CN-122020574-A - Urban rainfall flood risk early warning method and system based on multi-source data

CN122020574ACN 122020574 ACN122020574 ACN 122020574ACN-122020574-A

Abstract

The application relates to the technical field of data processing, in particular to a city rainfall flood risk early warning method and system based on multi-source data, wherein the method comprises the steps of constructing a risk index of any historical multi-source rainfall flood data; the method comprises the steps of dividing historical multi-source rain and flood data into a training set and a testing set, constructing a random forest by using the training set, obtaining the predicted risk level of each decision tree of the testing set in the random forest, constructing a decision matrix of each decision tree based on the risk index and the predicted risk level of the historical multi-source rain and flood data in the testing set, further obtaining the voting weight of each decision tree, voting the predicted risk level of each decision tree according to the voting weight, obtaining a real-time multi-source rain and flood data prediction result, and sending out early warning according to the prediction result. According to the technical scheme, the accuracy of the rainfall flood risk early warning can be improved.

Inventors

YU YING
Fang Kailun
XIA YUAN
ZHANG ZHITONG
XU LIJUN
HE GUANGLIANG
CHEN XU
LIU JIE
ZHU JINGLU
LIU SIWEI
KANG LE
Zhao Yuanyue

Assignees

广州市城市规划设计有限公司

Dates

Publication Date: 20260512
Application Date: 20260414

Claims (10)

1. The urban rainfall flood risk early warning method based on the multi-source data is characterized by comprising the steps of constructing a risk index of any historical multi-source rainfall flood data, wherein the risk index of the historical multi-source rainfall flood data is obtained by multiplying a difference value between a numerical value and a maximum value of any dimension data in the historical multi-source rainfall flood data with a change trend of the dimension data in a monitoring window, and taking the sum of products of the dimension data as the risk index of the historical multi-source rainfall flood data; the method comprises the steps of dividing historical multi-source rain and flood data into a training set and a testing set, constructing a random forest by using the training set, acquiring the predicted risk level of each decision tree of the testing set in the random forest, and constructing a decision matrix of each decision tree based on the risk index and the predicted risk level of the historical multi-source rain and flood data in the testing set; Normalizing the decision indexes to obtain voting weights of the decision trees, inputting real-time multi-source rain and flood data into a random forest, voting the prediction risk level of the decision trees according to the voting weights to obtain prediction results, and sending early warning according to the prediction results.
2. The urban rainfall flood risk early warning method based on multi-source data according to claim 1 is characterized in that the method for acquiring the variation trend of the dimension data in the monitoring window comprises the step of taking a sequence formed by any dimension data in the monitoring window as input of a Thai-Send estimation trend analysis algorithm to acquire the variation trend of the dimension data.
3. The urban rainfall flood risk early warning method based on multi-source data according to claim 1 is characterized in that the construction method of the decision matrix comprises the steps of taking a real risk level as a row, taking a predicted risk level as a column, counting historical multi-source rainfall flood data of each position in the decision matrix in a test set, and completing the construction of the decision matrix by taking an average risk index of the historical multi-source rainfall flood data of any position as an element value of the position.
4. The urban rainfall flood risk early warning method based on the multi-source data according to claim 3 is characterized in that the construction method of the decision index comprises the step of accumulating products of values of elements in a decision matrix and weight coefficients to obtain the decision index, wherein the weight coefficients are inversely proportional to absolute values of differences of real risk levels and predicted risk levels in classification results.
5. The urban rainfall flood risk early warning method based on the multi-source data according to claim 1, wherein the normalization of the decision indexes to obtain the voting weights of the decision trees comprises the step of dividing the decision indexes of the decision trees by the sum of the decision indexes of all the decision trees in a random forest to obtain the voting weights of the decision trees.
6. The urban rainfall flood risk early warning method based on multi-source data according to claim 1, wherein the multi-source rainfall flood data comprises meteorological data and water regime data.
7. The urban rainfall flood risk early warning method based on multi-source data according to claim 6, wherein the meteorological data comprise rainfall intensity and accumulated rainfall data, and the water regime data comprise water level and flow velocity data of key nodes of a main river channel, a waterlogged point and a drainage pipe network of the city.
8. The urban rainfall flood risk early warning method based on multi-source data according to claim 1, wherein the early warning method further comprises preprocessing the multi-source rainfall flood data, and the preprocessing comprises filling processing and normalization processing.
9. The urban rainfall flood risk early warning method based on multi-source data according to claim 1, wherein the monitoring window comprises a preset number of moments before the acquisition moment of the historical multi-source rainfall flood data.
10. The urban rainfall flood risk early warning system based on the multi-source data is characterized by comprising a processor, a memory and a communication interface, wherein the memory is stored with a computer program, and the computer program realizes the urban rainfall flood risk early warning method based on the multi-source data according to any one of claims 1 to 9 when the computer program is executed by the processor.

Description

Urban rainfall flood risk early warning method and system based on multi-source data Technical Field The application relates to the technical field of data processing. More particularly, the application relates to a city rainfall flood risk early warning method and system based on multi-source data. Background The urban rainfall flood risk early warning system is a comprehensive information management system for monitoring, analyzing, predicting and publishing flood disaster risks such as surface water accumulation, pipe network overflow, river channel flood bank and the like caused by heavy rainfall in a city in real time by integrating multisource data such as meteorological, hydrologic, geographic information and urban infrastructure. The system has the main effects that scientific decision support is provided for urban managers and the public by issuing early warning information in advance so as to start emergency plans, evacuate personnel and schedule flood control facilities in time, thereby reducing the loss caused by rain and flood disasters to the greatest extent. In the prior art, a random forest model is used for early warning of urban rainfall flood risks. The model builds a plurality of decision trees through an integrated learning strategy and predicts the result by integrating the decision trees. However, due to different data sources used in the construction of the decision trees, the early warning accuracy of different decision trees on the rainfall flood risk is different, and when the standard random forest model is finally decided, the simple voting or average mode is adopted to give the identical weight to each decision tree in the forest, namely, the prediction capability of different decision trees on different types and different stages of rainfall flood events is different, for example, part of decision trees can be more accurate in predicting short-time heavy rainfall events, and the other part of decision trees can be more accurate in predicting long-time accumulated rainfall events. Therefore, the indiscriminate voting mechanism ensures that the decision speaking right of the expert tree which is better in performance under a specific early warning scene is averaged, thereby limiting the whole prediction precision of the model and failing to meet the requirement of urban rainfall flood risk early warning on high precision. Disclosure of Invention The application provides a city rain and flood risk early warning method and system based on multi-source data, and aims to solve the problem that prediction accuracy is insufficient due to the fact that decision tree weights are the same in a traditional random forest model. In a first aspect, the application provides a city rain and flood risk early warning method based on multi-source data, which comprises the steps of constructing a risk index of any history multi-source rain and flood data, including multiplying a difference value between a numerical value and a maximum value of any dimension data in the history multi-source rain and flood data with a change trend of the dimension data in a monitoring window, taking the sum of products of the dimension data as the risk index of the history multi-source rain and flood data, dividing the history multi-source rain and flood data into a training set and a testing set, constructing a random forest by using the training set, acquiring a prediction risk level of each decision tree of the testing set in the random forest, constructing a decision matrix of each decision tree based on the risk index and the prediction risk level of the history multi-source rain and flood data in the testing set, weighting and summing elements in the decision matrix to obtain the decision index of each decision tree, normalizing the decision index to obtain the voting weight of each decision tree, inputting the real-time multi-source rain and flood data into the random forest, and carrying out prediction risk level prediction according to the voting weight to the prediction risk level to obtain a prediction result. And analyzing the change condition of the historical multi-source rain and flood data, combining the current state with the change trend, and constructing a risk index reflecting the degree of the potential rain and flood risk of the city. And constructing a decision matrix by combining the risk indexes and the predicted risk levels of the decision tree on the test set data, and evaluating the identification effect of the decision tree on the multi-source rain and flood data with different risk degrees in the subsequent step. And constructing a decision index reflecting the accuracy of the decision tree to the multi-source rain and flood data identification in high risk by analyzing the size and the position of each element value in the decision matrix. Different voting weights are given to different decision trees based on the decision indexes, the problem of insufficient prediction precision cause