Modelling and simulation are essential tools for concept evaluation and for predicting the performance of a hybrid energy system, since prototyping and testing each candidate design for such a complex system would almost always be prohibitively cumbersome, expensive and time consuming. To meet the modelling and simulation objectives, the various components of the system (sources, storage, loads, and converters) need to be characterised and modelled in a tractable way. The tuning of the models to reflect the actual system components is a key milestone in this process and requires reliable and comprehensive experimental data. Furthermore, environmental conditions such as ambient temperature may have a significant impact on the performance, which has to be taken into account. The complexity of hybrid energy systems and their dependence on embedded control software increases the difficulty in predicting interactions among the various components and subsystems. A modelling environment that can model not only the components but also control algorithms (such as Matlab/Simulink, Homer etc.) is therefore advantageous. Effective diagnosis of faults in an installed system also presents a challenge, because of the interactions between the components and the control system. Modelling may play an important role in diagnosis of the operating components. For example, running an electrolyser model and comparing actual electrolyser operating variables with those obtained from the model may help to diagnose a fault in the real electrolyser. This thesis focuses on modelling the principal components of hybrid solar energy systems that include energy storage in the form of hydrogen: a large photovoltaic array subject to manufacturer’s variability and temperature inhomogeneity; two types of electrolyser as commonly found in hydrogen energy systems; a metal-hydride hydrogen storage tank; a fuel cell. Attention is given here to building physics-based component models with minimum empiricism and to critically analysing the state of the art in modelling such components. The models have been realised in Simulink, so that they are mutually compatible and can be linked into a whole of system model. All the models were validated against experimental data and performed at least as well as models found in the literature. The thesis is based on six papers, four already published and two submitted.