【摘要】 随着淮北市相山区岩溶水开采量不断增大,区内岩溶水水位降落漏斗范围不断增大,为保障岩溶水的安全开采与地质环境安全,进行本区岩溶水安全开采量计算十分必要。目前神经网络模型已被广泛应用于岩溶水水位动态计算,但由于网络全局寻优能力不理想,网络训练容易陷入局部极小值,导致网络泛化能力不理想。针对人工神经网络的不足,利用遗传算法(GA)对较为常用的BP神经网络权值、阈值进行优化,将此方法应用于相山区岩溶水水位动态的预测,并以该区岩溶水临界开采水位为控制条件,经模型计算得到相山区岩溶水多年平均安全开采量为3 001.7×10~4m~3。计算结果表明:与BP神经网络相比,GA-BP神经网络具有更高的预测精度,遗传算法可以有效提高BP网络的泛化能力。更多还原

【Abstract】 With the increasing exploitation of karst groundwater in the Xiangshan District of Huaibei in Anhui province,the cone of depression of the karst groundwater in this district increases quickly. In order to guarantee the safety of karst water exploitation and the environmental geological safety,it `s necessary to calculate the safe exploitation yield in this area. Artificial neural network has widely been used in prediction of karst groundwater levels. However,the global optimization ability of the network is not ideal,and the training is easy to converge to the local minimum points,which causes the generalization to be not ideal. Aiming at the disadvantages of the neural network,the BP neural network weights and threshold are optimized by using the Genetic Algorithm,and the karst groundwater levels in the Xiangshan district are forecasted with this method.Taking the critical mining level of karst groundwater as the control condition,the annual average safe exploitation yield volume of 3 001. 7 × 10~4m~3 is determined. The calculation results show that compared with the BP neural network,the GA-BP network has higher accuracy,and the genetic algorithm can effectively improve the generalization ability of the BP network.更多还原