基于随机森林和RBF人工神经网络模型的新丰江水库枯季入库径流中长期预报

Medium-long-term Forecast of the Runoff of the Xinfengjiang Reservoir in Dry Season Based on the Random Forest and RBF Artificial Neural Network

基于随机森林和RBF人工神经网络构建了新丰江水库枯季入库径流中长期预报模型,首先采用随机森林模型从74个水文气象特征量和前期降雨、径流中筛选预报因子,之后利用筛选的预报因子作为RBF神经网络的输入层,利用RBF神经网络对新丰江水库枯季入库径流每月的流量进行预报。结果表明,基于随机森林和RBF人工神经网络模型的枯季径流中长期预报模型精度较高,其中训练期平均合格率为91.24%,平均相对误差为7.80%,检验期平均合格率为67.31%,平均相对误差为26.73%,模型有较高的可靠性,预报结果可作为东江流域枯季径流预报重要参考依据。

Based on the random forest and RBF artificial neural network, this paper establishes the medium-long-term forecast model for the runoff of Xinfengjiang reservoir in dry season, filters the forecast factors from 74 hydrometeorological features and pre-rainfall and runoff by the random forest, and then forecasts the monthly runoff of the Xinfengjiang reservoir in dry season by RBF neural network with the filtered forecast factor as the input layer. The results show that the medium-long-term forecast model for the runoff in dry season based on the random forest and RBF artificial neural network model has higher accuracy, among which the average pass rate in training period is 91.24%, with the average relative error of 7.80%;while the average pass rate in the test period is 67.31%, with the average relative error of 26.73%, so the model has high reliability. The results can be used as an important reference for the forecast of runoff in the Dongjiang River basin in dry season.