Engine working baseline plays an important role in aero-engine real-time monitoring and health management technology. In recent years, the method of baseline mining using QAR (Fast Storage Recorder) data has been widely concerned. In this paper, a baseline modeling method of lightGBM aero-engine considering the influence of bleed air and oil temperature is proposed. The data of GE90-115B engine in operation are extracted according to the steady-state working conditions of the engine, and the data are similarly corrected as basic data, and then the baseline model is obtained by training with the grid optimization lightGBM algorithm with cross-validation. Compared with the baseline model excavated in the past, the baseline model trained in consideration of bleed air and oil temperature has higher precision and stronger generalization ability.