Spatial load forecasting is one of key problems in the process of electric system planning. But due to the special complexity of the power load, spatial load forecasting is still unsolved and need to be studied further. Here spatial load forecasting is presented by data fusion technology. In detailed, the samples are firstly classified by grey clustering, which results in the better-performanced samples labeled with the grey leagues. After that, as limited samples as be concerned, various LS-SVMs are trained by the corresponding classified samples. In this way, complicated nonlinear regression modeling of spatial load and its influenced factors is accomplished by LS-SVM with more efficient training. Finally, real data are employed to test the proposed approach.