【摘要】 文章以济南市趵突泉泉域为研究区，采用核极限学习机(kernel extreme learning machine, KELM)建立泉域地下水流数值模型的替代模型，使用拉丁超立方抽样(Latin hypercube sampling, LHS)方法确定60组地下水开采方案用于训练KELM模型，通过对比地下水流数值模型的模拟结果与替代模型输出的结果，评价所建立替代模型的性能。结果表明：替代模型输出的地下水位值与地下水流数值模型模拟得到的地下水位值基本接近，且模型的运行时间减少了约99.62%。说明该模型可作为趵突泉泉域地下水流数值模型的替代模型，可提高区域地下水优化管理模型的求解效率。更多还原

【Abstract】 In this paper, the Baotu Spring area in Jinan City is used as the study region to establish a regional groundwater flow numerical simulation model. Based on this, the kernel extreme learning machine(KELM) is used to develop an alternative model to the numerical model of groundwater flow in the spring area. The Latin hypercube sampling(LHS) method was used to obtain 60 sets of groundwater extraction scenarios for training the KELM model. The performance of the developed alternative model was evaluated by comparing the simulation results of the numerical model with the output of the alternative model. The results show that the groundwater levels output by the alternative model are basically close to those obtained from the simulation of the numerical model, and the running time of the alternative model is reduced by about 99.62%, indicating that the KELM model can be used as an alternative model to the numerical model of the groundwater flow in the Baotu Spring area to improve the efficiency of solving the regional groundwater optimal management model.更多还原