Evaluation of rural groundwater quality is an important basis to ensuring the safety of drinking water in rural areas. On account of the importance of weight value to the TOPSIS (Technique for Order Preference by Similarity to an Ideal Solution) model, combinatorial weighting and optimization on the subjective weight and the objectiveweight respectively determined by the fuzzy analytic hierarchy process (FAHP) and the entropy weight method are carried out based on the game theory set model. In the meantime, a “virtual negative ideal point” is defined to replace the traditional negative ideal point, hence avoiding the equivalent Euclidean distance from the sample point respectively to the ideal point and to the negative ideal point. With the water quality monitoring data of five typical villages and towns in Deyang city as evaluation object, the water quality grade is determined by calculating the relative closeness degree. The results of the present model are compared with the evaluation results of F method, the Entropy-TOPSIS and the FAHP-TOPSIS model, which proves that the results of the improved TOPSIS evaluation model is objective, effective and practical. Besides, the present model could also accurately reflect the groundwater quality grade and the deviation from grading standard.