Abstract: Considering that the fault diagnosis of nuclear power plant (NPP) adopting the traditional threshold way can hardly realize early warning, the prediction model according to the variation trend of state parameter and making use of the good local characteristic of RBF neural network for predicting the trend of NPP operating conditions was introduced. The operating condition trends under the normal transition and the small-break loss-of-coolant accident were predicted. The results show that RBF neural network can predict the parameter’s change and the predicted value matches with the real value.