The invention relates to the technical field of electric power systems, and in particular to an electric power big data-based main transformer peak load prediction method and a data warehouse system.The method comprises the following steps of: S1, carrying out statistics and analysis on history data of a power grid system, so as to establish the data warehouse system; S2, analyzing an extracted knowledge according to a decision purpose of a final user by adoption of a sequence mode analysis and clustering analysis data mining method, distinguishing most valuable information and submitting theinformation to the user; and S3, preprocessing main transformer peak load data through the arranged history data, and establishing a support vector machine-based load prediction model to predict thefuture main transformer load condition. According to the electric power big data-based main transformer peak load prediction method, the data warehouse system is established, the data mining method isestablished, and the support vector machine-based method is adopted to predict the main transformer load change condition, so that correct and reliable main transformer peak load prediction is realized, and systems are reasonably scheduled to ensure safe and economic operation.