Yes, you are on the right track here. You want to decrease the element size until you reach a converged solution. It isn't a definite element size but you will know it when you can change the element size with negligable change in results.
Also, it is worth noting that, for thermal analysis, that some boundary definitions are not good/accurate reflections of real life. A "Temperature" boundary doesn't have a place in real life so if you have any of these you might want to find the appropriate convection/heat generation/etc. to replace that with.