Abstract: It is hard to measure the water content of crude oil precisely because the dielectric constant of crude oil varies greatly when displacing agent and emulsion breaker are injected. Based on field data collected from gauges and manual analysis values, using BP neural network to predict crude oil's water content and through the method of on-line hardware adjusting combined with software real-time predicting, measuring precision of crude oil's water content can be improved effectively, which is of remarkable significance in ensuring oil production and cutting down production cost.