一种基于决策树技术的中长期径流分级方法

Medium-and-long-term runoff classification method using decision tree technique

常用径流分级方法仅从径流本身特征出发,未与水库实际防洪发电调度相结合,因此关于径流丰平枯状态的描述及对应的调度策略并不能反映出水库实际运行特征。为此,本文提出径流分级决策树法,该方法利用决策树技术分析挖掘水库实际水文和发电调度数据,从中找出符合该电站实际运行特征的径流分级标准。通过二滩水电站发电调度实例分析,相对于保证率法的径流分级标准,径流分级决策树法基本上能反映出二滩水电站丰水期实际的运行特征,在发电调度中具有一定的优势。

This paper presents a new method for classification standards of medium-and-long-term runoff by analyzing a mass of practical hydrological data and reservoir operation information using decision tree technique.This model is established by taking into account the characteristics of flood control and hydropower operation,which is different from the traditional ones such as the cumulative occurrence frequency method.To validate the effectiveness of the proposed method,the reservoir operation polices and their simulations are also established for available standards using decision tree technique.The results show that the proposed method performs relatively much better and a higher hydropower production is obtained.