新能源汽车的开关磁阻电机神经网络预测控制
|
摘要
开关磁阻电机(SRM)常用作新能源电动汽车的驱动电机,但其转矩脉动大,影响整个调速系统正常工作,此处提出一种基于神经网络预测控制减小SRM转矩脉动的方法。首先,搭建系统数学模型,用建立的电机模型根据前一时刻和当前时刻状态值和采样的定子电流、转子位置及母线电压预测出下一时刻的转矩值;其次,定义一个基于预测转矩与参考转矩误差和相电流双目标控制的代价函数;然后,提出用离散空间矢量的方法评估一系列虚拟状态值,结合代价函数得到一个最优的虚拟状态值;最后,在神经网络预测控制系统中引入该虚拟状态值,实现对SRM转矩脉动的控制。为验证所提控制策略,在Matlab中对SRM转矩控制系统进行仿真,此外,搭建15 kW的SRM实验平台,仿真实验结果均表明,所提控制策略正确、可靠。
Switched reluctance motors(SRM) are often used as drive motors for electric vehicles,but their torque ripple are large.Therefore, a method to reduce the torque ripple of SRM based on predictive control is proposed.Firstly,the mathematical model of the SRM is established, and the torque value of the next moment is predicted according to the previous time and the current time state value stator current,rotor position and bus voltage.Sencondly,a cost function based on the predicted torque and the reference torque error defined.A series of virtual state values are evaluated by discrete space vector method and an optimal virtual state value is obtained by combining the cost function.Finally, the virtual state value is introduced in the neural network predictive control system to control the torque ripple of the SRM.In order to verify the feasibility of the proposed control strategy,a SRM torque control system is simulated in Matlab.In addition,a 15 kW SRM is used as an object to build a test platform.The results show that the proposed control strategy is correct and feasible.
Switched reluctance motors(SRM) are often used as drive motors for electric vehicles,but their torque ripple are large.Therefore, a method to reduce the torque ripple of SRM based on predictive control is proposed.Firstly,the mathematical model of the SRM is established, and the torque value of the next moment is predicted according to the previous time and the current time state value stator current,rotor position and bus voltage.Sencondly,a cost function based on the predicted torque and the reference torque error defined.A series of virtual state values are evaluated by discrete space vector method and an optimal virtual state value is obtained by combining the cost function.Finally, the virtual state value is introduced in the neural network predictive control system to control the torque ripple of the SRM.In order to verify the feasibility of the proposed control strategy,a SRM torque control system is simulated in Matlab.In addition,a 15 kW SRM is used as an object to build a test platform.The results show that the proposed control strategy is correct and feasible.
引文
[1]朱曰莹,赵桂范,杨娜.电动汽车用开关磁阻电机驱动系统设计及优化[J].电工技术学报,2014,29(11):88-98.
[2] S Mir,M E Elbuluk, I Husain.Torque-ripple Minimization in Switched Reluctance Motors Using Adaptive Fuzzy Control[J].IEEE Trans. on Industry Applications, 1999,35(2):461-468.
[3] Z Lin,D S Reay,B W Williams,et al.Torque Ripple Reduction in Switched Reluctance Motor Drives Using B-spline Neural Networks[J].IEEE Trans. on Industry Applications,2006,42(6):1445-1453.
[4]夏长亮,陈自然,李斌.基于RBF神经网络的开关磁阻电机瞬时转矩控制[J].中国电机工程学报,2006,26(19):127-131.
[5] Chen T,Chen H. Approximation Capability to Functions of Several Variables, Nonlinear Functional,and Operators by Radial Basis Function Neural Networks[J]. IEEE Trans. on Neural Networks, 1995,6(4):904-910.
[2] S Mir,M E Elbuluk, I Husain.Torque-ripple Minimization in Switched Reluctance Motors Using Adaptive Fuzzy Control[J].IEEE Trans. on Industry Applications, 1999,35(2):461-468.
[3] Z Lin,D S Reay,B W Williams,et al.Torque Ripple Reduction in Switched Reluctance Motor Drives Using B-spline Neural Networks[J].IEEE Trans. on Industry Applications,2006,42(6):1445-1453.
[4]夏长亮,陈自然,李斌.基于RBF神经网络的开关磁阻电机瞬时转矩控制[J].中国电机工程学报,2006,26(19):127-131.
[5] Chen T,Chen H. Approximation Capability to Functions of Several Variables, Nonlinear Functional,and Operators by Radial Basis Function Neural Networks[J]. IEEE Trans. on Neural Networks, 1995,6(4):904-910.