Limited to incompleteness of power load,electricity and other relevant data,the accuracy of network supply short-term load forecasting is hard to enhance.The data accumulation of distribution and utilization system and rapid development of big data technology provide data basis and technical support for big data based load forecasting.Firstly,the sources and types of'dirty data'are analyzed based on the characteristics of distribution and utilization big data,and the corresponding methods of data cleaning are put forward in this paper.Secondly,based on a large amount of historical power load,electricity consumption and meteorological data,industry load-temperature impact model,and industry electricity-holiday impact model are established.Finally,short-term load forecasting is carried out based on above models,and test cases show effectiveness and accuracy of proposed big data based short-term load forecasting method.