[EN] The target of the paper is to study how to devise an efficient discrete-event model for the yearly energy simulation of buildings. Conventionally, software tools use time-driven schemes and many components must be computed at every sampling time-point. Event-driven simulation aims at lowering this burden, by calling only those components whose state is evolving quickly. The article explores a model based on DEVS formalism and Quantized State Systems (QSS) techniques. Within this paradigm shift, our strategy was to reuse as much widely accepted knowledge as possible. One immediate difficulty was, that the well-known conduction heat transfer function (CHTF) of multi-layered walls is not suitable for DEVS in its traditional form since it is constrained to sample at a fixed time step. Instead, the paper introduces a non-conventional method: the Successive State Transition method (SST). Its distinguishing traits are: it allows variable time steps, has high accuracy and its computational workload adapts to the elapsed time between transitions. Unfortunately, although we found that SST and QSS work well together, the paper shows that the aforementioned transfer function is not adequate for event-driven simulations. Based on the paper outcomes, we propose a workaround for further research: a new transfer function, relating the conduction heat flux (input) to the time derivative of the wall superficial temperature (output) (recall that the traditional input-output relationship: superficial temperature and conduction heat flux, respectively).