Power Management Strategies (PMSs) to control stand-alone energy systems affect the reliability of meeting load demand as well as the cyclic operation of various subsystems. The hybridisation of sources through the integration of hydrogen fuel cells with energy storage means optimising the PMS should be "intelligently" done unless relying on rule-based PMSs which are simplistic to use but subject to lack of optimisation. This paper presents the methodology and validation of a lab-scale (desktop) energy system controlled by a predictive PMS. Validation of the intelligently based PMS can be done in the lab-scale before (costly) full deployment in the field, but experiments to support this have not been reported in relation to hydrogen systems. The experimentally tested hybrid energy system consists of an emulated renewable power source which can represent solar-PV and/or wind generators, battery bank and PEM fuel cell integrated with metal hydride storage. Experimental testing as well as the use of real-time predictions using Neural Networks is utilised. The effects of several control parameters which are either hardware dependant or affect the predictive algorithm are investigated with system performance, under the predictive PMS, benchmarked against a rule-based PMS. The results reveal that a predictive PMS is impacted by the prediction horizon used to forecast the availability of renewables or load, the decision time interval used for updating the PMS as well as time lags resulting from hardware sensors used to convey system status to the decision algorithm responsible for updating the PMS. The maximum thresholds of the above mentioned control parameters are 120, 15 and 3 s, respectively. Beyond these limits, the ability of the predictive PMS to effectively control the system degrades significantly. This study demonstrates the feasibility of using real-time predictions of renewable resources and load demand to optimise a PMS in a stand-alone energy system and experimentally validates this, which has not been previously reported. Copyright (C) 2015, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights reserved.