http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
미분없는 칼만필터 학습 알고리즘을 이용한 순환신경망의 파라미터 추정
권오신 국립7개대학공동논문집간행위원회 2004 공업기술연구 Vol.4 No.-
Recurrent neural network(RNN) trained with gradient based such as real time recurrent learning has a drawback of slow convergence rate. This algorithm also needs the derivative calculation which is not trivialized in error back propagation process. In this paper a derivative free Kalman filter for training a fully connected RNN is presented in a state space formulation of the system. A derivative free Kalman filter learning algorithm makes the RNN have fast convergence speed and good tracking performance without the derivative computation. Through experiments of nonlinear channel equalization, performance of the RNN with a derivative free Kalman filter algorithm is evaluated.
화력발전에서 비선형 모델링 기법 기반의 보상기를 갖는 다 변수 입출력 제어기의 설계
권오신,박성만,나상건,이용희,정원희,허훈 제어로봇시스템학회 2010 제어로봇시스템학회 국내학술대회 논문집 Vol.2010 No.5
Conventional control of steam temperature using feedback cascade-PID controller in the superheater and the reheater of thermal power plant is known to be efficient and robust to compensate disturbance. However, It is not easy to compensate nonlinearity and time delay due to complex and enormous structure of the thermal power plant. In order to compensate nonlinearity of superheater and reheater, the compensator is constructed via nonlinear modeling technique. Also, feedforward control loop is designed for time delay compensation. In this study, multi-loop MIMO controller that is comprised feedback/feedforward loop and compensator loop is proposed for performance improvement in steam temperature control. The designed multi-loop MIMO controller is implemented to superheater and reheater of thermal power plant, the nonlinear model of compensator is validated for various state of load change(100%↔90, 80, 75, 70%) and rate of load change(5, 3, 1%/min). The simulation of control performance is performed for 100% to 75% and 75% to 100% load changing state, also for 5%/min rate of load change, The proposed multi-loop MIMO controller with compensator based on nonlinear modeling technique reveals more stable and efficient performance than conventional feedback cascade-PID controller in load changing state.