At present, nitrogen production from air by pressure swing adsorption (PSA) is simulated almost exclusively at low product purity levels (< 99 % N2). However, with increasing global demand for highly purified gases provided by energy-efficient separation processes, the requirement for either extensive experimental research in the high-purity range or predictive computer simulations arises. Moreover, with increasing nitrogen purity, PSA plants require an over-proportional air demand with the consequence that high-purity PSA systems engender a distinct interest in energy-saving measures. This dissertation presents a mathematical model of a twin-bed PSA plant equipped with a carbon molecular sieve (Shirasagi MSC CT-350) adsorbent for the generation of high-purity nitrogen (99.9 – 99.999 % N2). The corresponding model is implemented in the process simulator Aspen Adsorption™. The influence of operating conditions, cycle organisation, as well as plant design on the PSA process performance is validated. Specifically, effects of adsorption pressure, operating temperature, half-cycle time, purge stream flow rate, cutting time, flow resistances, and volume of N2-receiver tank are studied. The precision of the performance prediction by numerical simulations is critically discussed. Based on the new insights, efficiency improvement strategies with a focus on reduced energy consumption are introduced and reviewed. Finally, a future outlook on the research is presented.