Abstract: Taking Weiyuan County in Southwest China as an example, the evaluation system and grading standard of water resources carrying capacity were constructed from five aspects, i.e. society, economy, ecology, water resources and shale gas development, and the genetic algorithm (GA) was used to optimize back propagation (BP) neural network. GA-BP neural network combined model was thus formed to evaluate the water resources carrying capacity status of the study area from 2014 to 2019. The results showed that the maximum relative error of the verification data calculated by GA-BP neural network was 6.5%, and the correlation coefficient between the expected output and the result was 0.995 98. With the increase in the scale of shale gas well groups, the water resources carrying capacity index of the study area had been decreased year by year. The water resources carrying capacity from 2014 to 2017 was in a bearable state, and from 2018 to 2019, it was in a weakly carrying state. The main impact indicators of the index were per capita water resources, shale gas well group scale and water consumption of 10 000 yuan of industrial added value.