I've done designs for over a hundred miles of fencelines in applications like this, mine all revolved around video analytics (of course), but a few things I've learned over the years:
There is no "optimal" product for all environments, they all have pros and cons. Some examples: climates with a lot of freeze/thaw ground shift can impact buried cable, areas in hot climates don't do well with thermal cameras, very dusty environments can cause problems over time with optical solutions (cameras or beams).
I would highly recommend doing some kind of long-term (1 month or longer) on-site test of one or more technologies before committing to a larger deployment. Pay attention not just to object detection, but also object rejection. A 500 meter range is useless if it triggers on anything that moves in the distance. A similarly priced product with a 200 meter range, but much higher accuracy will likely cost less to operate over time by reducing false alarms, even though you would need more devices in total.
No product I have come across is reliable enough to trust an alarm 100%, this means you will likely want cameras for verification purposes. In some cases a slew-to-cue PTZ will work, but most of the time the PTZ movement will be too slow and you wind up with a lot of cases of empty video, which makes troubleshooting false alarms WAY harder.
For electrical substations, and similar remote unmanned sites, multi-tier approaches are often used: video analytics looking beyond the fence line (the theory being that nobody generally goes near these sites, so anyone even coming close is worth knowing about and being aware of), and possible fence detection sensors, beams, etc on the actual perimeter, setup in an "AND" trigger fashion to generate a higher priority alarm (meaning that the video analytics AND fence sensor both need to detect something before an intrusion alarm is activated, particularly if doing some kind of automated lights/siren).
Be sure to pay attention to maintenance costs (recalibration with some analytics products, trimming along fence lines for some fence products, etc.) when factoring in TCO.
More likely than not you will require some amount of support from the manufacturer for setup, calibration, tweaking, whatever. Does the solution have solid local tech support? 24/7, or just normal business hours? Local support is doubly important for critical infrastructure particularly because getting remote access to these sites for off-site troubleshooting ranges from insanely difficult to impossible much of the time.
Systems like this are usually about getting an alarm event in front of an operator efficiently. Pay attention to upstream bandwidth at the site, so that events can be pushed out. For the actual alarms generated - do they contain images and/or video? Are the images clear enough to provide usable info? Does the event notification come through quickly enough to respond before the person damages equipment/property? Can they be opened on a mobile device? Can the operator initiate any kind of a remote response (voice-down, dispatch guards, turn on lights)?
Personally, I prefer sensor options that can cover a large area, and are not just "tripwire" based. There is too much opportunity to miss something if you are only looking across a very specific area or narrow zone. This can be video analytics, or other forms of adapted motion detectors that cover wide areas.