The increasing penetration of intermittent renewables and the accelerated climate change are challenging the power system operation in China, and understanding the cost of reducing power outage durations is essential in supporting the equipment maintenance, infrastructure investments and regulation policies. Therefore, this study first uses production theory combined with a parametric distance function approach to estimate the marginal costs (MCs) of reducing power outage durations by 1 h. Then, we establish a fixed-effects panel data model to investigate the impacts of different environmental factors on the estimated MCs. Finally, the estimated MCs are applied to the evaluations and designs of interruption compensation prices in the demand response mechanism. The significant findings are that: (1) The national MC shows an increasing trend during the period from 2002 to 2017 in China, ranging from 1.27 billion yuan/hour to 11.63 billion yuan/hour. (2) The MCs vary substantially among different provinces, and provinces with better reliability levels will have higher MCs. (3) The current compensations for power outages are only about 6% to 61% of the estimated MCs, indicating that grid companies would like to pay for the compensations rather than to enhance the system reliability from the supply side.