Control and Prediction of Chaotic Systems

• Main Author: 傅俊中
• Other Authors: 陳建忠
• Format: Others
• Language: zh-TW
• Published: 1995
• Online Access: http://ndltd.ncl.edu.tw/handle/26433375871084407234

碩士 === 國立中正大學 === 化學工程研究所 === 83 === 　　There are two topicws in this thesis: chaotic control and chaotic prediction. For the chaotic control two methods. (1) proportional feedback method and (2) self interaction method, are used to control the chaotic dynamics of a nonisothermal coupled CSTRs and can stabilize the chaotic dynamics to short or long period orbits. In method (1), unstable periodic orbits needed to be extracted first and then be used as set points such that chaotic dynamics can be controlled to these extracted orbits. For method (2) methematical model is required. Fixed point, periodic orbit, or tours can be obtained via small perturbations on the manupilable parametersl. All these two methods require only small energy to achieve the control purpose. 　　Next, I propose a new chaotic predictor. which combines the nearest neighbor concept of the local predictor (LP) and the artificial neural network (ANN). The results on the Lorenz time series show that 1.6×10-6 error in my predictor is superior to 4.2×10-6 of the LP and 4.75×10.-6of ANN . Moreover, the prediction horizon is longer that the above two methods.