Preselected Adjacent Vectors with Sphere Decoding Algorithm for Fast FCS-MPC in Three-Phase 2L-VSI

Finite control set model predictive control (FCSMPC) is a widely used formulation of MPC in power electronics converters that requires higher switching frequencies and, thus, often limits its prediction and control horizons for a single sample. Then, reducing the computational burden is one of the main goals when dealing with this formulation, since it can enable achieving better results by increasing the size of the horizons employed. This paper approaches this problem by building on two established concepts: the subset of adjacent vectors (SAV) and the sphere decoding algorithm (SDA), which optimizes computations by reducing the search space. A novel technique, the preselected adjacent vectors with sphere decoding algorithm (PAV-SDA), is introduced. The proposed approach is implemented in a three-phase two-level voltage source inverter and evaluated through simulation. The findings demonstrate that the PAV-SDA achieves a balance between computational efficiency and performance. For longer horizons, the PAV-SDA significantly reduces computational effort compared to the standard FCSMPC while maintaining the same THD and MSE values.