These temperatures were created (by me) as part of the very first edition of the PlanetsLib library. It is actually part of the origin story of the library: PlanetsLib allowed for the deprecation of https://mods.factorio.com/mod/Surfaces-Have-Temperature, a mod I created to support Hard Aquilo in my Space Age Hard Mode series. By the time of the Cerys release, there were multiple mods starting to define planet temperatures in contradictory ways and it was causing a bit of a mess. (This is not the only purpose of PlanetsLib, certain issues regarding the starmap had a similar coordination problem.)
Since that date it has been pointed out to me by Snouz, my collaborator and the author of Moshine, that other temperatures in Factorio use Celsius so it would have been better in Celsius than Kelvin. I can see both ways on that, perhaps it is my background in physics that leads me to think usage of Kelvin on astrophysical factsheets is more appropriate. A proposal to change PlanetsLib to Celsius has not been submitted by anyone, perhaps for the reasons I will explain next.
Planetary temperatures in Factorio have a predicament which is analogous to that of interplanetary distances. These predicaments are not noticed except by a small fraction of the audience, do not impact gameplay themselves, but flavor-wise involve the needs of the game pulling in multiple contradictory directions.
For interplanetary distances, the three contradictory pulls are 1) Factorio wants to represent an entire solar system that feels like a real solar system (i.e. huge distances when measured in km), 2) Factorio has a thruster entity which gives velocity to space platforms and the resulting speed should feel somewhat in concert with the physicality of the thruster, and 3) Factorio wants to be a game played in a sitting by a human being without spending weeks on every interplanetary mission. Therefore, if Factorio wants to maintain the s = v * t formula of kinematics, it needs to find which is most palatable and least noticeable out of unphysical distances, unphysical velocities and unphysical times. Wube chose #1, perhaps correctly because complaints about this on Reddit are quite rare.
For temperatures, the contradictory pulls are 1) You can build factories on planets which can process materials, and these machines and materials have realistic operating and working temperaturs, 2) The temperatures should realistically match astrophysical bodies, and 3) the engineer can walk on the surface of the planets unaided and does not recoil in horror when reading the factsheet of an upcoming planet.
There are lots of decisions one needs to make when launching one of the very first modded Factorio planets, and on balance my decision to respect #3 was not what I would do today. Although, similar to with large interplanetary distances, we have gone 10 months of it being very rare for someone to comment on it — partly because the temperatures are expressed in Kelvin which most players today have a worse intuitive grasp on than Celsius — the choice is probably less robust in the long-term because there are a great variety of possible materials and environments that one could represent with modded planets, but for the engineer you only have to 'cheat' a single time by imagining they have invulnerable plot armor.
I realized this about 9 months ago but have not entertained any proposals to fix it, partly as a contributor in my free time I choose what problems I try to solve very selectively, and partly because I do not have or have received any proposals for a workable migration strategy given the fairly large and unknown number of mods that depend on specific numerical values of the PlanetsLib temperatures.
