The loss of data packets affects seriously the stability of control system in modern network. This article mainly for the problem of packet loss uses a randomly distributed variable satisfying Bernoulli to describe the data loss during transmission, and then establishes a random and non-linear input model. A PID controller based on BP neural network is designed. The controller combines BP neural network algorithm and PID controllers effectively, to achieve the best PID control setted by BP neural network. Simulation results show that compared with the performance of traditional PID controller, the PID controller based on BP neural network can make the system owe a good control effect and show excellent environmental adaptability and robustness, when packet loss rate of the network control system is within a certain range.