The Control Algorithm Of The Direct Torque Under The Low-speed

Direct Torque Control System (DTC) is the frequency control system with ahigh-performance after the vector control technology. Direct torque control system whichbasically solve the complex control structures, the large amount of calculation andsensitive to changes in the structural parameters in the vector control have the fast response,robustness and simple control structure. The traditional direct torque control have thetorque ripple (with a poor load capacity) and the low speed control accuracy at low speeds.In order to reduce or eliminate speed ripple and improve the accuracy of speed at lowspeed, expand the speed range of the direct torque control system. The intelligent controlalgorithms are used for direct torque control system to achieve high performance atlow-speed stage.Main works are as follows:（1）The basic theory of BP neural network, direct torque control, BP neural networkalgorithm were introduced. According to the basic theoretical knowledge of the directtorque control principle, the control model was established. Selecting the optimalswitching sequence generate the optimized voltage waveform to control direct torquecontrol system by BP neural network. Through the analysis of simulation results obtainedthe torque ripple (±4N m), speed control error (±3r/min) and the circular trajectory ofthe stator flux and differ to expectations at low speeds. The reasons were analyzed forthese problems, and the feasibility of the network model can be proved by the speedresponse curve.（2）In order to improve the system with a better load capacity(reducing torque ripple,speed fluctuations and the flux path), the fuzzy algorithm and BP network were combined,by using the nonlinear and uncertainties of the fuzzy algorithm to improve theshortcomings of the BP network. The new model contains recognition association fuzzyinformation processing and adaptive. Experimental results show that the model is largelybetter than the BP neural network model such as the torque ripple (±1.5N m), the highprecision of the speed (±1.5r/min) and the stator flux trajectory closer to the circle. Bycomparing the results of the fuzzy network model and BP neural network control model itwas proved that the fuzzy neural network model is more feasible, but the model need along time. （3）In order to further improve the shortcomings of fuzzy neural network, the geneticalgorithm and BP network were combined. By using genetic algorithm’s global searchcapability to find the optimal weights and thresholds for the BP network establish thegenetic neural network model. Experiments indicate that the model’s network performance(convergence speed and the mean square error of test samples) are better than the BPnetwork model’s. Finally the results of BP neural network were compared with the geneticBP neural network model and fuzzy neural network, it show that the genetic neuralnetwork has better dynamic and static performance (±0.5r/min), with a better loadcapacity (±0.5N m) and the stator flux is basically a circular path. The genetic neuralnetwork model is more reasonable and practical.（4）All studies above showed that the direct torque control model based on geneticneural network has high recognition accuracy, practicality and versatility under thelow-speed and better meet the control requirements. Not only the model is suitable forlow-speed, but also it can meet the performance requirements of control at rated speed. Sothe genetic neural network model is more suitable for direct torque control.