机器学习在河流流量参数估计中的应用

针对经典水位流量关系模型在刻画河流动态变化特性时存在的局限性,提出采用局部加权回归算法估计河流流量;为了提高参数估计精度,提出一种聚类局部加权回归方法。首先对训练样本进行聚类,然后使用k-最近邻方法将新的水位样本划分进最恰当的聚类中,最后估计河流日流量。该方法在估计过程中,避免了不相关信息的干扰,从而提高了日流量数据估计的效率和精度。利用某水文站的实测数据对方法进行测试,仿真结果表明该方法估计精度较高,为水位流量关系模型参数估计提供了新的有效方法。

Parameters Estimation of River Flux Based on Machine Learning

Aiming at the limitation for classical water level and water flux relationship in accurately describing the dynamic characteristics of water,a locally weighted regression approach was utilized to estimate river flux.In order to improve the precision,a method named clustering locally weighted regression was used to estimate the variables such as water flux,water level.First,the train instances were clustered.Second,k-nearest neighbors were used to cluster new instances which were the testing data into the best fit clustering.Finally,river flux was computed.The test with factual data at some hydrological station showed that the method proposed had good performance for parameter estimation precision and efficiency,and provided new method for parameter estimation of water level and water flux relationship.