Well, one wild but not physically impossible idea I heard on the boat design forums, by Michael Praamsa, IIRC, was to use the latent heat of the ocean as a source of power. The ocean surface has a thin layer of humidity which prevents further evaporation, and of course the ocean's surface-to-volume ratio is quite low. By creating a spray or using a wick structure, water is evaporated and air is heated. Doing this inside a ducted fan windmill, the volume of air going out can be greater than that entering, it can make its own wind, which will be at a high velocity at the neck of the duct, enhancing evaporation. The energy is collected by a turbine rotor at the front (increasing turbulence and evaporation downstream. The energy can then be used to run any sort of marine propulsion. Evaporative cooling of the machine is a side benefit which can be used for refrigeration or air conditioning.
Another idea is to charge the droplets from the spray, allow the wind to accelerate them (well, really just move them away from the charged nozzle, making them attain a much higher voltage by capacitive voltage multiplication) and use the ocean as the current return path, putting a load in between. This could give a large wind-capture area as the droplet stream spreads out. Prof. Steven Salter had a great design for nozzles that would be ideal for this purpose, turning water with over twice the volume of the easily-manufacturable atomizing apparatus into 30-micron charged droplets each second, using only 60.5kW of solely mechanical power for each cubic meter of water per second. That's creating 11.9 km^2 of evaporative surface per kWh. Just lifting the same volume of water 10m would consume about 1.65 times the energy. ( http://www.mech.ed.ac.uk/research/wavepower/rain%20making/shs%20rain%20paper%20Feb.pdf - note that the design in this paper does not work for the original purpose of increasing rainfall - enhanced humidity near the surface suppresses ocean evaporation overall.)
Theoretically this idea could be turned around if one already had electrical energy - the droplets could be electrostatically accelerated, entraining air to create thrust, but I doubt it would be efficient.
