Abstract: Based on panel data of China's 20 industries from 2002 to 2012, this paper uses threshold regression method to study the relationship between industrial agglomeration and energy efficiency with the degree of industrial agglomeration and the level of opening up as threshold variables. The empirical results show that in terms of the whole industry sample and non-technology intensive industry sample, in the low or lower range of the degree of industrial agglomeration, industrial agglomeration is beneficial to improve energy efficiency; in the high range of agglomeration degree, industrial agglomeration has negative influence on energy efficiency. For technology-intensive industry, with the improvement of agglomeration degree, the positive effect of industrial agglomeration on energy efficiency becomes more and more dominant. When we use the level of industry opening as threshold variable, with the raising of the level of opening to the outside world, the positive role of industrial agglomeration on energy efficiency is gradually not distinctive. When the level of opening up is very high, the influences of industrial agglomeration on energy efficiency begin to appear negative. This requests us to guide and control the degree of industrial agglomeration, and to adapt to different levels of opening through independent innovation according to industry agglomeration, thus to make industrial agglomeration adequately play a positive role in improving energy efficiency.