In the view of electricity consumption pattern fuzziness, effected by various factors such as socioeconomic structure and climate change, a new system can be presented for these uncertainties, called gray system. At last, GM (1,1) gray model can be used for fitting and to predict the long-term and medium-term power consumption forecasting. With the popularity of electricity consumption and the complexity of power system, the traditional gray model is not enough to satisfy the accuracy of the prediction. Besides, because of the certain regularity in monthly network power, Markov transfer probability can be used as the weight to carry out the weighted calculation. To predict the volatile data, in this paper, a gray GM (1,1) Markov prediction method is presented, which can effectively compensate for the limitation of the two methods in data fluctuation and improve the precision in short-time power data. In the end, after building the model, simulations show that the gray GM (1,1) Markov prediction model effectively fits the change tendency of the consumption and greatly improve the prediction accuracy.