Research on flood peak prediction in the three gorges region based on similarity search with multisource information fusion

To improve the accuracy of flood peak prediction for rainfall in the Three Gorges Reservoir, this article proposes a multisource information fusion similarity search model. Therefore, the K-nearest neighbour algorithm is improved to mine the hydrological state similarity between rainfall instances, and a Siamese network is constructed to calculate the temporal and spatial dimension similarity between instances. Lasso regression multivariate linear fitting was used to obtain accuracy weights for multisource similarity, calculate the comprehensive similarity between instances, and flood peaks from similar historical instance were weighted and summed to obtain predictions. The results show that the fusion model of multisource information improves the R values by 2.6%, 16.0%, and 5.8%, respectively, compared to those from using independent models for prediction, and the RMSE values decrease by 3.2%, 22.1%, and 32.6%, respectively. This finding fully demonstrates that the multisource information fusion method in this article can effectively improve the accuracy of flood peak forecasting for rainfall.