Abstract: This paper briefly analyses the conventional reactive power optimization compensation. The new method proposes a new optimization reactive power compensation for electrical network that uses neural network to predict electric network's important parameters and nonlinear prime dual interior algorithm to optimize reactive power. This intelligent control system diminishs power losses, and settles the problems that the electric power has complicated parameters and it is hard to constitute the compensation system model. The result shows that the effect of this intelligent control system is good and this algorithm is valid.