降雨径流模拟神经网络模型及应用

Rainfall-runoff modeling with artificial neural network and genetic algorithm

针对水文系统的非线性,构建了基于遗传算法和人工神经网络的降雨径流模拟神经网络模型(GA-BP模型).采用附加动量法和自适应学习速率对BP神经网络进行改进,遗传算法用于优化神经网络的初始权重.以大别山及皖南山区月潭流域为例,将GA-BP模型、BP模型以及新安江模型应用于水文日径流过程模拟,进行应用比较以及分析GA-BP模型在水文径流模拟过程中的难点及其可行性.结果表明,GA-BP模型优化了网络结构,加快了算法收敛速率;可以用于降雨径流过程模拟,也为今后类似研究提供一种模拟技术.在实际应用中可以根据流域资料情况选择合适的模型进行水文模拟作业.

Considering the non-liner of hydrologic system, an intelligent model, based on ANN and GA, was developed for simulating hydrological daily rainfall-runoff process. BP model is improved with additional momentum algorithm and self- adaptive learning rate algorithm. The original weights of artificial neural networks are optimized by the genetic algorithm. Taking the Yuetan watershed in the south of Anhui province as an example, to evaluate the performance of the developed model, Xinanjiang model and the original BP model were conducted for comparing to the intelligence model so as Reasons and key technologies of applying the improved model in hydrologic simulation were analyzed. The simulation results show that the intelligence model can optimize network structure and accelerate arithmetic convergence and is a successful tool to simulate hydrologic process and it provides a good technique for simuating daily rainfall-runoff and the similar problems. The developed model can be applied in terms of real-time hydrologic information in practice.