【摘要】 综合客观评价河流水质的健康及污染状况，科学分析预测水环境质量的变化趋势，对于河流水污染精准防治具有重要意义。选取绵阳市境内主要河流为研究对象，基于2014-2022年绵阳市12个水质监测断面的4项污染物指标监测数据，结合模糊综合评价法及主成分分析法对其断面水质进行评价并判断污染程度，建立带外源输入的非线性自回归（NARX）神经网络模型，预测研究区域内2025年和2030年的水质变化趋势。结果表明：1）综合不同水期来看，在枯水期，芙蓉溪仙鱼桥与梓潼垢家渡断面的水质级别分别在2015年和2016年都未能达到水质目标，满足水质功能标准的断面占12个监测断面的92%，而在平水期与丰水期各断面水质均能达到各自的标准；2）凯江老南桥、梓潼垢家渡及天仙镇大佛寺断面水质受水期影响较大，平武水文站在所选断面中总体水质较好，而芙蓉溪仙鱼桥较差；3）溶解氧（DO）、高锰酸盐指数（CODMn）、五日生化需氧量（BOD5）及氨氮（NH3-N）的实测值与预测值之间高度相关，均方误差均较小，满足水质预测的精度要求；4）从水质变化趋势来看，除天仙镇大佛寺断面2028-2030年的水质级别由2023-2027年的Ⅰ级降为Ⅱ级以及芙蓉溪仙鱼桥断面2025年和2030年水质级别分别为Ⅱ级与Ⅲ级外，其余断面水质级别均为Ⅰ级，总体水质有所提升。该研究提出的基于模糊综合评价法-主成分分析法-NARX神经网络模型研究框架能够准确地对河流多断面各水期的水质进行客观评价，并在一定程度上弥补水文观测的“数据盲区”，适用于其他类似河流的水质预测，可为河流水质综合评价及预测预警提供科学依据与技术支持。更多还原

【Abstract】 In order to effectively build environmental water management and planning, it is inevitable to synthetically evaluate the pollution status of water quality and scientifically predict the varied trend of water environment quality in the future. In this study,based on the monitoring data of four pollutant indicators from 12 water quality monitoring sections of the major rivers in Mianyang City during 2014–2022, the water qualities of these sections were assessed by fuzzy comprehensive evaluation method, and the pollution degree was determined by principal component analysis method. Subsequently, the nonlinear autoregressive neural network model with external inputs(NARX) was established to predict the varied trends of water qualities in 2025 and 2030. The results showed that 1) the water qualities of the Xianyuqiao section of Furong stream in 2015 and the Goujiadu section in Zitong in 2016 did not reach their own water quality targets in the low water period. The ratio of sections satisfying the water quality function standard to the total 12 sections was 92%, and the water qualities in all sections can reach their own standards in the normal and high water periods. 2) The water qualities of three sections located in the south of the Mianyang City(Laonan Bridge section in Kaijiang,Goujiadu section in Zitong and Dafo Temple section in Tianxian Town) were affected greatly by the water periods. The best and worst overall water quality was presented in the Pingwu Hydrological Station section and Xianyuqiao section of Furong stream respectively. 3) The predicted values of dissolved oxygen, permanganate index, five-day biochemical oxygen demand, and ammonia nitrogen were highly correlated with their measured values, and the mean square errors were small, which met the accuracy requirements of water quality prediction. 4) In terms of the change trend of water quality, the water quality level of the Dafo Temple Section in Tianxian Town decreased during 2028–2030 compared with that during 2023–2027. The water quality of the Xianyu Bridge section of Furong stream in 2025 and 2030 was Grade Ⅱ and Grade Ⅲ, respectively. The water qualities of the other sections were Grade Ⅰ, and the overall water qualities were improved. In summary, the proposed research framework integrated with the fuzzy comprehensive evaluation method, principal component analysis method, and NARX neural network model can be used to accurately and objectively evaluate the water quality of rivers under various water periods, which can compensate for the shortcomings of mechanical model that needs more hydrological observed data. It is also suitable for predicting the water quality of other similar rivers and can provide a scientific basis and technical support for comprehensive evaluation and prediction, thus providing early warning of river water quality.更多还原