Application And Research Of Artificial Fish School Algorithm In Dtc System

Direct Torque Control(DTC)technique has been developed since the 1980's. It is a newly high performance AC drive control technique after the vector control technique. DTC calculates and controls the torque of induction machine in the stator reference frame, using stator field-orientation. And it directly controls the switching mode of inverter. It has a good prospect in the industrial variable speed field because of its novel control method, simple structure, and well stationary and dynamic characters.Sensorless speed identification technology has become a hot direction of research in modern AC drive control for it can increase system reliability and reduce the cost. BP Neural network is used to recognize machine's speed in this dissertation and because the traditional BP Neural network is easy to plunge into local extremum and has a low searching speed, An improved artificial fish school algorithm(AFSA) is used to optimize BP neural network in order to chase speed well.In this dissertation, the history of the adjustable speed of AC drives, electric power electronic technology and speed sensorless technology is firstly reviewed and explained. Then the mathematical model and coordinate transformation is discussed as the base of voltage space vector, which has effects on the asynchronism machine stator flux and electromagnetic torque and circular flux track of DTC.Secondly, on the bases of analyzing the principle, advantage and disadvantage of basic AFSA, some improvement is made on AFSA for precision and speed of optimizing and completes research of optimizing BP neural network, which proves to be effective comparing with other algorithms. The optimized BP neural network is used as a speeder identifier in DTC system in matlab/simulink, whose results is particularly analysed.Finally, the asynchronism machine speed sensorless DTC experiment is carried out on the existing equipment, including programming, debugging and putting the BP neural network speed identifier into DSP and working on line. The results shows that the BP neural network speed identifier can identify speed very well, which proves the scheme is correct.