针对现有高程异常拟合在复杂地形时拟合精度不高的问题，提出多策略改进灰狼优化算法的BP神经网络拟合方法。首先，利用精英反向优化策略初始化灰狼种群，提高初始种群的多样化；通过动态权重因子，实时调整最优狼引导能力的影响；利用分段搜索策略，增强灰狼算法全局搜索与局部搜索能力的平衡。然后，通过多策略改进灰狼算法优化BP神经网络收敛因子提高高程拟合精度。结合工程实例数据构建IGWO-BP高程拟合模型，结果表明IGWO-BP高程拟合模型的收敛效率和高程拟合精度优于BP模型和GWO-BP模型，具有较高的拟合精度。本文可为提高高程异常拟合精度提供一定的参考。

In response to the issue of low fitting accuracy in existing elevation anomaly fitting when confronted with complex terrain， a BP neural network fitting method based on a multi-strategy improved Grey Wolf optimization algorithm is proposed. First， an elite reverse optimization strategy is utilized to initialize the Grey Wolf population， enhancing the diversity of the initial population. A dynamic weight factor is introduced to adjust the influence of the optimal wolf's guiding capability in real-time. A piecewise search strategy is employed to strengthen the balance between global search and local search capabilities of the Grey Wolf algorithm. Subsequently， the multi-strategy improved Grey Wolf algorithm optimizes the convergence factor of the BP neural network to improve elevation fitting accuracy. By combining engineering instance data， an IGWO-BP elevation fitting model is constructed， and results indicate that the IGWO-BP elevation fitting model shows superior convergence efficiency and fitting accuracy compared to the BP model and GWO-BP model， demonstrating high fitting precision. This study provides a reference for improving elevation anomaly fitting accuracy.