A new quantum harmony search based improved discounted mean square forecast error (QHS-IDMSFE) combination model is proposed in order to combine the information of single forecasting result and reduce the risk of choosing model. Considering the influence of time difference and single model difference, the discounting factor (β) in DMSFE is extended to the matrix form. Quantum harmony is employed in Harmony Memory (HM) to increase the information of harmony based on quantum states superposition, which can effectively improve the performance of search efficiency. Thus, the best β value can be determined through optimizing Mean Absolute Percent Error (MAPE) objective function by QHS algorithm. So, the corresponding weight for each single model can be determined based on optimal β value. The QHS-IDMSFE combination forecasting method is established and tested for annual electricity consumption prediction for two areas. The empirical analysis confirms the validity of the presented method and the forecasting accuracy can be increased in a certain degree. The proposed method is suitable to mid-long term electricity consumption prediction; meanwhile, the optimal β value can be determined intelligently.