Geothermal energy extraction through closed systems is a secure approach responding to the global heating demand without contaminating subsurface water and causing seismic events. However, the generated power of conventional closed systems is much lower than those of open systems. Therefore, this study is dedicated to planning a novel closed system, which can produce a significant amount of thermal power. For this purpose, the performance of a single closed-loop deep system with a lengthy horizontal extension is preliminarily assessed. Based on the achieved results, it is feasible to produce roughly 3 MW thermal power while operating with thermosiphon flow. It is a big step forward in designing a new type of closed geothermal system that operates without pumping power and produces a considerable amount of thermal power, comparable to the power generation of open systems. Nevertheless, the low ratio of generated power to the total length of the wellbores and long payback period are big barriers to the spread of this system. Therefore, in the next step of this research project, enhancing the lateral heat exchange area by designing multilateral closed deep systems is proposed to increase this ratio. It is demonstrated that operation with multilateral systems can remarkably improve the performance of the system. Hence, working with multilateral systems is more reasonable than operating with several single systems to generate the same amount of power. However, it requires an extensive sensitivity analysis for different numbers of lateral wellbores and flow rates to identify the best operation scenarios. Additionally, some criteria are set as functions of extraction temperature, produced power, and relative drilling expenses to define successful cases. The interpretation of the results revealed that a successful project requires a specific relation between local vertical and horizontal flow rates. Finally, it is found that the long-term performance of a multilateral system can be predicted as a function of its short-term behaviour.