Abstract: To solve the RFID tag-arrival-rate prediction problem in dynamic environment,the tag-arrival-rate prediction algorithm,which can dynamically self adapt to the change of arrival rate,was proposed based on the gray model．The changing trend of arrival rate was judged by the rule knowledge denoting reverse mutation of the tag arrival rate,in order that the model length can dynamically self adapt to the arrival rate change,as a result,the contradiction between prediction tracking speed and accuracy was overcome,which improves the prediction accuracy．The tag-arrival-rate change was modeled by the Non-Homogeneous Poisson Processes with sine intensity function．Simulation experiments show that the algorithm can effectively predict the arrival rate based on a few data in a few random data environment,and the accuracy of prediction is above general gray prediction algorithms and exponential smoothing algorithm.

Key words: Arrival rate prediction,NHPP,Dynamic self-adaptive,Gray model,Modeling length