An improved neural networks-based vector control approach for permanent magnet linear synchronous motor

A permanent magnet linear synchronous motor (PMLSM) control strategy based on improved neural networks is proposed for the problem of inaccurate decoupling and poor adaptive performance of the traditional proportional-integral (PI) control method. In this paper, incremental broad learning (IBL) is selected as the neural networks training model. IBL is improved on the basis of broad learning, which can dynamically change the number of nodes while reducing the model training time. The IBL controller design process does not require a mathematical model of the control system. It can be designed based on data obtained from existing systems or factories without the necessary expertise, and can better enhance the nonlinearity of the controller. First, the IBL controller is trained and built by collecting relevant data based on the entire dynamic model of PMLSM. Then, test datasets and error evaluation metrics are used to determine the optimal hyperparameters of the IBL model. Finally, simulations and experiments in dSPACE show that the method proposed in this paper outperforms the conventional vector control method in terms of robustness and adaptivity.