基于改进重叠投影法的空间耦合水污染评估模型

Space coupling water pollution evaluation model based on improved overlapping projection method

为准确评估库区上游河流突发水污染的危害程度,针对库区上游区域河床中上游窄、下游宽的结构特点,在传统重叠投影法的基础上,于重叠区域上边界引入流速、水位、污染物浓度等物理量的二次横向分布函数,基于神经网络的思想,采用最速下降法构建训练算法,对各分布函数中的未知参数进行训练,将收敛后的分布函数作为模型的连接条件,进而建立改进的一维、二维水动力和水质空间耦合模型;以青龙河流域为例,将训练得到的分布函数的必要参数作为一维、二维连接,模拟了河流环境中污染物的时空变化规律。结果表明,改进的耦合模型计算结果稳定,显著降低了物理量由一维向二维过渡时的突变性,相比传统重叠投影的单数值传递,改进的方法以训练后的分布函数作为一维、二维耦合的连接,可以大幅度提高耦合的精确性。

To accurately assess the damage degree of sudden water pollution in the upstream of the reservoir area,according to the structural characteristics of narrow upstream and wide downstream in the upstream riverbed of the reservoir area,one-dimensional and two-dimensional hydrodynamic and water quality space coupling models were established and improved.Based on the traditional overlapping projection method,a quadratic transverse distribution function of the flow velocity,water level,pollutant concentration and other physical quantities were introduced at the boundary of the overlapping area.Based on the idea of neural network,the steepest descent method was adopted and a training algorithm was constructed to train the unknown parameters in each distribution function.Finally,the converged distribution function was used as the connection condition of the model.In the case of Qinglong River Basin in Taolinkou,the necessary parameters of the distribution function were obtained through training,and it was taken as the connection of one-dimensional and two-dimensional to simulate the temporal and spatial variation law of pollutants in the river environment.The results show that the calculation results of the improved coupling model are stable,which significantly reduces the mutation of physical quantity in the transition from one to two dimensions.Compared with the single numerical transfer of traditional overlapping projection,the improved method takes the trained distribution function as the connection of 1D-2D coupling,which can greatly improve the accuracy of coupling.