Abstract: Load forecasting is critical for management and security of smart grid system. Traditional methods are usually on the basis of historical power consumption data, and the popularization of multi-meter integration technology makes analysis of integrated energy consumption data more efficient. Towards the issue of load forecasting, with water/power/gas consumption data collected by integrated smart meter as features, two mid-and-long term power consumption forecasting methods are proposed: gaussian process regression (GPR) and relevance vector regression (RVR). Experimental results show the superiority of the proposed method and the significance of integrated energy consumption data for load forecasting problem.