To meet the rising comfort requirements of non-residential buildings like office buildings mechanical ventilation in the form of e.g. air handling units (AHU) is needed. As of today, the fans in AHUs are designed for a specified operating point. The real operating range does rarely match this design point, which causes the fans operating in ranges with suboptimal efficiencies. This leads to higher electrical consumption than needed and therefore higher emissions of CO2. In this work we propose a control strategy which can be applied to an AHU where several smaller fans are operated in parallel. We show that when this fan array as a whole is operating in a suboptimal operating point, it is beneficial to switch off one or more fans. This shifts the remaining fans to a operating point with higher efficiency, resulting in an overall higher efficiency. The experimental AHU is equipped with four parallely mounted radial fans and additional dampers upstream of the fans to prevent backflow in case a fan is switched off. The electric consumption of each fan is measured separately to calculate a global system efficiency. Several fan performance maps are generated from measured data, which enable the controller to predict the system efficiency and decide on the optimal number of active fans. We benchmark the new control strategy against the standard strategy with no fan-shutoff. The use case covers a demand controlled load profile, which is based on a generic office building with parallel ventilation branches which can be shut off separately.