Abstract: The concept of operating mode is proposed by analyzing the influence of the thermal power plant operating mode on economy. A kind of model mergering algorithm based on clustering analysis is designed with full use of power plant operation data, which is used to establish a typical set of operating modes of thermal power units. In this algorithm, data of dividing range of operating condition are set to mine objects through comparison and analysis of the similarity among these typical thermal parameters. The operating records meeting requirements are classfied into one kind of operation mode. After the collection of mining operation mode is completed, a typical operating mode library is established. The result of data mining to Panshan Power Plant historical operation by this merging algorithm proves its effectiveness.