Abstract: In order to understand the dynamic of groundwater level and master the earthquake precursor dynamic, we designed groundwater level prediction system based on BP neural network. According to the groundwater level of Deyang, Sichuan Province, SWY-II digital water level meter is used to collect the groundwater level data of Deyang. Based on the collected water level data in 2015, the BP neural network is used to predict the change of groundwater level, and the data collected for one year are trained and tested. The structure of BP neural network is designed with three input nodes and one output node. In order to further validate the proposal, the groundwater level from July 1 to October 26, 2017 is predicted. The experiment shows that the scheme can predict groundwater level effectively and provide reliable data for earthquake precursor work.