Variance reduction technique is one of the most efficient methods to accelerate Monte Carlo (MC) results convergence. Among all techniques, MC self-couple variance reduction method is highly praised for it does not need any secondary model or code. This method can be applied in high-fidelity geometry simulation. In this paper, a new mesh-based bias parameters generation method based on multi-group Monte Carlo adjoint calculation for global variance reduction has been proposed. Forward flux is used to calculate global adjoint source. Source bias parameters and transport bias parameters are constructed by using global adjoint flux. The mesh definition of mesh-based adjoint flux tally and mesh-based weight window could to be fully consistent with each other. H.B. Robinson-2 in-vessel and ex-vessel neutron dosimetry experiment has been used to validate the calculation efficiency. According to numerical results, the calculation efficiency has been improved about 1–2 orders for H.B. Robinson-2 benchmark and the goal of global variance reduction has been achieved. The mesh-based bias parameters generation method based on multi-group Monte Carlo adjoint calculation would be a strong tool for Monte Carlo global deep-penetration shielding calculation.

方差降低技术是加速蒙特卡洛（MC）结果收敛的最有效方法之一。在所有技术中，MC自耦合方差减少方法由于不需要任何辅助模型或代码而受到高度赞扬。该方法可应用于高保真几何仿真。提出了一种基于多网格蒙特卡洛伴随计算的基于网格的偏置参数生成新方法，用于减少全局方差。正向通量用于计算全局伴随源。源偏置参数和输运偏置参数是通过使用全局伴随通量构造的。基于网格的伴随通量计数的网格定义和基于网格的权重窗口可以彼此完全一致。乙肝 Robinson-2容器内和容器外中子剂量测定实验已用于验证计算效率。根据数值结果，对于HB Robinson-2基准，计算效率提高了大约1-2个数量级，并且达到了减少全局方差的目标。基于多组蒙特卡洛伴随计算的基于网格的偏置参数生成方法将是蒙特卡洛全深穿透屏蔽计算的有力工具。