实数编码遗传算法优化的神经网络模型在岩溶水水位预报中的应用

Neural Network Model Based on RNCGA and Its Application to Karst Ground Water Evaluation

岩溶地下水资源的准确评价是一个重要而又难以解决的问题 .采用了遗传算法的最新成果 ,建立了以实数编码的遗传算法优化的前馈型神经网络模型以预测岩溶水动态水位 .模型继承了传统遗传算法的优点 ,兼具神经网络强大的函数逼近功能 ,同时又克服了传统神经网络优化方法易陷入局部最优解的缺陷 .实例的训练和预报结果表明 :实数编码遗传算法优化的神经网络预报模型精度较高 ,适合于岩溶地区的地下水资源评价 .

Karst ground water is widely dispersed in China. And the evaluation of karst water resources is an important but difficult problem, for the transport law and the runoff track of karst water are difficult to be ascertained clearly. In this paper, the authors combine the latest achievement of the genetic algorithms-the technology of real number coding genetic algorithms (RNCGA) with an artificial neural network model, and establish a neural network model based on RNCGA to predict the table of karst groundwater. The model takes the advantages of traditional neural network and genetic algorithms, and gets rid of their defects. In order to demonstrate the feasibility of the model, a case of Jinan area is selected for discussion. In the case, the predicted results are in good agreement with the measured data. This shows that the method presented in this paper is feasible in forecasting the karst water table. And it is of important reference value for the evaluation of groundwater resources in other karst zones.