针对现有高程异常拟合方法拟合精度低的问题，提出了多策略融合蚁狮算法（ALO）优化反向传播（BP）神经网络的高程异常拟合方法。首先通过引入双策略随机游走机制增强ALO算法全局搜索能力，并结合蚂蚁位置自适应更新策略和精英蚁狮的个体突围机制进一步提高ALO算法的参数寻优能力和效率；然后利用多策略融合ALO算法优化BP神经网络初始参数的随机获取策略，提高BP神经网络高程拟合精度与稳定性；最后以柴达木水资源配置工程为例，将该方法与BP、MVO-BP和ALO-BP等神经网络方法进行对比分析。结果表明，采用不同的拟合点时，相较于BP、MVO-BP和ALO-BP方法，该方法的外符合精度分别至少提高了17.02%、18.24%、30.45%。

Addressing the issue of low fitting accuracy in the existing elevation anomaly fitting methods, we proposed an elevation anomaly fitting method for back propagation(BP) neural network improved by multi-strategy fusion ant lion optimization(ALO) algorithm. Firstly, we introduced a dual-strategy random walk mechanism to enhance the global search capability of ALO algorithm, and integrated the individual breakthrough mechanism of elite ant lions and an adaptive position update strategy for ants to improve the parameter optimization capability and efficiency of ALO. Subsequently, we employed the multi-strategy fusion ALO algorithm to optimize the random initialization strategy for initial parameters of BP neural network, thereby improving the accuracy and stability of its elevation fitting. Finally, taking the Caidam water resource allocation project as a case study, we compared and analyzed this method alongside the BP, MVO-BP and ALO-BP neural network. The results demonstrate that when different fitting points are adopted, the method achieves at least 17.02%, 18.24%, and 30.45% improvements in external consistency accuracy compared to BP, MVO-BP and ALO-BP, respectively.