The result of dividing external operating conditions by using the historical operating data of power plants depends on the adaptability of the mining algorithm to the data. In this paper

the k-means algorithm for the external operation of the historical operation data of the power stations was proposed

and the initial clustering numbers of the k-means algorithm and the calculation method of the clustering center were analyzed and improved. Moreover

this algorithm was applied in data mining of the external conditions of the unit running data for the power station

and the historical data of the external temperature of the power station was clustering analyzed by the equal width method. The mining results show that the improved k-means algorithm and the equal width method are more reasonable

and the optimal combination of external operating conditions can be obtained to describe the unit operation

which can provide more reasonable data references for the field operation personnel.