Hydrological multi-model time-varying weight combination forecasting method

Abstract

The invention discloses a hydrological multi-model time-varying weight combination forecasting method, and the method comprises the following steps: determining a plurality of hydrological models participating in combination forecasting, and obtaining each piece of hydrometeorological data in each historical flood session of a target watershed; calculating a weight value of a single hydrological model at each moment in each historical flood by using a Bayesian model averaging algorithm; screening out each preferred flood sample most similar to the current flood; selecting k flood samples with the highest matching degree with the current flood by using a KNN method, obtaining a forecast weight value of each hydrological model through the idea of inverse distance weighting, and carrying out weighted summation by combining the flood flow forecast values of each hydrological model at the forecast moment to obtain a fused flood flow forecast value of the target drainage basin. And forecasting of the flood flow of the target drainage basin is realized. According to the method, the multiple hydrological models are fused and forecasted through the Bayesian theory, updating combination of the time-varying weights is achieved with the help of the KNN learning algorithm, the precision of the flood forecast value is improved, and a basis is provided for flood control decision making.

Description

technical field [0001] The invention belongs to the technical field of hydrology, and in particular relates to a combined forecasting method of hydrology multi-model time-varying weights. Background technique [0002] Medium- and long-term runoff forecasting is of great significance to flood control and drought relief, water resources management, and water conservancy project scheduling. Therefore, how to improve the accuracy of medium and long-term runoff forecast is one of the most difficult problems in the field of hydrology. [0003] In today's era of big data, data mining technology has attracted the attention of scholars. As a classic data classification method, KNN (K-Nearest Neighbor) algorithm can effectively organize and manage a large amount of text data, and is currently widely used in many fields such as weather forecasting and intelligent learning. The KNN method has a simple idea, clear logic, easy to understand and implement, and does not need to estimate p...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the problems of poor applicability to historical big data and low forecasting accuracy in medium and long-term runoff forecasting, the present invention provides a hydrological multi-model time-varying weight combination forecasting method, Fully consider the flood process data of all historical events, realize the time-varying weighted combination forecast of multiple hydrological models, and improve the forecast accuracy

Images