| Summary: | 碩士 === 中原大學 === 電機工程研究所 === 90 ===
This thesis proposes an optimal FNNC whose initial setting of parameters and learning rate are done by AGA. The FNNC is based on TSK fuzzy model and is realized from the network point of view. The parameters of the fuzzy model are tuned on-line by a backpropogation algorithm. Usually fuzzy logic controllers use error and error change rate as inputs, our design use plant output error and input instead. Therefore, the proposed FNNC generates control signal according to plant input and output information directly.
In FNNC, the initial setting of parameters has decisive effects for control results. The initial values of parameters are usually chosen by trial-and-error or by experience. In the thesis, an optimal FNNC is obtained by AGA. Compared to the traditional GA, the proposed AGA has varying crossover-rate and mutation-rate to prevent the falling of local optimum and to speed up convergence.
The optimal FNNC obtained by AGA is applied to the simulation of a second order linear system, a nonlinear system and a highly nonlinear system with instantaneous loads. When compared with the initial parameters of FNNC chosen by trial-and-error, the results show that the controlled systems have good tracking ability even for different trajectories and instantaneous loads.
|
|---|
