In this article, the non-parametric data envelopment analysis method (DEA) is employed to evaluate energy efficiency (EE) of 34 coal-fired power units in China. Input-oriented CCR (Charnes, Cooper and Rhodes) model is used for EE analysis. Two efficiency indices, generalized EE and special EE are defined and analyzed. The generalized EE is calculated based on four input parameters: coal consumption, oil consumption, water consumption and auxiliary power consumption by power units. The special EE is only based on two input parameters: coal consumption and auxiliary power consumption. Relations between these two EE indices and non-comparable factors including quality of coal, load factor, capacity factor, parameters of main steam and cooling method are studied. Comparison between EE evaluation results of the two indices is conducted. Results show that these two kinds of EE are more sensitive to the load factor than the capacity factor. The influence of the cooling method on EE is larger than that of main steam parameter. The influence of non-comparable factors on the special EE is stronger than that on the generalized EE. (c) 2015 Elsevier Ltd. All rights reserved.