In view of the problem that abnormal cellular load will bring about adverse effects to the spatial load forecasting，a probability spectrum method for ascertaining the maximal value of cellular load is proposed.By proceeding from the planning demand and combining with the actual operation mode of the urban power network，the method classifies the abnormal data in cellular load and clarifies their different modes of impact on spatial load forecasting.It also reveals the internal connection between the data needed and load transfer，non-stationary growth and abnormal conditions of cellular load.By computing the probability spectrum of cellular load，the characteristics of different kinds of abnormal data are described and depicted.Then the maximal value of cellular load is limited and restrained on the cellular probability spectrum curve treated by Gauss fitting technique.Finally，a case study shows that the prediction accuracy is appreciably improved by the maximal load using the method.