The present invention discloses a thermal power plant set identification data selection method based on historical operation data. The method comprises: obtaining the input and output data of a target loop from the historical database of the thermal power plant set; detecting whether the data obtained from the time domain is located in the same work condition or not; performing point-by-point computation of the power spectrum of the data obtained from the time domain, employing the numerical value quantization of the power spectrum to identify the suitable and unsuitable degrees of the thermal power plant set, selecting out the most suitable data segment in the obtained data, comparing the value of the quantization of the most suitable data segment with an experience value, and finally determining whether the most suitable data segment is suitable for the identification of the thermal power plant set; and performing identification of the selected most suitable data segment, and performing cross validation of the data at different data in the same work condition. The thermal power plant set identification data selection method based on the historical operation data displays the process that the historical operation data is employed to help modeling and combines the actual condition to improve the accuracy of the historical data selection for identification and improve the identification precision based on the historical data.