Abstract: In order to guarantee the demand of power grid dispatching command and office telephone,and reasonably allocate communication resources,it is necessary to accurately predict the traffic dynamic telephone traffic of electric power company.The integer generalized auto-regressive conditional hetero-scedasticity（INGARCH）model was introduced as the traffic prediction model of electric power company.Firstly,the definition of INGARCH model and the mathematical principle of N-step prediction method were introduced,and the parameters of INGARCH process were obtained by using conditional maximum likelihood estimation method.Then,the Poisson parameters of the future process were predicted through four steps of "pretreatment,training,prediction and update",so as to establish the prediction model. Finally,two indexes of the model were given to measure the performance. The real telephone traffic data of State Grid Shandong Electric Power Company was used for the experiment. By comparing the training set and test set of the three common models,the optimal training ratio was obtained,and the prediction performance of INGARCH model was compared with the three commonly used models.The experimental results show that the prediction effect of INGARCH model is very good,and the N-step prediction effect of INGARCH model is better than the other three models.Therefore,INGARCH model can be used as the telephone traffic prediction model for electric power company.