基于人工神经网络的降雨径流模拟研究

Study on rainfall-runoff simulation based on artificial neural networks

针对水库径流难以预测的问题,采用改进的动量-自适应学习率调整BP神经网络方法,以南告水库作为研究对象,对水库的日资料进行径流模拟,并对该模型在径流模拟中的方法和难点问题进行分析和探讨。改进的BP模型模拟的结果与三水源新安江模型的模拟结果相比较,探讨改进的BP模型应用于水文模拟的可行性。研究结果表明,改进的BP模型用于水文模拟是可行的。

In view of the difficult predictions of reservoir inflow runoff, an improved back-propagation （BP） neural networks model is proposed in this paper. It is the joints of additional momentum algorithm and self-adaptive learning rate algorithm. Approaches and key technologies when applying the improved model in runoff simulation are discussed. The improved BP model is applied for simulating daily streamflows in the upper area of Nangao Reservoir at Shanwei City, Guangdong Province, China. The experiment results demonstrate the applicability of the improved BP model. Comparison with the Xinanjiang model shows that this method is feasible and effective. It also demonstrates that the ANN is a promising tool for simulating various hydrologic processes.