The exponential reference trajectory algorithm used in traditional predictive function control results in the fact that the dynamic response can not be tuned to track a predefined curve.A new reference trajectory self-tuning scheme for predictive function control is presented in this paper.The damp parameter of reference trajectory is on-line optimal tuned by comparing current reference trajectory with a predefined trajectory，which guarantees that the dynamic response can reach the set point at an expected time and follow the expected response.The simulation shows that the proposed method can result in favorable dynamic performance and good robustness.