Multi-Scale Storm Flood Risk Prevention Intelligent Decision-Making Based On Disaster Prediction

Rainstorm and flood disasters pose a serious threat to people’s lives and property and social stability,so it is very important to take targeted decisions in a timely and effective manner.our country has established a set of emergency response systems based on the severity of disasters.However,the traditional method of disaster statistics still relies on the reporting of governments at all levels,which not only consumes manpower and material resources,but also has the hidden danger of delaying the best time for rescue.In this regard,this thesis uses deep learning and risk assessment technology to conduct a rapid and detailed analysis of the storm and flood disasters,and provides multi-level intelligent decision-making suggestions in combination with disaster-causing factors and emergency response systems in various regions.Due to the wide range of rainstorm and flood disasters and the huge amount of statistical data,disaster reporting is slow and inaccurate,making it difficult to make timely and effective emergency response decisions.Aiming at these problems,firstly,this thesis constructs a comprehensive rainstorm and flood disaster evaluation system;secondly,using historical disaster data,a real-time forecasting research of rainstorm and flood disaster based on the combination of deep learning and risk assessment is carried out;finally,the disaster prediction results are compared with the actual Some emergency response systems are combined to build a practical decision-making framework for risk prevention.The main work content and research results are as follows:(1)Construct a rainstorm and flood disaster evaluation system.A comprehensive and accurate evaluation of the severity of rainstorm and flood disasters is extremely important for formulating effective disaster prevention and mitigation measures.To this end,this thesis draws on relevant literature and national standards to design four disaster indices that reflect different aspects of losses: population index,housing index,agricultural index and economic index.And a comprehensive disaster index that quantifies the overall loss degree is obtained from these four indexes.(2)Provincial disaster prediction based on deep learning.Aiming at the problem of insufficient utilization of spatial information in existing methods for predicting storm and flood disasters.Based on raster data(such as precipitation,DEM,etc.)and VGGNet,this thesis designs a lightweight regression convolutional neural network model to predict the provincial disaster index,and achieved good results.At the same time,in view of the important impact of precipitation on the disaster situation,this thesis uses the Conv LSTM module to capture the temporal and spatial characteristics of precipitation data,and uses the method of feature fusion to construct the TSVGG-Light model,which further improves the prediction accuracy.In addition,this thesis also uses the interpretability algorithm to verify the reliability of the model.(3)Prediction of spatial distribution of disasters based on risk assessment.Due to the lack of city-level disaster data,the TSVGG-Light model cannot obtain the spatial distribution of city-level rainstorm and flood disasters.Therefore,based on the mechanism of storm and flood disasters,this thesis constructs a short-term storm and flood risk assessment model,uses the output results of the model to approximately reflect the spatial distribution characteristics of disasters,and verifies the effectiveness of the method through historical data.(4)Risk prevention decision-making based on disaster prediction.In order to make full use of the experience of flood control and disaster reduction in various regions,this thesis combines the disaster index predicted by the TSVGG-Light model,the risk assessment results and the existing emergency response plans for rainstorms and floods in various regions to construct a framework that integrates disaster prediction and risk prevention decision-making.