Abstract: The Particle-In-Cell (PIC) particle simulation method is one of the important numerical simulation methods for conducting plasma kinetic physics research. In this paper, by constructing an end-to-end neural network model from the phase space to the electrostatic field and embedding it into the one-dimensional electrostatic PIC particle simulation algorithm, the effective integration of the neural network algorithm and the traditional particle simulation method is achieved by replacing the traditional Poisson solver. The feasibility of the method is verified by carrying out simulations of the plasma two-stream instability physics problem, observing the phase space evolution and the growth trend of potential energy during the running of the algorithm, and its computational performance is tested and obtained. The results show that the method is able to achieve an accurate simulation with four times the dataset duration in the case of input training with short duration training set. At the same time, the method is able to break the constraint of explicit particle simulation time step to a certain extent, which in turn can effectively reduce the total number of iterative computations of the simulation problem.