Multisampled model predictive control of inverter systems: A solution to obtain high dynamic performance and low distortion

At low switching frequency, the dynamic performance of most pulse-width-modulated schemes is limited due to the limited sampling rate, while direct control schemes feature a good dynamic performance but poorly controlled switching frequency and harmonic performance. The paper proposes a new multisampled model predictive control scheme suitable for the control of low switching frequency AC inverter systems such as employed in high power. The scheme deals optimally with the switching constraints that arises when the duty cycles are modified more than twice in each switching period and it eliminates the inverter ripple with a ripple observer, which improves dynamic performance compared to filter based solutions. It achieves lower distortion than direct control schemes, while featuring similar dynamic performance and lower computational complexity.