Proposing memory computing and the development of technology, is based on the actual situation needs. For industries, its data processing is facing all kinds of problems and difficulties, such as large amount of data processing, low efficiency of data processing and slow processing speed, load forecasting of electric power industry also meet obstacles, to large quantities of data real-time analysis make predictions become a big problem, in this paper, the calculation based on memory, combined with bp neural network prediction model, make the research problems on load forecasting, compared with traditional methods, experimental proof has qualitative improvement.