In Shunt Active Power Filter, the DC level applied to load caps can be switched ΔV to match the level of current, needed to cancel harmonic load noise from Ic = C ΔV/Δt when in sync with the noise.
From a different student thesis with ref to PQ theory by Akagi.
Essentially an AC to DC converter is needed to power the harmonic active filters and these DC losses must be minimized while attenuation and losses are inverse to the Vdc created for active filtering. The IEC guidelines are <5% THD out.
As we know transferring DC charges from one capacitor to another incurs I^2R loss, so too this applies in correcting AC harmonics with active current negative feedback using the DC converted source. So the filter losses; include DCR of coil and Rce of IGBT's and ESR of Caps with all real and stored energy currents in the filter creating losses.
Thus cost and Active attenuation are tradeoffs with component quality related losses.
Note this thesis improved THD from 30% to 1%, but only above the 5th harmonic (=250Hz) on 440Vac but did not indicate any actual Pwr levels or efficiency. Such omissions are common in academic papers. But they did mention DC issues that need resolution with imbalanced harmonics on 3 phases by "pulse swapping" on phases in the last sentence.
Also we have no idea how much this improves harmonic attenuation over passive methods if bulk low ESR plastic VAR caps that come in 19" racks.
It left some things to imagination, but the PQ theory is interesting. e.g.