China's total energy consumption, according to the official data, decreased impressively during 19971998 and increased sharply during 2003-2007, which in turn resulted in energy intensity fluctuation. Many literatures explained this "unusual phenomenon" from the perspectives of technical change, economic structure shifting and statistical data quality. They measured aggregate energy in thermal units by using linear summation approaches. In this paper, from the perspectives of heterogeneity and imperfect substitutability among diverse energy types, we further examine China's aggregate energy consumption by using Divisia (Sato-Vartia) approach. The results show that China's aggregate energy consumption and intensity fluctuated slightly less than values calculated by using conventional linear approaches, and the "unusual phenomenon" is partly explained. It also implies that China's energy intensity changes in 2006-2007 are slightly more optimistic than those officially reported, and the official communique of provincial energy intensity reduction achievements are partly bias. Some provincial achievement are underestimated or overestimated on some provinces. Our empirical results are also helpful to further research, such as energy-economic modeling, energy price elasticity, and elasticity of substitution among capital-labor-energy-material (KLEM). The difficulties or defects when using Divisia approach are also discussed in this paper. (C) 2009 Elsevier Ltd. All rights reserved.