Research On Speed Identification For Variable Frequency Drive

As the development of power electronics and microelectronics.AC drive system,for which speed dection is an essential part, has gradually supplanted DC drive system.The traditionally way, in which a shaft encoder is used to detect speed.increase cost of complexity and decrease environmental adaptability. Because sensorless vector control, in which there is no speed sensor and the speed is estimated from machine terminal voltages and currents, can simplfy the system and enhance robustness,it gradually become a research hotspot. Three methods are applied to identify the speed of synchronous motor in the research work.Current situation in speed estimation of adjustable-speed drive is described, principles of vector control is introduced and mathematical models of ac motor in various reference frames are built. Rotor flux oriented vector control is highlighted and simulation model is built using Matlab/Simulink.According theory of popov hyperstability, a adptive law is designed.Based on basic theory of model reference of adaptive system (MRAS), speed identification algorithm is proposed. Simulation results in different conditions, including no-load,with load.are analyzed. The results prove that the proposed algorithm can acheive better dynamic performanc.However, the proposed algorithm can't ensure zero steady state error on the setimated speed.Using state equation of ac motor.identification methods for speed and rotor flux are developed based on extended Kalman filter.Appropriate filter parameters are selected and identification algorithm is coded.The simulation results show the proposed methods have good filtering effect,anti-jamming capability and stability.Neural network has superior self-learning ability and robustness and through trainig it can approximate any nonlinear function.A 2-layer neural network is designed according to multilayer feedforward network model.Appropriate learning rate parameters are selected and BP-based speed identification algorithm is designed. The results for simulation experiments show speed identification using neural network has excellent static and dynamic performance and good anti-jamming capability.Contrastive analysis are made when the three proposed methods are applied in different condition and parameters and with noise. The simulation results show the speed identification based on MRAS and Neural network can acheive rapid response,good following performance but its performance is limited by stator resistor. The one based on extended Kalman filter can achieve good stability and anti-jamming capability,but its performance mainly deponds on motor parameters.