| Type-2 Fuzzy system is a kind of type-2 fuzzy set based non-linear inference system. To compare with traditional fuzzy system, namely type-1 fuzzy system, type-2 fuzzy system exhibits excellent performance in treating the uncertainty of language variable, experts knowledge uncer-tainty, internal and external disturbances, etc. Type-2 fuzzy system is widely applied in system control, data classification, prediction and system modeling fields. However, the parameter di-mension and complexity of type-2 system are higher than typical fuzzy system due to the three dimensional fuzzy set characteristic. Therefore, the difficulty of parameter and rule distribution process will further increase, and it is easy to cause the curse of dimensionality problem in multi-variables system.Therefore, aiming at the problems of parameter optimization, rule adjustment and system modeling, the corresponding research works are done. The main contributions of this dissertation are summarized as follows:(1) To the problems of type-2 fuzzy system parameter complexity, an improved RNA-GA is proposed with DNA double helix complementary. This algorithm adopts nucleotide bases based encoding method and introduces the reconstruction operation and the novel selection scheme to improve the higher dimension search efficiency. The validity of this algorithm is confirmed by testing some benchmark functions. Then, the RNA-GA based type-2 fuzzy controller is adopted to a double inverted pendulum system. The effectiveness of the proposed method is verified by simulation and comparison results.(2) On the problem of type-2 fuzzy system parameters and rule base assignment, a DE al-gorithm based type-2 fuzzy system is proposed. DE algorithm is a real encoding global search algorithm with excellent global search performance and can decrease the encoding length of pa-rameters and rules efficiently. The DE algorithm based type-2 fuzzy system is applied in designing and optimizing of the power system stabilizer, and the validity of this method is confirmed by simulation results.(3) Given the curse of dimensional problem of the multi-variable type-2 fuzzy system, a type-2 fuzzy sliding controller is designed to reduce the complexity of type-2 fuzzy system. For those complex dynamical nonlinear system, however, it is difficult to set the sliding surface function and controller parameter efficiently. In order to solve this problem, inspiring by the bee colony forag-ing mechanism, a hybrid DE (HDE) algorithm is proposed by introducing two kinds of neighbor search operations for improving the local search performance. The effectiveness of the proposed method is confirmed by testing some benchmark functions. Then, the type-2 fuzzy sliding model controller and the HDE algorithm are employed in the overhead crane system to demonstrated the effectiveness.(4) For solving the model deterioration problem, a type-2 T-S fuzzy neuro-network (T2T-SFNN) model is presented and some adaptive type-reduction parameters are introduced. In order to identify plenty of complex nonlinear parameters in the proposed model, inspired by genetic mechanism, a hybrid adaptive DE (HADE) algorithm is proposed by introducing two kinds of dynamical adaptive parameters to enhance the global and dynamic search performance. The va-lidity of this algorithm is verified by some benchmark function tests. Then, the proposed model and HADE algorithm are used in time series prediction and overhead crane system modeling to demonstrate its validity respectively.
|
PDF Full Text Request