Yes, the devil is in the details. I guess 2kW per square meter of the working face, in a four knot current, is a decent output; and tidal currents are reliable, if intermittent. 500 sq meters to produce a megawatt is a very big machine, and the moving parts are complex.
I'd like to OP to explain whether this approach has any inherent advantages over a shrouded turbine, as used by most hydro stations I know of. Something to do with low-speed flow?