Flood forecasting scheme real-time optimization method based on machine learning

Abstract

The invention discloses a flood forecasting scheme real-time optimization method based on machine learning. The flood forecasting scheme real-time optimization method comprises the following steps: 1)collection and processing of basin hydrological data; 2) rainfall and flood session division and association; 3) generation of a rainfall flood event sample set; 4) flood grading; 5) construction ofa flood forecasting scheme; 6) classifier training based on machine learning; and 7) forecasting scheme real-time optimization based on early rainfall. According to the flood forecasting scheme real-time optimization method, the sample set is divided into subsets on the basis of flood levels, and the flood forecasting schemes are compiled respectively and associated with the early rainfall processthrough a machine learning method, and optimization of the flood forecasting schemes during real-time forecasting is achieved, and the real-time flood forecasting precision of the drainage basin canbe effectively improved.

Description

technical field [0001] The invention belongs to the technical field of water conservancy engineering, in particular to the technical field of flood forecasting, and specifically relates to a machine learning-based real-time optimization method for a flood forecasting scheme. Background technique [0002] As an important part of non-engineering measures, flood forecasting can effectively improve the disaster prevention and mitigation capabilities of river basins and regions. At present, more than 1,700 national basic hydrological stations have achieved normalized forecasting work, more than 200 control sections of large rivers and lakes and more than 700 medium-sized reservoirs have realized routine flood forecasting, and the national hydrological system produces and releases floods of important river, lake and reservoir sections every day during the flood season More than 5,800 stations have been forecasted, and flood forecasts with different forecast periods and accuracy ha...

AI Technical Summary

Problems solved by technology

However, how to construct multiple schemes for the same forecast section and how to quickly select an appropriate scheme in real-time forecasting are problems that have not been well solved in practical applications

Images