Brendan - I don't have a good answer for your question, but just saying hello.  We are working on an un-funded project here related to UAA to create an Arctic power conversion board, but the concept is to use fuel-cell (PEM) power from H2 combined with renewable sources as available (wind, solar, tidal) as well as grid/generator for testing convenience.  It's a very modest design for 20 Watts to power sensor network camera and uplink nodes (composed of a Jetson and 3 cameras enclosed in a box).  The common wisdom is that batteries just won't work reliably below -40C so transient power storage uses ultracapacitors as we deal with load/generation dynamics.  When I was researching all of this, I found Battery University to be quite useful - http://batteryuniversity.com/  I'm new to the technology as well, but this helped me to understand fundamental limits and concepts.  Hope it helps.  I'll be up in Anchorage sometime this summer to test the camera, so if you're around, it would be great to catch up.

Thanks Sam. Interesting work. I am also interested if there could be a breakthrough in supercapacitors so that you could use them in a variable energy system using renewables to handle sudden drops in energy. Enough for back up reserves to spin up or some combo with batteries, on demand load control and other tricks to get a higher penetration of renewables.  Possibly using predictive analytics on load and supply to spin up reserve even if it meant false positives but greater penetration. It would be great to say hello this summer. 

A worthy research project indeed - I wish I could say we have the intelligent power conversion board for you to use to try out ideas, but we don't - we had a student team that has made an effort as a project, but not operable - I am considering applying for a small grant to complete and build the platform - DC/AC inputs from external sources and fuel-cell input, microcontroller for source and load monitoring (switching really does not appear to be needed - all resolved at power-electronics circuit layer) - I would need collaboration form a true power electronics expert (working on that), but likewise interested in how to power sensors up to 20 Watts for a full year with no batteries - the goal - I have seen H2 fuel-cell powered UAV demonstrations that look quite promising as well (e.g. http://www.uasvision.com/2016/01/18/fuel-cell-quadcopter-flies-2-hours-with-4k-video/ - not sure how well verified, but I've seen similar claims and there are some older published peer reviewed papers - http://www.engr.colostate.edu/~thb/Publications/2006-01-3092.pdf).  Anyway, our camera system uses 2 Watts standby and about 10 Watts for 3-camera continuous transform processing in real-time.  So, overall, we'd go between stand-by and active and based on duty cycle, be somewhere just below 10 Watts average - it's a fair amount of H2 to feed the PEM, but not totally unfeasible for applications of interest to us.

How far away are the sensors? Could some kind of mesh network be used to transmit data between them? I guess it wouldn't save much energy. I used to work at ACEP at UAF and they might have some resources or testing facilities. http://acep.uaf.edu/ Also is it possible to send a drone to refuel H2 to your devices via some kind of cartridge system. Or generate H2 when you have excess renewables?

The camera nodes are the uplink to the Cloud - the smaller sensors are in a mesh and send data through the mesh to camera nodes for uplink - at least that's the concept.  The cameras themselves probably would not mesh as they'd be too far apart (miles or more).  The theory is that the USCG or other security personnel would refuel them once a year and swap out any parts needed or swap in a whole new unit and take the defective unit back for repair.  But, it's just a concept.  For buoys there may be enough tidal power and reliable enough to just use that, although bouys will also get trapped in sea ice.  Definitely a challenging problem - thanks for the suggestions.  I will look into the ACEP at UAF - we are working with them, so much appreciative for the pointer.

Sam, this wouldn't solve your problem because of the winter and other factors, but your question did remind me of visiting my friend Jan at the Bristol Robotics Lab in 2007 and they had a separate room they kept shut that was working on powering robots and sensors with decaying flies. It was one of the most pungent smells I have ever experienced but they were able to attract flies and they would get stuck and die and fuel a powercell. If you could do it for mosquitoes in some regions you would get good power and points for reducing the mosquito population. http://www.brl.ac.uk/brlresearchprojects/microbialfuelcells/ecobotii.aspx