This papers proposes a novel climate control strategy for mine extraction rooms based on the receding horizon optimal control scheme. Being a model-based procedure, the development of a pertinent prediction model is one of the keystones. According to recent technological advances, we consider that distributed measurements are available and provided by a wireless network. An enhanced modeling approach, based on stratification and sigmoid description of concentrations in the extraction rooms, is then proposed and allows for an optimal use of information provided by the wireless sensor network (WSN). The complexity of the resulting model, due to the nonlinearities, different time scales and time-delays, is handled by using an on-line shape prediction, included in the design of an optimal sequence of control actions over a finite horizon. Physical and communication constraints are successfully handled at the design stage and the resulting closed-loop system is robust with respect to variations in the pollutant dynamics.