In statistical load modeling methodologies, aggregate electric load behavior is derived by propagating the ensemble statistics of an individual load process which is representative of the loads in the aggregate. Such a modeling philosophy tends to yield modets whereby if physical meaning is present at the elemental level, it is preserved at the aggregate level. This property is essential for applications involving direct control of power system loads (for peak load shaving purposes, for example). The potential applicability of statistical load models is a strong function of one’s ability to limit the volume of unusual data required to build those. An identification algorithm for a previously proposed stochastic hybrid-state Markov model of individual heating-cooling loads is presented. It relies only on data routinely gathered in power systems (device energy consumption over constant time intervals). It exploits an alternating renewal viewpoint of the load dynamics. After deriving some general results on the occupation statistics of time homogeneous alternating renewal processes, the analysis is focused on the specific model. In the process, however, some intriguing features likely to be shared by a wide class of alternating renewal processes are revealed.