A Virtual MPC-Based Artificial Neural Network Controller for PMSM Drives in Aircraft Electric Propulsion System
Shengzhao Pang, Yonghui Zhang, Yigeng Huangfu, Xiao Li, Bo Tan, Peng Li, Chongyang Tian, Sheng Quan
Abstract
Model predictive control (MPC) has great potential in PMSM drives due to the advantages of fast dynamic response and multi-variable control. However, due to its exponentially increasing computational load and a large number of online calculations, it greatly increases the computational complexity and resource consumption of the microcontroller. Therefore, overcoming the barriers of computational burden has become a key point for the large-scale application of MPC strategies. This article proposed a novel virtual MPC-based artificial neural network controller (ANN-MPC) for PMSM drives in aviation electric actuators, to reduce computational burden and improve the system control performance. Firstly, a traditional MPC controller is designed under circuit simulation to generate the input and output data for training. Next, the design of the ANN-MPC controller is trained offline with massive training datasets. The ANN-MPC controller replaces the heavy online calculation of the MPC controller through simple mathematical expressions, so the ANN-MPC controller significantly reduces the computational burden and resource consumption. Moreover, the simulation and experimental results reveal that the proposed ANN-MPC controller has an approximate control performance compared to the conventional MPC controller.