Faults occurrence is inevitable in real world and a practical alternative approach is the reduction of fault consequences. Despite of this, the majority of buildings control systems do not have such mechanisms. Thus, this work proposes a procedure for the modeling of control systems for intelligent buildings which considers their functional specifications in normal operation, and in case of faults occurrence. The procedure adopts the concepts of discrete event system, holon, and Petri net and its extensions. It is presented some models derived from case studies, and mechanisms to fault-diagnosis, decision and reconfiguration.