The optimal output speed of ESPCP system is caused by the coupling of multi-factors and it is a typical nonlinear optimization problem. A genetic neural network model for output speed, which takes the three factors of the content of oil viscosity, pressure difference of the pump two ends, magnitude of interference between stator and rotor as the input variables of this model, has been built with the help of Matlab (Matrix Laboratory) software. The model using genetic algorithm optimize the weights and thresholds of neural network, so the convergence and prediction accuracy of the network were improved effectively. After the simulation through samples study and by combining field data, it shows that the predicting results is basically in accordance with the observation data, and acquire satisfactory results. The research fruit provides a new approach and a route for optimal output speed optimization of ESPCP system when more factors are considered.