pin for fastening the rotor of motor and the propeller. Through the basic design of motor size and optimum design considering the proposed structure, it was satisfied with the requirement specifications and performance improvement of the motor. The validity of the design was verified by stress analysis of the propeller and electromagnetic analysis and thermal analysis of the motor.

선박 동적 모델을 고려한 하이브리드 시스템의 운영 제어 알고리즘 개발 Operation Control Algorithm Development of Hybrid RDP System considering Ship Dynamic Model

https://doi.org/10.5370/KIEE.2020.69.7.1016

김성안(Sung-An Kim)

The performance of a hybrid electric propulsion ship depends on the interaction of the diesel generator and the energy storage system. Therefore, this paper presents a basic design to replace the mechanical-electric propulsion system with the hybrid propulsion system. In addition, an operation control algorithm is proposed to extend the operating time and reduce the fuel consumption of ship. The proposed control consists of three modes based on the output curve according to the motor speed of rim driven propulsor. ON and OFF operation of diesel generator and the charging and discharging of the energy storage system are determined depending on the control mode. The feasibility of the proposed algorithm is verified by using the load simulation considering the ship dynamic model which improved the conventional simple load profile. As a result, fuel consumption is expected to be reduced by about 4% compared to conventional mechanical-electric propulsion systems.

노이즈 제거 및 전압 위상 지연을 고려한 부분방전 패턴 인식 Partial Discharge Pattern Recognition Considering Noise Elimination and Voltage Phase Lag

https://doi.org/10.5370/KIEE.2020.69.7.1024

윤성호(Sungho Yoon) ; 안범(Beom An) ; 이상군(Sanggoon Lee) ; 김정태(Jeongtae Kim) ; 정연하(Yeonha Jung) ; 장태인(Taein Jang)

In this study, the effect of noise elimination and voltage phase lag on the partial discharge pattern recognition was studied to improve the accuracy of the partial discharge diagnosis of the XLPE transmission cable system. Recognition rates were compared by applying Neural Network and SVM techniques using statistical feature values extracted from physical quantities in PRPDA data such as discharge numbers and discharge amounts according to the voltage phase, which were measured through the commercial partial discharge diagnostic system for four different types of partial discharge models. As a result, it was found that the minimum noise elimination method showed high pattern recognition rate because it relatively preserved the partial discharge information, even though the background noise would not be clearly eliminated. In addition, for the effect of the voltage phase lag, the neural network did not show any meaningful effect, whereas SVM showed significantly lowered recognition rate.

데이터 증대를 활용한 딥러닝 기반 위 병변 분류 시스템 Deep Learning based Gastric Lesion Classification System using Data Augmentation

https://doi.org/10.5370/KIEE.2020.69.7.1033

이신애(Sin-ae Lee) ; 김동현(Dong-hyun Kim) ; 조현종(Hyun-chong Cho)

Gastrointestinal symptoms and functional gastrointestinal disorders comprise a large proportion of primary care and gastroenterology practice. We propose a Computer-aided Diagnosis (CADx) system that analyzing the traditional gastroscope images and help the medical experts improve the accuracy of medical diagnosis. To improve the performance of the CADx system, a data augmentation has also been implemented to increase both the amount and the diversity of the training images. Augmentation method finds the enhancement parameters through RNN through large-scale verified three data, ImageNet, SHVN and CIFAR-10. In this study, we compared the performance of applying data augmentation method using four networks, Inception-V3, Resnet-101, Xception, and Inception-Resnet-V2. For Inception-V3, Resnet-101, Xception, and Inception-Resnet-V2 in normal and abnormal classification, the highest Az values were 0.87, 0.85, 0.88 and 0.82 respectively. The Xception networks and CIFAR-10 data is a promising CADx configuration for gastric lesion which had relatively simple structure and good classification performance.

합성곱 신경망을 활용한 장미잎 병충해 분류 시스템 개발 A Development of Rose Leaf Disease Classification System using Convolutional Neural Network

https://doi.org/10.5370/KIEE.2020.69.7.1040

함현식(Hyun-sik Ham) ; 조현종(Hyun-chong Cho)

The classification of plant disease by images has been studied over past decades. In this paper, convolutional neural network models were applied to perform rose leaf disease diagnosis using simple leaves images of healthy and diseased rose leaves, through deep learning methodologies. Training of the models was performed with the use of an open database of 13,125 images, containing field and laboratory images with five different disease and healthy leaves. Based on experiments, the precision and recall are 98.7% and 97.4% and the F1-score is 0.98. The significantly high success rate makes the model a very effective advisory or early warning tool, and an approach that could be further expanded to support an rose leaf disease identification system to operate in real cultivation conditions.

보안관제 정탐률 향상을 위한 인공지능 알고리즘 연구 A Study on AI algorithms to Improve Precision Rate in a Managed Security Service

https://doi.org/10.5370/KIEE.2020.69.7.1046

최승환(Seunghwan Choi) ; 장민해(Minhae Jang) ; 김명수(Myongsoo Kim)

Cyber attacks are becoming intelligent and mass-produced. Thus, precision rate is very important in terms of a managed security service. Currently, cyber attacks are detected using various security protection devices and mass security events from the security devices are inevitable. Most devices are carring out defense on predetermined rulesets. Due to the problem of classifying non-attacks as attacks, people are unable to handle massive events log. Researches have been conducted to solve this problem by fine-tuning rulesets, but there is a limit to improving precision rate. As a solution, applying AI technology to the security monitoring areas have been researched in recent years. However, research on improving precision rate, which is the basis of managed security service, has not been conducted as much. In addition, the dataset used in the research is different from the log collected in the security devices in the real environment. This paper describes the logs that can be collected in the real network environment, the datasets that are used in the past studies, and artificial intelligence algorithm research for improving precision rate based on datasets collected in real network.

딥러닝 기반 무인전력설비 감시시스템 개발을 위한 학습 데이터 셋 구축 연구 A Study on Creation a Learning Dataset for the Development of an Unmanned Power Facility Monitoring System based on Deep Learning

https://doi.org/10.5370/KIEE.2020.69.7.1053

정남준(Nam-Joon Jung) ; 채창훈(Chang-Hun Chae) ; 황명하(Myeong-Ha Hwang) ; 이인태(In-Tae Lee)

Since the emergence of deep learning technology, research and development using it in various industries has been in progress. In the power field, deep learning technology based on image recognition has been in the spotlight for the development of condition diagnosis and monitoring systems. However, in order to increase the performance of the deep learning model, it is necessary to build a large data set for learning it as well as the deep learning model. Therefore, in this paper, we propose a dataset creation method for development of unmanned power facility monitoring system based on deep learning. In an environment in which a PTZ camera and a 360-degree camera are operated simultaneously, various method for generating learning data were proposed, and the performance of the learning data set was proved through the detection system experiment.

특징 추출 기법에 따른 신장암 데이터의 분류 Classification of Kidney Cancer Data based on Feature Extraction Methods

https://doi.org/10.5370/KIEE.2020.69.7.1061

손호선(Ho Sun Shon) ; 김경옥(Kyoung Ok Kim) ; 차은종(Eun Jong Cha) ; 김경아(Kyung Ah Kim)

Recently, Numerous data mining methods in the bioinformatics field have been developed for processing biodata. We extracted significant genes (60,483 of gene expression data from TCGA) for the prognosis prediction of 1,157 patients using gene expression data from patients with kidney cancer and applied classification methods based on data mining. Significant genes were extracted using least absolute shrinkage and selection operator (LASSO) and principal component analysis (PCA), and classification accuracy and performance were compared using a classification algorithm. Combined clinical data from patients with kidney cancer and gene data were used to determine the optimal classification model and estimate classification accuracy as risk factors by sample type, primary diagnosis, tumor stage, and vital status representing the state of patients. Classification accuracy based on sample type showed the best performance, particularly for the logistic regression and support vector machine algorithms. These results can be applied to extract biomarkers for prognosis prediction of kidney cancer from various causes and for preventing kidney cancer and early diagnosis.

엑스레이 수하물 영상에서 위해물품 인식 인공지능 시스템 개발 Development of Artificial Intelligence System for Dangerous Object Recognition in X-ray Baggage Images

https://doi.org/10.5370/KIEE.2020.69.7.1067

이정남(Jeong-nam Lee) ; 조현종(Hyun-chong Cho)

The importance of flight safety has been highlighted lately due to the increase of aviation industry. Baggage screening tasks are still difficult and the failures of dangerous object detection are frequent despite the improvement of screening equipment. The purpose of this study to develop the AI system on dangerous object detection for improving aviation safety. The convolutional neural network model, Xception, were applied to perform dangerous object recognition using X-ray baggage image dataset which contains 25,405 images of twelve items. Based on experiments, the accuracy and F1-score are 0.9939 and 0.9942. The significantly high success rate makes the model a very effective advisory or early warning tool, and an approach that could be further expanded to support a dangerous object identification system to operate in real airport screening process.

전방 이미지를 위한 세포정량화 도구 Cell Quantization Tool for Anterior Chamber OCT Images

https://doi.org/10.5370/KIEE.2020.69.7.1073

이영섭(Yeongseop Lee) ; 강태신(Taeseen Kang) ; 한용섭(Yongseop Han) ; 김진현(Jinhyun Kim) ; 김경훈(Kyong Hoon Kim) ; 이성진(Seongjin Lee)

In general, ophthalmologists visually grade the state of a patient by counting the cells within the anterior chamber OCT image. The manual cell counting method is highly inaccurate and spends a lot of time to determine the progress of the patient. In this work, we develop a new tool to count cells in anterior chamber OCT images to aid doctors in analyzing the state of patients. We exploit image processing to remove noises from images, segment the anterior chamber, and quantize the cells in OCT images. We also provide statistics to aid the doctors in determining the progress of the patients.

음성의 음향적 특징과 얼굴 이미지 시퀀스를 이용한 감정인식에 관한 연구 A Study on Emotion Recognition using Speech Acoustic Features and Face Images

https://doi.org/10.5370/KIEE.2020.69.7.1081

손명진(Myoung-jin Son) ; 이석필(Seok-pil Lee)

Generally, people recognize other people's emotions by their voices and facial expressions. So, speech signals and facial images have been actively studied in the field of emotional recognition. Therefore, in this paper, we present effective acoustic features for emotion recognition and a method to recognize emotions by combining speech signals and facial image sequences. To combine these the two inputs like speech signals and facial image sequences, three models are designed. And these three models are combined by using the Joint Fine Tuning method. The result shows that the performance of our model is very promising for emotion recognitions in comparison with other models using speech signals and facial image sequences.

사용자의 온오프 라인 쇼핑에서 상품 추천을 위한 전이학습 모델 비교 Comparison of Transfer Learning Models for Goods Recommendation in User’s Online and Offline Shopping

https://doi.org/10.5370/KIEE.2020.69.7.1087

정종진(Jongjin Jung) ; 강동구(Donggu Kang) ; 김지연(Jiyeon Kim)

Recently, deep learning based services has been actively developed in various fields. Especially, in the online distribution environment some Applications with deep learning based on large amounts of data and user information began to be applied to real systems. In this paper, we develop an application system which recommends goods after learning and predicting using deep learning by presenting good image that user wants to purchase in online market sites. At this time, we apply three revised models based on the transfer learning model for automatic image recognition. After image recognition of goods, the system produces candidate goods and requires filter condition input for goods he wants to purchase. Finally, the system recommends goods by content-based recommendation method. In this paper, we experiment three revised models of image recognition which are used as core techniques in the system. These models are fully connected network, k-nearest neighbors search and linear SVM and they act as a classifier within the transfer learning model architecture. We experiment and compare performance evaluation for the revised models.

측정 잡음을 고려한 기 무한모선 시스템의 고장 판별을 위한 기반 외란 관측기 설계 Design of a DQN-based DOB for Line Fault Detection of a Single Machine Infinite Bus System Against Measurement Noise

https://doi.org/10.5370/KIEE.2020.69.7.1095

양선직(Sun Jick Yang) ; 장수영(Su Young Jang) ; 손영익(Young Ik Son)

Increasing demand on electric power supply results in a need of an intelligent method for fast detection of various failures in the power system. This paper presents a reinforcement learning-based disturbance observer (DOB) design for the determination and protection against a line fault occurred in the single-machine infinite bus (SMIB) power system. Whilst a high gain disturbance observer could estimate the system states and the external disturbance successfully, the high gain of the observer can cause problems in the presence of the measurement noise. When measurement noise exists in the output, fault detection methods based on the estimated states may often result in false alarms. To solve the problem, this paper designs an adaptive DOB using Deep Q-Network (DQN) which is one of reinforcement learning algorithms. For the proposed observer design, this paper explains the definitions of the state, the action, and the reward for the reinforcement learning. Matlab simulations have been conducted based on the observer gains trained using the power angle data from the swing equation. The results show that the estimation performance of the proposed DQN-based observer can be satisfactory against both an external disturbance and the measurement noise.

환경적 요소 선별을 통한 딥러닝 기반 우울증 분류 예측 Deep Learning based Depression Classification using Environmental Factor Selection

https://doi.org/10.5370/KIEE.2020.69.7.1102

남원우(Wonwoo Nam) ; 김병욱(Byung Wook Kim)

Previous studies have examined whether symptoms found in annual health examinations could be predictive to classify a person with depressive disorders. In this paper, Convolutional Neural Network(CNN) and Light Gradient Boosting Machine (LightGBM)-based depression classification models were proposed based on physical and environmental information of health examinations. For this, input data of CNN and LightGBM were pre-processed by adding and excluding several environmental information that could highly affect the prediction results. And the optimal model of CNN and LightGBM were obtained through hyperparameter analysis to maximize the depression classification performance. Performance results proved that the predictive accuracy of 2D-CNN was 78.71% and AUC values for 1D-CNN, 2D-CNN, LightGBM were 0.750, 0.716, 0.731, respectively. By comparing performance results, our proposed classification models outperformed the ANN and DNN-based conventional models in terms of accuracy and AUC.

인공지능 해석 기법을 이용한 태양광 발전량 예측 성능 향상 Improvement of Solar Power Forecasting Using Interpretation of Artificial Intelligence

https://doi.org/10.5370/KIEE.2020.69.7.1112

오재영(Jae-Young Oh) ; 이용건(Yong-Geon Lee) ; 김기백(Gibak Kim)

Artificial intelligence (AI) has been effectively applied to various industries thanks to the increased availability of data and computing power. Advanced machine learning techniques also contribute to the widespread application of AI. However, it is becoming more difficult to interpret the AI implemented by advanced and highly complex machine learning algorithm. In this paper, for solar power forecasting system, we conduct SHAP value analysis which is one of the explainable AI techniques. We aim to improve the performance of the solar power forecasting by employing feature selection which is based on the feature importance computed by SHAP values. In the experimental results, three different machine learning algorithms (SVM, ANN, XGBoost) are applied for solar power forecasting and shown to improve the forecasting performance in all three methods.

도심비행환경 맵을 이용한 복합항법 성능 검증을 위한 기반 설계 PILS Design based on Unity3D for Performance Verification of Multi-Sensor Integrated Navigation System using 3D GIS Map in Urban Canyon

https://doi.org/10.5370/KIEE.2020.69.7.1118

고은학(Eun-Hak Koh) ; 유강수(Kang-soo Ryu) ; 성상경(Sangkyung Sung)

This paper presents PILS using Unity3D, a game development tool, for verification of navigation performance in various environments. The navigation performed with PILS is a fusion/complex navigation using GPS navigation and 3D GIS maps and multiple distance sensors in urban environments where satellite signals are not reliable. Accordingly, the topography and the building are constituted by utilizing the Unity3D program, and the virtual data are created by modeling sensors, vehicles, and vehicle’s guidance and controller. To create a more realistic simulation, actual sensor noise characteristics, and actual flight maneuvers are analyzed. Data are transferred to a navigation module used for actual flight by using serial communication, a navigation solution obtained from the navigation module is plotted in GCS. Solutions are transferred to a PC again where data are generated and finally compared with a true value in real-time.

심층 신경회로망 앙상블을 이용한 걸음걸이 인식에 대한 연구 A Study on Gait Recognition using Deep Neural Network Ensemble

https://doi.org/10.5370/KIEE.2020.69.7.1126

홍성준(Sungjun Hong) ; 이수형(Soohyung Lee) ; 이희성(Heesung Lee)

The recognition of a person from his (her) gait has been a recent focus in computer vision because of its unique advantages such as non-invasive and human friendly. Gait recognition, however, has the weakness that it is not reliable compared with other biometrics. In this paper, we applied deep neural network ensemble to the gait recognition problem. The deep neural network ensemble is a learning paradigm where a collection of deep neural networks is trained for the same task. Generally, the ensemble shows better generalization performance than a single deep neural network such as convolution neural network and recurrent neural network. To increase reliability of the gait recognition, gait energy image (GEI) and Motion silhouette image (MSI) are extracted for gait features and convolution and recurrent neural network ensemble are used for classifier. Experiments are performed with the NLPR and SOTON databases to show the efficiency of the proposed algorithm. The performance of proposed method is 4.55%, 4.85%, 2.5% and 2.43% better than single CNN, respectively in two databases. As a result we can create a recognition system with accuracy of 100%, 100%, and 94% in the NLPR database and 97.35% in the SOTON database.

장단기 메모리를 이용한 새로운 선박 이동 경로 예측 방법 A New Vessel Path Prediction Method using Long Short-term Memory

https://doi.org/10.5370/KIEE.2020.69.7.1132

김종희(Jonghee Kim) ; 정찬호(Chanho Jung) ; 강도근(Dokeun Kang) ; 이창진(Chang Jin Lee)

In this paper, we propose a new vessel path prediction method using long short-term memory (LSTM). LSTM is one of recurrent neural networks which contains memory cell in order to deal with long-term data. In order to fully utilize the advantage of LSTM, our proposed method employs 3-layer LSTM instead of a fully connected layer. We also propose new input and output vectors well suited for the vessel path prediction. In order to prove the effectiveness of the proposed method, we compare the proposed method with a baseline method which consists of a LSTM and a fully connected layer. In comparison between the proposed method and the baseline method, the proposed method outperforms the baseline method based on LSTM.

전기안전관리 플랫폼 보안 위협분석 및 대응방안 연구 A Study on the Security Threat Analysis and Countermeasures of Electric Safety Management Platform

https://doi.org/10.5370/KIEE.2020.69.7.1136

김장훈(Jang Hoon Kim) ; 이주찬(Joo Chan Lee) ; 신지호(Ji Ho Sin) ; 서정택(Jung Taek Seo)

The electric safety management platform using IoT(Internet of Things) technology is a platform for real-time monitoring of power consumption and generation. also, It checks the power demand and production using a sensor to each home, factory and proper power supply. However, the electric safety management platform has become more and more connected to the internet. Therefore, the electric safety management platform becomes vulnerable to cyber security threats. so, In this paper, we analyze the electric safety management platform using IoT, and we study the security threats that can occur for each interface and their countermeasures.

무변압기형 태양광발전용인버터와 태양전지 모의 직류전원장치 연동 시 고장분석 Failure Analysis when Interlocking a Transformerless Type PV Inverter with a PV DC Simulator

https://doi.org/10.5370/KIEE.2020.69.7.1144

정성인(Sung-In Jeong)

Normal characteristics test of grid-connected photovoltaic inverters using test facilities such as a PV dc simulator power supply, a simulation system power supply (AC-simulator), a measuring device (power meter, analyzer, etc.) and a load device (RLC dummy load), and protective function tests. Recently, a transformerless type photovoltaic inverter that can increase power generation efficiency, reduce weight, size, and increase power generation time has been developed and popularized. The transformer-less inverter for solar power generation does not reflect the insulation function, which is the basic role of the transformer, internally. Therefore, when applying the transformerless inverter, abnormal operation may occur through breakdown between the devices and mutual interference, and sometimes it may cause a serious failure in the test equipment (simulated DC power supply for the solar cell, simulated power supply, etc). Therefore, in this paper, the transformerless type inverter for photovoltaic power generation is connected and connected to the device as shown in Fig. 1 to configure one system to protect and solve the test facility through failure analysis based on the problems that occur when the devices are interlocked. I would like to suggest a method through experiments.