Urban inland inundation rapid forecasting method based on multi-output machine learning algorithm

Abstract

The invention discloses an urban inland inundation rapid forecasting method based on a multi-output machine learning algorithm. The method comprises the following steps: constructing an urban inland inundation model according to data of an existing research area; historical rainstorm rainfall time history distribution information is extracted, rainstorm with different characteristics is generated, and a rainstorm waterlogging database is constructed through simulation of the constructed urban waterlogging model; constructing a multi-output random forest model, and training and testing the multi-output random forest model by taking the rainfall factor as an independent variable and the submerging depth as a dependent variable and utilizing the rainstorm waterlogging database; based on rainfall forecast input conditions, rapid and real-time forecast of two-dimensional waterlogging inundation is realized through the constructed multi-output random forest model. The method is of great significance to early warning and prevention of urban waterlogging disasters, waterlogging prevention and disaster reduction and the like, can effectively improve the real-time waterlogging disaster situation forecasting efficiency and precision of areas where waterlogging disasters occur frequently, and can provide guidance to a certain degree for rapid forecasting of the waterlogging disasters.

Description

technical field [0001] The invention relates to the technical field of urban waterlogging disaster prevention, in particular to the technical field of waterlogging simulation, and in particular to a method for rapid prediction of urban waterlogging based on a multi-output machine learning algorithm. Background technique [0002] In the context of global climate change and rapid urbanization, rainstorms and waterlogging have become more frequent, resulting in increasing casualties and economic losses. Rainstorms and waterlogging have caused huge casualties and property losses to the country and society, seriously affecting social and economic development and people's happiness. For example, Guangzhou was hit by the "May 22" torrential rain in 2020, and the hourly rainfall at 42 stations exceeded the historical record. This led to serious waterlogging in many areas of the city, resulting in flooding of 443 sections. Among them, Metro Line 13 was out of service due to inversion...

AI Technical Summary

Problems solved by technology

At present, most of the previous machine learning models are based on the water depth prediction of a single or a small number of water accumulation points (Xu Weihong, Wang Shan, Gao Jianbiao, Zhang Nianqiang, Yu Qian, Li Na, Han Song, Wang Jing, Wang Yanyan, Ding Zhixiong. Based on machine It is difficult to predict the water depth on the urban waterlogging surface at the same time, and it is impossible to provide information on the submerged range; at present, some studies build a corresponding number of machine learning models for each grid point to realize For the prediction of all grids, however, this type of method does not consider the spatial correlation between each water depth point, and the accuracy is limited to a certain extent. In addition, this type of method has the disadvantages of complex algorithm model construction process, which consumes a lot of time and storage space

Images