The ongoing advancement of power coupling mechanisms in the direction of high speed and high power density has brought to the fore the significance of addressing the uneven distribution of lubrication among components and the suboptimal lubrication effect of parts in the system. These factors have emerged as pivotal considerations affecting the lubrication and cooling efficacy of power coupling mechanisms. This paper presents a study of the flow distribution law and optimisation of oil circuit parameters in power coupling mechanism lubrication systems, with the objective of ensuring the accurate flow distribution of such systems. Firstly, a computational fluid dynamics (CFD) model of the lubricating oil path of the power coupling mechanism is established. The lubrication flow distribution law was investigated utilising the aforementioned model. Secondly, the design parameters with a minimal impact were eliminated through correlation analysis, thus reducing the subsequent computational burden. Finally, an elliptic basis neural network approximation model and a multi-island genetic algorithm (MIGA) were established to propose an optimisation design method for key components of an electromechanical composite power coupling mechanism, with the objective of achieving accurate flow distribution and improving the lubrication effect. After CFD model verification, the error of each outlet flow before optimization and the target flow is about 40%; After the optimization algorithm proposed in this paper, the average error of each outlet flow is about 2%.