The energy consumption characteristics of coal-fired power units are main indicators of energy saving and consumption reduction of power generation units. The big data mining technology was applied to analyze the historical operating data of the units. Based on the attribute reduction method of the fuzzy rough set

the operating characteristic parameters which affect the energy consumption characteristics of the units were reduced

and the simplest attribute set was obtained. Then

the sensitivity analysis model of the unit energy consumption was established by using the support vector machine technology. The sensitivity coefficient of the operating parameters to the coal consumption of power supply under different load conditions was analyzed

to provide guidance for the actual operation optimization and energy saving of the unit. Taking a 600 MW unit coal-fired unit as an example

the energy consumption sensitivity was analyzed. It is found that the precision of the test sample and the training sample of the unit energy consumption sensitivity analysis model based on data mining is basically within ± 0.04%

and the accuracy is less than 0.5%. The model is feasible and effective. In addition

the sensitivity coefficient of operating characteristic parameters to the coal consumption of power supply change greatly but irregularly under different load conditions. The corresponding adjustment measures should be taken according to the change of the sensitivity coefficient in the actual optimization operation of the unit.