Abstract:

The accuracy of the peak flood flow prediction for earth-rock dam breaches is of decisive significance for the conclusions of dam break analysis. To enhance the prediction accuracy of the peak flood flow following a dam breach, this paper introduces a prediction model based on General Regression Neural Network (GRNN), coupled with Fennec Fox Optimization (FFA) for hyperparameter optimization, to forecast the peak flood flow resulting from dam breaches. Utilizing a database of domestic and international dam failure incidents, the model employs three factors as input variables: the reservoir capacity above the breach bottom, the water depth above the breach bottom, and the breach depth, to construct the FFA-GRNN dam break peak flood flow prediction model. To assess the precision and fitting accuracy of the model in predicting peak flood flow values, a comparative verification was conducted against four other intelligent algorithms. The results indicate that the proposed FFA-GRNN model exhibits a lower RMSE (592), MAE (995), and a higher coefficient of determination R² (0.973) in comparison to other models, demonstrating superior computational precision and fitting performance overall. By analyzing the applicability of the model in predicting dam breach peak flood flows, it can provide technical support for dam break analysis.

Key words: dam break, peak outflow, earth-rock dam, Fennec Fox Algorithm, Generalized Regression Neural Network