RBF神经网络在径流时间序列预测中的应用

Application of RBF Neural Network in Runoff Time Series Prediction

利用RBF神经网络建立了径流的时间序列预测模型,对其原理和相应的计算步骤进行了介绍。实例应用结果表明:①该模型收敛速度快、预报精度较高,结果优于传统的AR模型;②应尽可能采用更大容量的训练样本,获得更好的预测性能;③历史径流资料应选取未受人类活动干扰或受人类活动影响较小的时间序列来进行分析。

RBF neural network is applied to the runoff time series prediction,and the detailed model principle and the corresponding calculation steps are explained and discussed.The application case shows that ①the model has the ability of fast convergence and high precision and the forecast precision is higher than traditional automatic regressive model;②for better prediction performance,greater samples capacity should be adopted in the training process;③historical runoff data should be analyzed by selecting time series that are unaffected or less affected by human activities.