The methylcyclohexane (mch) dehydrogenation step to recycle toluene and release hydrogen (hyd) for various applications is being studied as part of a hydrogen energy storage project. The reaction is performed catalytically in a fixed bed reactor, and the efficiency of this step significantly determines overall system economics. The fresh catalyst kinetics and the deactivation of the catalyst play an important role in the process analysis. To predict the catalyst performance at various coke content levels, laboratory experiments at different scales were carried out in three stages : 1) isothermal experiments at low mch conversions in a differential microreactor; 2) non-isothermal experiments in PC-controlled integral reactors; 3) increasing temperature experiments in two integral reactors in series to deposit significant coke levels. A wide range of operating conditions were used : 200-400 hours continuous operation, high liquid hourly space velocities (LHSV) of 3-20 1/hr., 0-10 hyd/mch feed ratio, 6-16 arm overall pressure, Z50 degrees C - 550 degrees C temperature in the reactors. Axial species concentrations and temperature profile measurements at different pressure levels and time-on-stream allowed a study of the catalyst deactivation effects for mch conversions up to 99%. The main reaction kinetics were selected by using a parameter sensitivity analysis for model discrimination and the kinetics short-cut identification software package of Maria (1993), Maria and Rippin (1995). The resulting two-parameter model was further tested with experiments in the dual integral reactor system. The activation energy for the main reaction was estimated at 220 +/- 11 kJ/mol. A similar sensitivity analysis was used to derive a reduced model for catalyst deactivation, by using data recorded in the dual non-isothermal integral reactor system. The best results were obtained with a zero-order deactivation kinetics. A simple dependence between the catalyst activity and the coke content on the catalyst (up to 17 wt% for significant time-on-stream) was tested.