Experiments in Long Timeframe Timelapse

I am embarking on a large scale, long timeframe time lapse project that will require cameras in remote locations to function for months and years on end.  There are a few commercial systems on the market but their price is seems abnormally high for the hardware involved.

I am trying to engineer a simple but robust version from off the shelf components.  The current idea looks something like this: