Two dimensional cooperation prediction algorithm of communication network traffic in smart grid

With the evolution from traditional power grid to smart grid, the network traffic has become much larger in the integrated services network of the smart grid. It is necessary to predict these network traffic and make early analysis warning to ensure the performance of the integrated services network. The existing algorithms show large prediction error in the several consecutive points after the turning point, which is called one step lag problem. In this paper, a two dimensional cooperation prediction algorithm is proposed to solve the one step lag problem by combining the advantages of two dimensions. The proposed algorithm uses the Autoregressive Integrated Moving Average Model as the vertical dimension prediction algorithm to capture the daily periodicity of the integrated services network´s network traffic, and uses the wavelet neural network model as the horizontal dimension prediction algorithm to capture the unique trends of the forecast day. Simulation results show that the proposed algorithm solved the one step problem and significantly improved the prediction accuracy.