遗传编程在降雨径流模拟中的应用研究

Application Study on Genetic Programming Approach for Rainfall-runoff Simulation

针对降雨径流过程的复杂性,介绍了遗传编程的基本原理和方法,建立了王家厂水库流域不同输入变量情况下基于遗传编程的降雨径流模拟模型,讨论了遗传编程符号集和终止集对模拟精度的影响。结果表明,遗传编程可用于降雨径流模拟并可给出明确的降雨径流关系式,便于理解复杂的降雨径流过程。

Rainfall-Runoff modelling plays an important role in flood control and water resources sustainable manage- ment. For the comlexity of rainfall-runoff process, Genetic Programming （GP） provides a new data-driven modelling frame. In this paper, the theory and method of GP is introduced. Developed wangjiachang reservior catchment model using GP is discussed and the models with different number of input variables that represent the parameters of GP are compared in order to improve the modelling accuracy. The results indicate that GP can be considered as an efficient method to rainfall-runoff modelling. And compared with Artificial Neural Network, GP can provide the understanding for complicated rainfall-runoff process by giving the explicit modelling equation.