The fuzzy-optimized, multi-objective and multilevel comprehensive evaluation model for the comprehensive benefits of water resources in irrigation area was established on the premise of improving the weight vector decision method of different layers. The entropy weight evaluation of the first layer could be calculated as per the utility degree of information, which embodied variation extent of evaluation indices. The weight vectors of the intermediate layer could be determined by means of combining the entropy weights with linear subjective weights or mean value, which gave attention to both the preference of decision-makers and the objective data. Finally, the output of the topmost layer was used to evaluate the whole solution. A certain irrigation area in Shaanxi was taken as a verification example, and the results showed that all the benefit indices increased progressively. Scheme three of the year 2015 was the best one, in which the comprehensive benefits had increased 22.6% and 39.6% comparing with that in 1997 and 2005. The model not only to be used for program optimization, but also to evaluate the each scheme. The model is considered more reasonable and more feasible for comprehensive benefit evaluation in irrigation area than the other two methods in practice.