China is the world's largest CO₂ emitting country, and coal-fired thermal power generation accounted for over 50% of the total electricity generation in China in 2015. This study reports the changes in the power generation efficiency of coal-fired thermal power plants in China from 2009 to 2011 and elucidates how the differences in the production scale of the power plants and regional heterogeneity affect the power generation efficiency. We propose a metafrontier data envelopment analysis (DEA) decomposition framework to investigate the sources of inefficiency in power generation. The results suggest that, on average, the power generation efficiency of large-scale power plants is 13% higher than that of small-scale power plants. Although operational inefficiency is the main source of inefficiency in eastern and central China, the technology gap - the differences in the quality of coal consumed for electricity production and in the equipment of the power plants among regions - is the main source of inefficiency in western China. This study uses the results of the framework to discuss the scrapping policies for coal-fired thermal power plants in China. For large-scale power plants in western China, the components of inefficiency vary and, thus, policymakers should consider scrapping thermal power plants based not only on the level of inefficiency but also on their components.