Research on optimization algorithm of BP neural network for permanent magnet synchronous motor based on Cloud Computing

The low-efficiency design method of traditional motor usually needs a long time, so Cloud Computing is introduced for the traditional FEM and BP neural network to reduce the time consuming. In this paper, the electromagnetic axial torque of permanent magnet synchronous motor (PMSM) is used as calculation case. The utilizing Cloud Computing curtails approximately 4 times hour to calculate this case. The Cloud Computing can achieve mixed computing between windows and Linux, with running MATLAB on windows and running COMSOL on Linux in the form of non-GUI. BP network training needs hundreds of group samples. Though it won't takes long time that a group FEM calculation data is achieved, hundreds of group data will cost a plenty of time. High performance cloud computing is utilized for shortening samples achieved time, synchronous parallel computation is executed on the cloud elastic cluster, which can obtain multi-group samples of PMSM in the meantime. Then achieved data regard as BP network input and output samples, which is used for training and learning. The results show that motor optimization combine cloud computing with BP algorithm has feasibility and high-efficiency.