【摘要】 针对水质监测数据多源、非线性、不确定性大的特点，提出了一种基于神经网络与改进D-S证据理论相结合的水质评价模型。该模型首先利用3种前馈神经网络对水质监测数据进行初步评价，将初步评价结果归一化后作为基本概率分配，然后引入证据权重修正冲突证据，根据D-S合成规则得到融合评价结果，最后利用迭代思想修正评价结果。其中，基于证据权重修正冲突证据并进行融合结果的迭代修正，能有效解决传统D-S证据理论无法处理高冲突证据的缺陷。冀南地区5个监测断面水质评价结果表明，该水质评价模型能够提高水质评价准确性。更多还原

【Abstract】 Considering the characteristics of multi-source, nonlinear and high uncertainty of water quality monitoring data, a comprehensive evaluation model of water quality was proposed which combined multiple neural network and improved D-S evidence theory. Three feed-forward neural network models were used for preliminary assessment, the results were normalized as the basic probability assignment, then the conflicting evidence was corrected by the weight of evidence, and the fusion result was obtained by D-S rule. Finally, the iterative method was used to modify the evaluation result. It could effectively solve the defect that traditional D-S evidence theory could not deal with high-conflict evidence by modifying conflicting evidence based on evidence weight and iteratively modifying fused results. Through the water quality evaluation of the monitoring sections in southern Hebei Province, the results show that the proposed model can effectively reduce the uncertainty in the evaluation process and improve the accuracy of water quality evaluation.更多还原