If using cellular load annual maximum of measured data directly in spatial load forecasting for urban power grids, errors caused by measurement, random fluctuations and communication are introduced to the predicted results, which led to reduce prediction accuracy. A method for ascertaining maximal value of cellular load in spatial load forecasting based on PCA is proposed. By analyzing the cellular load historical data and using PCA, the cellular load is decomposed into the principal components which can characterize cellular load general information and non-principal components which can characterize cellular load random fluctuations. By excluding non-principal components to inhibit the adverse effects caused by random fluctuations, and extract the principal components of cellular load as the reasonable maximum value of cellular load, and it is used for spatial load forecasting. Case study shows that the method is effective.