距离度量学习法在水库静库容调洪演算中的应用

静库容法和动库容法是水库调洪演算最常用的方法,与动库容法相比静库容法简单快速,所需资料较少,在水库蓄水期和消落期可直接用于水库调洪。但由于未考虑入库向出库转换的时间延迟,在汛期易造成较大的库水位计算误差。对此,设计了距离度量学习算法,对当前的预报样本,先使用静库容法计算库水位;然后根据学习的距离度量,从历史样本中寻找若干个与当前水情相似的样本,用这些相似样本的库水位计算误差的线性组合对当前样本的计算库水位进行修正,改善其精度。以三峡水库2009～2017年的数据作为训练集,以2018年的数据作为测试集,试验结果表明距离度量学习算法能够显著改善汛期库水位的计算精度。

Application of Distance Measurement Algorithm in Calculation of Flood Regulation Based on Static Reservoir Capacity

Static storage capacity method and dynamic capacity method are the most commonly used methods for flood regulating calculation of reservoirs. Compared with dynamic storage capacity method, static storage capacity method is simple and fast, and requires less data, which can be directly used for flood control of reservoirs during the storage period and down period. However, because it fails to take into account the time delay of the flow transition in large reservoir, it is easy to cause large water level prediction error in the flood season. Aiming at this problem, this paper designs a distance measurement learning algorithm. For the current prediction sample, the static reservoir capacity method is used to calculate the water level of the dam. Then, according to the learned distance measurement, several samples similar to the current situation are found from the historical samples. The calculation error of the current sample are corrected by using the linear combination of the calculation errors of these similar samples, so as to achieve the goal of improving the accuracy of water level prediction. Taking the historical data of the Three Gorges Reservoir from 2014 to 2017 as the training set and the data from 2018 as the test set, the experimental results show that the distance measurement learning algorithm can significantly improve the prediction accuracy of the water level of the dam for flood season.