Abstract: Objectives Addressing the low efficiency and high safety risks associated with multi-vessel anchoring during offshore wind farm construction, an intelligent anchoring position management method is proposed based on an Optimal Brain Surgery-Multi-Layer Perceptron (OBS-MLP) fusion algorithm.Methods Initially, a three-level optimization framework of feature decoupling-dynamic pruning-preference decision was constructed, which integrates the advantages of decision tree and neural network., utilizing OBS pruning strategies to reduce model parameter redundancy, thus overcoming the dimension disaster issue faced by traditional algorithms in high-dimensional decision spaces. Subsequently, a multi-modal feature fusion mechanism was established for hierarchical extraction and collaborative optimization of environmental static features and dynamic interaction features. Finally, a preference decision-making model under mixed-integer programming constraints was designed to quantify the economic and safety weightings of different adjustment strategies.Results Taking an offshore wind farm in Guangdong as an example, the average accuracy of the algorithm is 95.3 %, which is 8.05 % higher than that of the single algorithm. The average response time is shortened to 0.9 seconds, which increased by 30.77 %. The potential economic loss is reduced by about 240,000 yuan per year, which effectively reduces the safety risk in the anchoring operation of ships.Conclusions The research provides an efficient solution for multi-ship anchoring collaborative optimization and promotes the digital transformation of maritime safety management.