Abstract:The purpose of this paper is to use large data technology to mine the power data systematically, provide accurate load forecasting, and provide the basis for the safety monitoring, preventive control and emergency treatment of power supply. Based on the problems existing in the operation of distribution network in Xi''an area, this paper comprehensively analyzes the external environment of the distribution transformer overload, the distribution transformer information, the distribution transformer property information, the type of distribution transformer power supply customer and so on, and applies the big data machine learning algorithm, the big data preprocessing technology, the data mining modeling technology and the big data visualization technology. On the basis of analyzing the correlation and importance of the influence factors on the heavy overload of equipment, using the classification prediction mining method and the random forest algorithm, analyzing and calculating the influence variables and the target variables, establishing the relational model mining model, The heavy overload analysis and early warning model of power distribution transformer are constructed to complete model verification and correction, to realize accurate early warning of heavy overload in the next week, evaluation of safety coefficient of distribution transformer, factors of heavy overload distribution and variation, and feature analysis and display, and so on, which will provide a strong support for the operation and maintenance of the power grid. Through heavy data overload and early warning analysis based on big data, the efficiency of operation and maintenance will be improved, and the safe and reliable operation of power grid can be realized.