The remaining useful life (RUL) prediction of aero-engine is of great significance for flight safety. The existing methods do not fully consider the influence of the individual differences of engines on the degradation process. And these methods are difficult to extract key features in highly coupled data, resulting in low prediction accuracy. To solve the problems, an adaptive prediction method for individual differences of engines is proposed. Firstly, the position of the First Prediction Time (FPT) is determined by calculating the maximum volatility value of the Health Indicator (HI) curve, so as to obtain the accurate RUL label. On this basis, the improved Temporal Convolutional Network (TCN) model is used to capture both local degradation patterns and global features. Then enter the bidirectional cyclic network to model the temporal features in the forward and backward directions. Finally, the accurate RUL value is obtained through the deep regression network. Compared with the existing methods, the proposed method has advantages in both prediction accuracy and robustness.