The accurate power prediction of actuator plays an important role in electric energy management. In order to improve the prediction accuracy in actuator system, an improved multivariable grey neural network model is established. Taken this aspect of the relative factors of producing a great impact on the needs of power in system into account, the PCA is used to extract the comprehensive variables as an input of the prediction model. By so doing, this effectively reduces the dimension on the basis of the accuracy of prediction. In the process of selecting the initial values and background values by utilizing the incremental method, the PSO is used to calculate the optimal background value to improve the prediction accuracy to a large degree. In addition, taken this aspect of the influence matter on sequence fluctuation into account, a BP neural network model is established to correct the error in the prediction value. The results show that the model is valid and advantageous to the energy dispatching of aircraft.