基于神经网络的永磁同步电机模型预测电流控制  

Neural-network-based model predictive current control forpermanent magnet synchronous motor

作　　者：李耀华[1] 刘东梅 陈桂鑫 刘子焜 王孝宇 童瑞齐 LI Yaohua;LIU Dongmei;CHEN Guixin;LIU Zikun;WANG Xiaoyu;TONG Ruiqi(School of Automobile,Chang’an University,Xi’an 710064,China)

机构地区：[1]长安大学汽车学院,陕西西安710064

出　　处：《电机与控制学报》2024年第10期109-122,共14页Electric Machines and Control

基　　金：陕西省自然科学基金(2021JM-163);西安市碑林区科技计划项目(GX2252)。

摘　　要：针对备选电压矢量有限导致永磁同步电机有限集模型预测电流控制性能较差及计算量较大的问题,提出基于神经网络的永磁同步电机模型预测电流控制。基于7个基本电压矢量和121个扩展电压矢量的永磁同步电机模型预测电流控制分别建立7分类和121分类神经网络。随着备选电压矢量的增加,模型预测电流控制性能提升,对应的神经网络控制性能也得到改善,但分类任务数也随之增加。对于多步模型预测控制,计算量随步长呈指数上升,但输出电压矢量不变。因此,基于两步模型预测电流控制建立7分类神经网络。仿真结果表明:以上神经网络控制均可行,性能与相对应的模型预测电流控制基本相当。实时性实验结果表明相较于单步模型预测电流控制,神经网络控制并不占优势,但相较于两步模型预测电流控制,神经网络实时性有明显优势,计算耗时减小29.58%,表明神经网络控制更适于多步模型预测电流控制。Aiming at the problems of poor performance of finite-control-ser model predictive current control(MPCC)for permanent magnet synchronous motor(PMSM)caused by limited candidate voltage vectors and large calculation burden,a neural-network-based MPCC for PMSM was proposed.Based on the MPCC for PMSM with 7 basic voltage vectors and 121 extended candidate voltage vectors,the neural networks with 7 and 121 classification tasks were established.With the increase in candidate voltage vectors,the control performances of MPCC and the corresponding neural network were improved,but classification tasks were increased,too.For multi-step control,calculation burden will increase exponentially with the increase of steps,but output voltage vectors will not change.Therefore,a neural-network with 7 classification tasks was established based on two-step MPCC.Simulation results show all proposed neural networks operate well.And neural networks’control performances are almost the same as the corresponding MPCC.Real-time experiments show that compared with one-step MPCC,the real-time performance of neural network is worse. But compared with two-step MPCC, the real-time performance of neural networkis better and its calculation time is decreased by 29. 58%. Thus, neural network is more suitable formulti-step MPCC.

关 键 词：永磁同步电机 模型预测电流控制 神经网络 备选电压矢量 实时性 多步预测 

分 类 号：TM351[电气工程—电机]