Flood Forecasting Method Based on Cluster Analysis and Real-time Correction

Abstract

The invention discloses a flood forecasting method based on cluster analysis and real-time correction. The steps are as follows: firstly, using principal component analysis (PCA) to perform dimension reduction processing on model input. The second is to use the K-means clustering method to cluster the original data. Divide the flood data into different categories, and then train different SVM models. When the test sample is input, the cluster centroid is used to judge the category of the test sample, and the corresponding model is used to predict it to obtain the predicted value q; the third is BP The neural network corrects in real time. Calculate the error sequence between the predicted value and the real value, use the error sequence data to train the BP neural network error correction model, and obtain the error correction value q e , the final forecast result is the model forecast value q plus the error forecast value q e . The invention has the advantages of: dividing the original hydrological data into several categories through clustering analysis, training models respectively to realize multi-model forecasting; and then realizing real-time correction through the BP neural network to improve the forecast accuracy of flood peak time.

Description

technical field [0001] The invention belongs to the technical field of water flow forecasting, in particular to a flood forecasting method based on cluster analysis and real-time correction. Background technique [0002] Flash floods are sudden floods caused by rainfall in small and medium-sized rivers in mountainous areas. However, there are many small and medium-sized rivers in my country, and most of the small and medium-sized rivers have a sparse network of stations. The necessary emergency monitoring methods are lacking, and the forecasting scheme is not perfect. In recent years, due to the increase in extreme weather events, local heavy rainfall and sudden torrential rain often occur, while mountainous areas have high mountains and steep slopes, short river sources and rapid streams, and mountain disasters such as mountain torrents, mudslides, and landslides are prone to occur in heavy rainy weather , resulting in the loss of people's lives and property. Therefore, th...

AI Technical Summary

Problems solved by technology

Although the combined model solves the multi-model problem contained in the historical flood data, the problem of the general deviation of the data-driven model for the forecast results of the flood peak time has not been well solved

Images