基于模糊控制及神经网络的开关磁阻电机转子位置估计

摘要

The operation of switched reluctance motor (SRM) requires a position sensor, but the position sensors will increase the system cost and complexity and reduce the reliability. Thus, carrying out rotor position estimation and making the SRM run independently from the position sensors is worthwhile researching. According to the feature of neural network which can approach any nonlinear function precisely, a rotor position estimation model based on fuzzy logic control model and neural network theory was established. Moreover, on the MATLAB/Simulink platform, the SRM rotor position estimation system with 60 kW and 6/4 pole was simulated. The simulation results show that, this estimation system can accurately predict the position angle of the SRM rotor, with an error of 2° and mean square error of 0.7144. Furthermore, this estimation system was used to control the SRM, and the results show the motor's three-phase current output is uniform, the torque ripple is small, and the whole operation process is stable.%开关磁阻电机(SRM)运行需要转子位置传感器,而添加转子位置传感器使系统成本以及电机结构复杂度提高,可靠性降低.因此,对SRM转子位置进行估计,使SRM不依靠位置传感器而独立运行具有重要意义.本文利用神经网络对非线性函数高精度逼近的特性,基于模糊控制模型和神经网络理论,建立了SRM转子位置估计系统,并采用MATLAB/Simulink对该60 kW、6/4极的SRM转子位置估计系统进行了仿真.仿真结果表明:该系统对SRM转子位置角的估计较准确,输出角度精度较高,误差在2°左右,均方误差为0.7144;利用该转子位置估计系统对SRM进行控制,该电机三相电流输出均匀,转矩脉动小,运行全过程稳定.