Bifacial photovoltaics (bifacial PV) offer higher energy yields as compared to monofacial PV. The development of appropriate models for simulating the energy yield of bifacial PV power plants is a major topic in both research and industry. In particular, the adequate calculation of the energy yield from ground-reflected irradiance (GRI) is challenging. The purpose of this work is to investigate the currently available energy yield models and suggest areas for improvement. A new model with the proposed enhancements is used to investigate the behaviour of bifacial PV power plants in more detail. The model calculates the absorbed irradiation originating from eight irradiance contributions for the front and rear of each cell string: DNI, DHI, GRI from DHI (GRI(DHI)) and GRI from DNI (GRI(DNI)). The model was tested using a defined case study power plant. The breakdown of absorbed irradiation (subscript "ab") into its contributions revealed that while in summer months GRI(DNI-ab-rear) is significantly larger than GRI(DHI-ab-rear), both are roughly the same in winter months. Furthermore, for the calculation of GRI the common simplification of infinitely long module rows was avoided by implementing an algorithm for the view factor calculation for a three-dimensional space. This procedure allowed for the assessment of impact of the ground size on the annual energy yield. In a sensitivity analysis, it has been shown that the extension of the relevant ground area resulted in an asymptotical increase of the energy yield. Additionally, the impact of ground shadows on the power plant's performance was quantified. The presence of ground shadows reduced the annual electricity generation by almost 4%, compared to a hypothetical scenario where no ground shadows existed. Finally, five different ground surfaces and the resulting bifacial gains were analysed. The results show that while dry asphalt (12% reflectivity) gave less than 6% of bifacial gain related to generated electricity (BG(el)), the use of a white membrane (70%) would result in 29% of BG(el).