A data-driven spatial load forecasting(SLF)method based on Softmax probabilistic classifier is proposed. The functional land plots in the grid system are used as SLF units of spatial load forecasting and the attribute is described through the multi-dimensional indicator system. Based on a large amount of research data

the law of spatial load density distribution and typical shape of load curve of different land plot types are extracted by data mining method. The Softmax probabilistic classifier is introduced to forecast load levels of unknown land plots. Bottom-up superposition for load forecasting results of adjacent land plots in time domain is conducted

which obtains forecasting information of larger area including load levels and forecasted load curve. The simulation results of the example show that the proposed spatial load forecasting method has a certain improvement in forecasting accuracy.