The problem of reconfiguration control for a three?craft coulomb formation in the deep space is investigated in this paper. First of all, considering external disturbance and Debye effect, the accurate dynamic models of three?craft coulomb formation are established. According to the characteristics of the coulomb formation dynamic model and the effect of solar radiation pressure on control precision, a PID charge feedback controller based on BP neural network is designed. PID control has simple structure and good stability, while BP neural network has abilities of autonomic learning and nonlinear approximation. Connecting PID control with BP neural network, the optimal PID control parameters are outputted to change the charge of satellites in order to change coulomb forces between satellites, which eventually make the formation converge to equilibrium. The simulations show that the controller is accurate and simple which is better than traditional PID control, and exhibits good control robustness.