Real-time adaptive control of distributary channel reservoirs using Monte Carlo Tree Search planning under geological uncertainty

Reservoirs in distributary channel systems (DCS) serve as critical carriers for hydrocarbon accumulation, yet their strong heterogeneity poses persistent challenges for reservoir characterization. Although the Monte Carlo Tree Search (MCTS) algorithm has demonstrated potential in reservoir modeling, conventional implementations often overlook complex geological constraints, resulting in fluid distributions that violate physical laws, such as the “oil-above-water” principle. To overcome these limitations, this study proposes an innovative MCTS-based modeling framework that systematically integrates fault constraints and oil–water relationships into the stochastic modeling process. The primary novelty of this research lies in the transformation of geological rules into intrinsic algorithmic search constraints. Unlike traditional methods that treat oil–water contacts as posterior filters, our approach directly embeds a conflict avoidance mechanism within the MCTS expansion and simulation phases. This enables the algorithm to dynamically suppress potential inconsistencies during the decision-making process. The superiority of the proposed method is validated through comparisons with mainstream algorithms. Experimental results indicate that the optimized algorithm effectively eliminates oil–water contradictions and reduces modeling time. By systematically fusing multidimensional constraints, this approach significantly enhances inter-well prediction accuracy and geological plausibility, providing robust technical support for the refined development of complex oil and gas fields.