Decoupled Multistep Model Predictive Control of Permanent Magnet Synchronous Machines

We propose a novel multistep model predictive current control framework for synchronous machines. The proposed framework is based on the mathematical structure of the system model in different coordinate systems. Since the dynamics of the stator currents of the machine are independent of each other, it is possible to decouple them within the cost function in stationary coordinate frames. Based on this properties, a multistep prediction model is established. In order to obtain the optimal solution efficiently, we furthermore propose a specialized sphere decoder-based algorithm that is tailored to decoupled current control. Compared to existing approaches the proposed method does not require any matrix inversion ir decomposition for the formulation of the optimization problem. For this reason it is especially suitable for electric drive systems with large and or rank deficient system matrices. The framework is verified with the help of numerical simulations.