Single solution sphere-decoding algorithm model predictive control for high-current applications

This paper presents a Sphere-Decoding algorithm (SDA) Model Predictive Control (MPC) for a parallel-connected H-Bridges Power Supply (PS). The proposed converter topology faces the very high current peaks (tens of kiloamperes) required by Central Solenoid coils of the Divertor Tokamak Test (DTT) facility for nuclear fusion, quite unusual in industry applications. The choice of the control strategy aims at exploiting the very fast transient response of MPC over linear control schemes and the computational burden reduction of SDA. As a result, this approach is able to guarantee a low load current tracking error and an effective current sharing among H-Bridges, thus proper operations for tens of years. In order to implement the SDA-MPC on a FPGA-based control board, fast but characterized by limited memory, the mathematical model of the PS is first introduced and the SDA-MPC procedure is then mathematically modified to find a single optimized solution. This simplification guarantees a remarkable reduction of the computational burden, avoiding the analysis of a set of possibilities, without losing in control effectiveness. Its performances are verified through simulations and experimentally validated with Hardware-In-the-Loop and prototype tests.