Abstract: Storm surge disaster is one of the most serious marine disasters in China, and the pre-assessment of storm surge disaster has an important role in disaster prevention and mitigation.In this paper, 40 sets of storm surge disaster data from 2002 to 2014 were selected for experiment.This paper establishes the index system of storm surge disaster and simplify the indicators by gray correlation analysis method.The combination model with SVM and BP neural network forecasts the direct economic losses and the affected population number of storm surge respectively, and with single prediction methods were compared.It can be found that the combination forecasting model can reduce the error, provided effective basis for decision makers during disaster management.