DFNN based information fusion technology for Early Warning of Marine Organism Detection in Nuclear Power Plant

Abstract

In order to improve the accuracy and efficiency of early warning of marine organism invasion in nuclear power plant cold source system, a multi-information fusion model based on dynamic fuzzy neural network is proposed to deal with the problems of single-index evaluation, arbitrary decision-making and incorrect combination of information in the traditional analysis and processing. Taking the intake of Daya Bay Nuclear Power Plant as the research object, the early warning information sources were screened from five aspects: marine organism density, ocean current, sea breeze, sea water temperature and sea water salinity. Aiming at the non-linearity and uncertainty of the early warning system, a dynamic fuzzy neural network model based on Gauss membership function, T-norm product operator, error reduction rate pruning strategy and linear least squares method is designed. And then early warning system based on multi-source information fusion and dynamic fuzzy neural network method is obtained. Based on this model, the possibility of water intake blockage in nuclear power plant can be predicted. The experimental data verify that the learning error and generalization error of the proposed prediction algorithm are controlled within [-0.1, 0.1], and the Euclidean distance is 10-5 orders of magnitude.