基于灰色神经网络和证据理论预测地下水位的研究

Study on Ecological Agriculture Groundwater Level Forecast Based on Grey Theory BP Neural Network and D-S Theory

用灰色GM(1,1)模型、BP神经网络模型和证据理论相结合而成的灰色证据神经网络模型,对山东枣庄区域地下水位进行预测。基本思路是运用灰色GM(1,1)模型所得到的预测值按前两年来预测下一年的组合规律分别作为BP神经网络的输入输出;再用BP神经网络输出作为证据理论基本可信度分配函数,使用D-S理论将信息进行二次融合;并用地下水位实测数据对模型进行验证。结果表明,预测模型具有可行性和实用性,为生态农业规划发展提供了科学依据。

In this article the grey D-S BP neural network model that is composed of grey GM（1,1） model, BP neural network model and D-S theory was used to forecast ecological agriculture groundwater level of Zaozhuang Regional in Shandong Province. The method was to take the forecast value of GM（ 1, 1） model as the input and the output of BP neural network model respectively according to the compounding rule by using the former two-year forecasted data to forecast to the data of later year data, then to take the output of BP neural network as BPAF（Basic Probability Assignment Function） to fuse information twice by using D-S theory. And also the real data of groundwater level were taken to validate this model. The result indicated that this model had feasibility and practicability. This model provided scientific evidence for ecological agriculture planning and development.