경주시 바이오가스 생산과 에너지 이용에 관한 연구

고택범 경주지역발전협의회 부설 경주발전연구원 2009 경주연구 Vol.18 No.2

공기열원 수축열식 히트펌프 시스템의 냉방 성능평가

고택범 경주대학교 건설환경연구소 2009 建設環境論叢 Vol.- No.11

The main objective of the present study is to investigate the cooling performance characteristics of a water storage air source heat pump system. In order to evaluate the cooling performance, the water storage air source heat pump system for a condominium in Gyeongju was designed and constructed. Four kinds of field tests were performed to obtain the experimental data from October 1, 2004 to January 1, 2005. The field tests show that the cooling performance characteristics including total cooling load, hourly maximum cooling load, insulation performance for thermal storage reservoir, and efficiency of total energy utilization of the system are satisfied with the authorization standards of KEPCO. Downsizing of thermal storage reservoir is essential to enlarge the application of water storage heat pump heating and cooling systems to resident and medical buildings.

하수처리 시스템 활성오니공정의 퍼지 모델링

고택범 경주대학교 2002 論文集 Vol.15 No.-

The modeling of the activated sludge process, which is widely used in modem wastewater treatment plants, is difficult due to the following reasons 1) the complexity of the wastewater components, 2) the change of the wastewater influent, and 3) the adjustment errors in the control process. Because of these reasons, it is difficult to obtain mathematical models that reflect the relationship between the variables and parameters in the wastewater treatment process correctly and effectively. This paper presents an application of self-organizing fuzzy modeling (SOFUM) to build the model of an activated sludge process. The suggested algorithm SOFUM is composed of four steps coarse tuning, fine tuning, cluster creation and optimization of learning rates. In the coarse tuning, fuzzy C-regression model clustering and weighted recursive least squared algorithm are used, and in the fine tuning gradient descent algorithm is used to adjust parameters of the fuzzy model precisely. In the cluster creation, a new hyperplane-shaped cluster is created by applying multiple regression to input/output data with relatively large fuzzy entropy based on parameter tunings of fuzzy model. And learning rates are optimized by utilizing meiosis-genetic algorithm in the optimization of learning rates. The efficacy of the proposed algorithm is evaluated by means of computer simulation based on measurement data in the activated sludge process.

최적의 베이지안 순환 신경망을 이용한 시계열 예측

고택범 慶州大學校 2005 論文集 Vol.18 No.2

In this paper, the optimized Bayesian recurrent neural network is proposed to predict time series data. The Bayesian neural network estimates not the single set of weights but the probability distributions of weights. The weights vector is set as a state vector in the state space method, and its probability distributions are estimated using Bayesian inference. In the Bayesian inference process, multi-dimensional integrals are easily evaluated in accordance with the recursive Monte Carlo method, so called the particle filter. The recurrent neural network is expected to show higher performance than the feedforward neural network for the problem of time series prediction. The structure and parameter of the Bayesian recurrent neural network are optimized utilizing meiosis-genetic algorithm. To check the effectiveness and feasibility of the suggested algorithm, two representative examples for time series prediction are examined and the prediction performance of the optimized Bayesian recurrent neural network is demonstrated in comparison with that of the conventional time series models.

주성분분석을 이용한 자기구성 퍼지 모델링

고택범 경주대학교 정보전자기술연구소 2003 情報電子技術論叢 Vol.2 No.-

This paper proposes a self-organizing fuzzy modeling which can create a new hyperplane-shaped cluster and adjust parameters of the fuzzy model in repetition. The suggested algorithm is composed of five steps input data transformation, initial cluster creation, coarse tuning, fine tuning and new cluster addition. Input data is transformed by principal component analysis to reduce the correlation among input data components. Initial clusters are created by fuzzy neural network clustering which uses the control structure similar to Adaptive Resonance Theory-1. In the coarse tuning, fuzzy C-regression model clustering and weighted recursive least squared algorithm are used and in the fine tuning, gradient descent algorithm is used to adjust parameters of the fuzzy model precisely. In the cluster creation, a new hyperplane-shaped cluster is created by applying multiple regression to input/output data with relatively large fuzzy entropy based on parameter tuning of fuzzy model. Finally, we applied to the proposed method to the gas furnace data modeling utilized by Box and Jenkins, and obtained a better performance than previous works.

신경회로망과 유전 알고리즘을 이용한 상수처리시스템의 응집제 주입공정 최적제어

고택범,김선필,임찬호 경주대학교 건설환경연구소 1999 建設環境論叢 Vol.- No.2

This paper presents the methods on intelligent process modeling to determine the optimum dosage of coagulant in water purification plant. The method of intelligent process modeling is carried out using the neural network models to predict the optimum dosage of coagulant, PAC(Polymerized Aluminium Chloride), of coagulant dosing process based on jar test results. The network structure and learning parameters in neural networks are derived by optimization based on genetic and complex algorithms. The optimum dosage of coagulant fir minimizing turbidity in flocculation is obtained using the neural network models based on water quality data (turbidity, temperature, PH, alkalinity of raw water and turbidity in flocculation). The results of experiments performed on the water purification plant and computer simulation are summarized as follows: (1) The prediction error of optimum neural network model utilizing genetic algorithm was less than half compared to that of the statistical model. (2) Because jar test data based on operator's experiences, normal and abnormal state models were utilized for the process modeling, the prediction performance of model has been improved.

최적 신경회로망을 이용한 플라즈마 식각공정의 시계열 모델과 실시간 이상검출

고택범 경주대학교 1999 論文集 Vol.12 No.2

This paper presents the methods on process modeling and real-time malfunction detection to provide an accurate and efficient operation for plasma etching process. The method of process modeling is carried out to predict an accurate ADI CD(after deposition inspection critical dimension) of gate polysilicon using time series neural network model which represents the changes in process dynamic characteristics due to process drifts. The network structure and learning parameters in neural network are derived by optimization based on genetic and complex algorithms. The main features of EPF(endpoint detection) signals are extracted by the application of PCA(principal component analysis) to EPD signal trajectories. ADI CD can be predicted using time series neural network model which represents the relation between ADI CD and the main features of EPD signals. The process malfunctions are detected correctly by applying SPC(statistical process control) method to the predicted CD. All of the procedures in malfunction detection can be performed on-line, wafer-by-wafer.

신경회로망과 유전 알고리즘을 이용한 냉방부하예측

고택범 경주대학교 정보전자기술연구소 2005 情報電子技術論叢 Vol.4 No.-

In this paper, the method of estimating temperature and humidity and forecasting the cooling load of ice-storage system using neural networks is suggested. The learning parameters of the neural networks are derived to reduce the prediction errors of cooling load by utilizing meiosis genetic algorithm. The forecast of the temperature, humidity and cooling load are simulated. and the forecasted data approach to the actual data as a result. Operating the ice-storage system with night electric power by the forecast of cooling load can improve the ice-storage system more efficiently and reduce the peak eletric power load during the summer season.

빙출열 시스템의 냉방부하예측과 최적운전에 관한 연구

고택범 경주대학교 정보전자기술연구소 2007 情報電子技術論叢 Vol.6 No.-

In this paper, the method for cooling load forecast and optimal operation of ice-storage system using self-organizing fuzzy models and dynamic programming is suggested. Optimal operation of the ice-storage system with night electric power by the forecast of cooling load can improve the ice-storage system more efficiently and reduce the peak electric power load during the summer season. To check the effectiveness and feasibility of the suggested method, the actual example for forecasting cooling load, temperature and humidity of ice-storage system in KEPCO training institute, Sokcho, is examined. The results of computer simulation and actual operation show that the accuracy of cooling load, temperature, humidity forecast using the suggested algorithm is higher than that of the conventional methods and the optimal operation for ice-storage system provides stable control of 3-way valves and efficient energy consumption.

최적 신경망을 이용한 빙축열 시스템의 냉방부하예측

고택범 경주대학교 지역개발연구소 2006 地域開發論叢 Vol.- No.7

This paper presents the method of forecasting temperature, humidity and cooling load of ice-storage system using neural networks in order to reduce the electric power load during the peak time in summer and improve energy efficiencies. The learning parameters of the neural network models, such as learning rates, momentums and output ranges, are optimized by meiosis-genetic algorithm to reduce the prediction errors of temperature, humidity and cooling load of ice-storage system. To check the effectiveness and feasibility of the suggested algorithm, the actual example for forecasting temperature, humidity and cooling load of ice-storage system in KEPCO training institute, Sokcho, is examined. The experiment results show that the prediction capabilities of the suggested algorithm are superior to those of the conventional method. Moreover the suggested algorithm is simpler than the conventional method in forecasting temperature and humidity of ice-storage system.