基于梯度自适应优化的入库流量数据平滑算法研究

Research on smoothing algorithm of reservoir inflow data based on gradient adaptive optimization

为了解决入库流量等时序数据跳变而影响对数据变化趋势规律判断的问题,结合水量平衡原理计算入库流量现实存在“跳变”的局限性,提出了梯度自适应迭代平滑算法,设计了数据序列光滑度评价计算方法,并阐述了其详细的计算原理和计算步骤。以三峡水库实际计算入库流量为例,将梯度自适应迭代平滑算法的平滑过程分别与3点、5点、7点线性平滑,5点、7点二次函数拟合平滑,5点、7点三次函数拟合平滑的结果进行对比。结果表明:该算法在入库流量平滑计算中具有较好的表现;三峡水库两段比较具有代表性的2月份和4月份小时计算入库流量数据序列经平滑计算后,数据序列的平均光滑度分别达到了0.9584和0.9803。该算法能够很好地适应原始数据光滑度较差的序列平滑计算,具有很好的实际应用推广价值。

In order to solve the problem that the jump of time series data such as inflow affects the judgment of the data change trend,we proved the limitation of"jump"in inflow based on the principle of water balance,proposed a smooth gradient adaptive iteration algorithm(SGAIA),designed a calculation method for evaluating the smoothness of data series,and expounded its detailed calculation principle and calculation step.Taking the actual inflow calculation of the Three Gorges Reservoir as an example,we compared the smoothing results of SGAIA with other methods,including 3/5/7-point linear smoothing algorithm,5/7-point quadratic function fitting smoothing algorithm,and 5/7-point cubic function fitting smoothing algorithm.The results showed that SGAIA has better performance in the smoothing of inflow data.Then we execute smooth calculation on two more representative data series of hourly calculated inflow of the Three Gorges Reservoir in February and April,and the average smoothness of the data series after smoothing calculation reached 0.9584 and 0.9803,respectively.SGAIA can be well adapted to data series smoothing calculation in case of poor original data smoothness.