Abstract: The state-based maintenance is by far the most advanced maintenance mode with the development of condition monitoring technology. Since the 1990s, the state-based maintenance of power generation equipment has been implemented in China, but with very few remarkably successful cases. In order to practically implement the state-based maintenance of power generation equipment, a power generation group has introduced, developed and popularized an early fault warning system based on similarity modeling technology for the power generation equipment since 2008. This paper introduces the mechanism of the system, and illustrates how the early fault warning system was effectively applied in the state-based maintenance of coal-fired power generation equipment with the case studies in steam turbine, boiler, electrical elements and chemical elements, respectively. At last, some suggestions are put forward on how to integrate them in a better way.