基于箱形图数据清洗的水电站特性曲线修正方法

Characteristic Curve Correction Method of Hydropower Plant Based on Box-Plot Data Cleaning

摘要: 水电站和机组长期运行后，下泄流量~尾水位、水头~耗水率等特性曲线相较于设计参数存在偏差，影响水电计划编制的准确性，该问题在西南地区巨型梯级水电站群普遍存在，需要研究行之有效的解决方法。基于南方电网水电调度生产实践，提出基于箱形图数据清洗的水电站特性曲线修正方法。利用水电站海量历史运行数据构造大数据样本，通过箱型图模型去除异常数据，采用多项式拟合技术对设计曲线进行修正，能够更好地反映发电特性曲线真实关系。实际应用算例证明了方法的有效性。

Abstract: The characteristic curves, such as tailwater level and net head curve, deviate from the design parameters during the long-term operation of hydropower units, which is difficult to make an accurate hydropower scheduling plan. It is necessary to find an effective solution to address the problem, especially for the giant cascade hydropower stations in Southwest China. A correction method of characteristic curve of hydropower unit based on the box-plot data cleaning is proposed and applied in the hydropower dispatching in China Southern Power Grid. We utilized the massive historical operation data of hydropower station to construct large data samples, used the box-plot model to remove the abnormal data, and adopted the polynomial fitting technology to modify the characteristic curve. The results show that this method can better describe the power generation characteristic curve, and its effectiveness is proved by practical application examples.