All examples on the net use the temperature example because that is the standard terminology for simulated annealing-- SA is a physics-inspired technique, modeled after a real-world phenomenon called annealing. It is much the same as how all examples for genetic algorithms talk about genes and chromosomes.
If you trace the mathematics back far enough, there are some fascinating connections between various optimization meta-heuristics and some physical processes, usually bridged by the notion of entropy.
But, in very rough terms, the temperature T in simulated annealing corresponds to the willingness or ability of the algorithm to "jump" out of a local minimum in the search for a global (or at least, better local) minimum. High temperatures correspond to higher randomness, jump around more, and may even end up in worse configurations; low temperatures correspond to lower randomness (and eventually purely greedy algorithms) and cannot escape any local minima no matter how shallow.
As to how to use that idea for your applications, well. It takes some insight and some creativity in order to get most metaheuristics to work right. ut you're never going to find a discussion of SA that doesn't talk about temperature.