Accurate forecast of mid-term power demand ensures the stable and efficient operation of power systems, and is essential for the construction of energy interconnections and renewable energy microgrids. However, the implementation of strategies aimed at reducing carbon emissions such as electric energy substitution increases the uncertainty of power demand. In order to effectively extract the changing characteristics of electricity demand, this paper firstly proposes a secondary decomposition model based on a seasonal-trend decomposition procedure based on Loess (STL) and variational mode decomposition (VMD) to reduce sequence complexity. Then, different models such as grey wolf optimized support vector regression (GWO-SVR) for different sequences were used to achieve the best prediction effect. In addition, this study used the Markov chain model to further improve the prediction accuracy based on interval optimization. To verify the effectiveness of the hybrid model, a case study was conducted on the monthly electricity consumption in Zhejiang Province, China. The results show that the proposed model effectively extracts the characteristics of changes in electricity demand and greatly improves the forecast accuracy. •A hybrid model based on secondary decomposition and interval optimization is proposed.•Setups for determining the number of components of VMD are proposed.•The intelligent reconstruction method can be used in the VMD model.•The Markov chain can be used to further improve prediction accuracy.