Research On Comprehensive Evaluation Model Of Flood Disaster Based On Intelligent Algorithms

With the growing awareness of flood’s formation, development and evolution, more and more person realizes human activities play an important role on flood formation. In order to achieve the harmonious development between men and water, changing the original "Flood relief to "Flood management" has been proposed. Flood management means that start with the analysis of the floods’mechanism, study the resulting conditions and the regularity of the evolvement of time and space, and explore the mutual restrict influence between human and society. So that we will be enable to scientifically prevent, mitigate, and avoid the disaster. Among them, flood assessment is an important part of flood management because it can supply scientific basis and analysis method for management and decision. It includes two aspects such as the flood risk assessment before flood and the flood disaster loss evaluation after flood. This paper majorly studies the flood disaster loss evaluation.Floods comprehensive evaluation is a complex nonlinear system which includes multiple uncertain factors and continuously evolves in space and time. A single method is difficult to adequately solve the various problems of flood assessment, so the urgent need of research work at present is to update the evaluation method and improve the evaluation system. This study is supported by the National "973" Key Basic Research Program of China (Project No.2007CB714107)-"Dam break mechanism and risk control theory under complicated conditions" and the Special Research Foundation for the Public Welfare Industry of the Ministry of Water Resources (Grant No.201001080)-"key technology research and application on Big East Lake water reticular scheduling". By analyzing the characteristics of the floods, it introduces the intelligent optimization methods to the fields of flood evaluation, with the representative of the cloud model and the support vector machine. According to the specific problems of the flood evaluation, the research improves and optimizes these intelligent algorithms and establishes appropriate flood intelligent assessment model, and the main research work and innovative results are as follows:(1) Based on the complex characteristics of flood evaluation, the research analysis the formation mechanism of the flood, Study the assessment effect based on gestation factors, formative factors, hazard-affected factors of flood calamity and propose that flood assessment is an nonlinear complex system which is collected by multi-attributes, fuzzy uncertainty and randomness. Furthermore, it explores the guiding significance of these characteristics on how to select intelligent methods and how to optimize the assessment models.(2) For some assessment factors is difficult even unable to accurately quantify in the course of flood disaster, the cloud model is used to create quantitative model for qualitative concepts. Then transfer the quantitative samples by qualitative reasoning rules, so that can get the comprehensive reasoning between qualitative concepts and quantitative expressions. During the reasoning process, it analyzes the fuzzy membership of cloud model, elaborates the lapse principle and validity index of maximum membership principle, and gives the handing methods if the maximum membership principle is failure. The cloud model which is combined with fuzzy membership has been applied into the flood disaster assessment of JingJiang area and achieves good results. The studies show the model can finish assessment work under incomplete or imprecise quantitative information, which reflects the fuzziness and randomness and shows strong adaptability and generalization ability in extensive assessment problems.(3) Flood assessment can be considered as a composite nonlinear property of the pattern recognition problem. In this study, support vector machine model is introduced to the flood assessment to solve the classification problem of pattern recognition. In order to reflect the significant difference of assessment factors and avoid biased problem by single weights, the study analyzes the weighting method of evaluation factor, combines objective weights and subjective weights into combination weights. And then propose a feature combinative weighting support vector machine (FWSVM), meanwhile, analysis the unclear parameters optimization of support vector machine, improve the grid search method, and comes up with a quadratic optimization method depending on multiple suboptimal solutions. Finally, FWSVM which is improved by feature weighting and parameter optimization is effectively applied in the flood disaster evaluation. It can solve the assessment without assessment criteria and shows good application prospect in multiple factors comprehensive evaluation problems.(4) In the flood comprehensive grade evaluation, the sample distribution is seriously uneven and the small sample set is easy to be misclassified because of less training. For this issue, the research develop the support vector machine by extending the kernel width, combine the feature weighting method the multi-width, put forward a support vector machine model with multiple-width based on feature weighting to break through the technical bottleneck of misclassification. The case study shows that the model has certain advantages in small sample classification. It is a new method in flood disaster assessment and enriches the evaluation method system.