In order to reduce the impact of imbalance and uncertainty of the sub-area development on spatial load forecasting accuracy, a multi-stage spatial load forecasting model is proposed by combining cluster analysis and Markov theory. First, the maximum load, electricity consumption and average load percentage of unit area are extracted as indicators representing the imbalance of the sub-area development, and a k-means algorithm is adopted to cluster the sub-area to determine the load density of each development stage. Secondly, the transition probability between different development stages is summarized to form the state transition matrix of a Markov Chain, and reveal the change rule of spatial load, so as to deal with the uncertainty of the sub-area development. Thirdly, the short-, medium- and long-term spatial load forecasting models are established using information of industry expansion, the classification saturation density and state transfer vector. Finally, an example proves that the model can effectively consider the uncertainty of economic development and the difference of power consumption level, and the forecasting results of each stage have high reliability. This work is supported by National Key Research and Development Program of China (No. 2017YFB0902800) and Scientific and Technological Project of Henan Province (No. 182102210258).