The alternator provides power to vehicle electrical loads with the battery, and its maximum current depends on various factors such as electrical load, engine speed, thermal condition, and other variables. Above all, thermal effects make alternator simulations more complicated. For example statically similar conditions may show different results according to the temperature variation for each alternator operation. This paper proposes a two-stage statistically-based model structure which separates dynamic thermal effects from steady state performance. The method was validated by experiments and shows good predictive performance, suitable for use in test reduction.