Different from the load forecasting of lines above 10kV, the super short term forecasting of a single user tends to be highly inaccurate. The paper first analyzes the source of this inaccuracy, indicating that this inaccuracy is inevitable. In view of this problem, the load curve is no longer taken as the forecast target, but the idea of using heatmap to represent the load forecasting result is put forward in this paper. Then, this paper combines the heatmap with the basic principle of Markov process, and designs a single user super short term load forecasting algorithm. This algorithm includes the initialization, incremental update and generation of forecasting data. Finally, the experimental verification of user data in Gusu district of Suzhou city shows that the algorithm presented in this paper has high accuracy and practical value.