As an essential part for understanding and analyzing the status and problems of the water environment, water environmental quality assessment is an important basis for river planning, governance, and management. In the view of fuzziness, randomness, and non-rigidity of the maximum membership degree in water environment assessment, a normal cloud-fuzzy variable evaluation coupling model was proposed to quantify the fuzzy boundary problem, identify the membership function and water quality level identification, and adopt the minimum Relative Entropy Coupling Weights. This paper presents a quantitative treatment of the fuzzy boundary problem and identifies membership functions and water quality level identification by the normal cloud-fuzzy variable coupling model. In addition, we used the minimum relative entropy to couple the Shannon entropy and AHP weights. This model was used to evaluate the water environment in the Qinhuai River in 2016. The results were compared with the single-factor index evaluation method, fuzzy comprehensive evaluation method, and the cloud model. It was found that the evaluation results of water environment were more reasonable, effective and feasible. The analysis of model-level eigenvalue (H) shows that most of the water environment in Qinhuai River upstream and Qinhuai New River flood season is better than the non-flood season; however, it shows the pollution characteristics of urban industry and living; the water environment in Lower Qinhuai River flood season is worse than the non-flood season and it shows the typical pollution characteristics of urban living and catering. Research methods and results presented here will have certain reference and guidance for the future evaluation and research of regional and catchment water environment.