The vulnerability of water resources in Wenshan autonomous prefecture was evaluated. Analytic hierarchy process was firstly employed to determine the assessment indexes and grades suitable for water-abundant areas. In subsequence, comprehensive evaluation models for regional water vulnerability were established based on single and double hidden layer BP neural network respectively. Interpolation was used to construct network training samples, and the grading standard value was taken as the evaluation sample. Results showed that: 1) the assessment results of both models were generally the same, indicating that the models and evaluation methods were reasonable and feasible. Compared with single hidden layer network, double hidden layer network has better generalization ability and higher prediction accuracy, but longer training time; 2) The evaluated vulnerability of water resource in Wenshan in different target years was between grade Ⅲ and grade Ⅴ, which meant invulnerable to moderately vulnerable. The results objectively reflect the vulnerability of water resources  in the region and can serve as a reference for research and evaluation.