By analyzing and comparing the common models and methods of state forecasting, a novel neural network technique, artificial immune network (AIN) in state forecasting of dyna-mical system is proposed to deal with the prediction problem. This paper is mainly focused on the AIN immune adjustment and immune planning, the network system structure designing, and the final model building. In order to examine the feasibility of AIN in state forecasting, the practical vibration data measured from some turbo-generator set are used to validate the performance of the AIN model by comparing it with a traditional BP neural network and RBF network model. The experiment results show that the proposed AIN model outperforms the BP neural network and RBF neural network based on the criteria of normalized mean square error, and it can capture the system dynamic behavior quickly, and track system responses accurately.