| In the last few years, the development of the technique of modern electric power and electronics, micro-processor, and advanced control provides very important pledge for high-performance control of motor. The DTC technique, which belongs to high performance variable-frequency adjusting-speed technique in the AC drives, is researched widely. However, for high-performance speed close-loop system, the feedback of speed is wanted. Speed-sensorless technique is very important to reduce its cost and strengthen its reliability, and has become the focus of AC adjusting-speed field.The dissertation presents the development and the principle of the Direct Torque control (DTC) systematically and detailedly. And then brings forward a kind of method used to identify parameters, based on chaos intelligence algorithm. With the help of the intelligent control theory and direct torque control theory, a kind of Neural Network Group (NNG) trained by multiorbit chaos optimization algorithm is put forward, and the method of identifying parameters through speed identifier is modified. In our research, we find samples used to train the neutral network are inconsistent, so-called "samples inconsistence". Even though the fitness is satisfying in the simulation environment, most of them will produce periodical errors, when these algorithms are employed to the DTC system to observe speed. The solution multiorbit chaos optimizing NNG is presented ultimately to resolve that problem. On one hand, the NNG trained by the multiorbit chaos optimization algorithm is able to produce better result and performance than single-orbit and single-nonlinear function algorithm. On the other hand, DTC system has better character with it.Chipset TMS320F240, specially used in the motor control field, and intelligent power module (IPM), with protection circuit to improve the system dependability, are employed in the experiment. The experiment supports the validation and feasibility of the solution presented in the dissertation.
|
PDF Full Text Request