This paper takes the layered water injection technology in oil production engineering as research object. This technology is characterized by strong nonlinearity, strong interlayer interference, time variation and complex model. Focusing on these characteristics, this paper proposes a novel control strategy based on PID-NN algorithm to satisfy the requirements of layered water injection technology. Many performance indexes of the controller are optimized, including model establishment of nonlinear controlled object, multi-parameter decoupling, controller acceleration and precision improvement. The simulation and experiment show that compared with traditional controller based on PID and controller based on BP neural network, the optimized controller has stronger robustness, faster response and better flow control, so it has an advantage over traditional controllers of dealing with discrete events and multi-variable decoupling problems.