LAUSR

Abstract Random walk algorithms have unique advantages in reservoir modelling of distributary channel systems (DCS). However, the existing random walk algorithms often face several shortcomings, and typically do not perform adequately in complex scenarios. To overcome this problem, this paper introduces our recent development and application of magnetic random walk algorithm. The algorithm absorbs the concept of attraction and repulsion in the magnetic field and allows the walking seeds to wander randomly in a magnetic field composed of "channel well" observations and "non-channel well" observations, based on magnetic force. After post-processing of the walking trajectories, a DCS reservoir can be reconstructed. In a case study, the magnetic random walk algorithm efficiently completed the reservoir modelling, and parallels the manually drawn version completed by experienced geologists. This algorithm can effectively overcome several of the shortcomings of the existing methods and reduce manpower to a great extent in solving redundant work, which has a great application prospect.