【摘要】 由于高铁超声波测风仪碍于安装环境限制易受外部干扰产生秒级异常风速，而现有的基本质量控制方法与气象领域分钟级时距质量控制算法无法准确识别秒级异常风速，导致监测风速不能很好地满足高铁秒级风速监测预警需求。为了解决该问题，基于高铁沿线风速监测异常值数据特征，提出一种基于PI-BBI的秒级风速监测数据质量控制算法。该方法首先利用基于物理信息的双向长短期记忆网络对风速异常值进行预测，并得到预测误差；接着通过改进的指数加权移动平均法对预测误差进行平滑；最后使用孤立森林检测平滑误差序列中的异常值，从而能够识别原始风速序列中的异常值。实验结果显示，该质量控制算法能有效改善高铁沿线风速监测数据的质量，从而提高监测预警的准确性。更多还原

【Abstract】 The ultrasonic anemometer of high-speed rail is susceptible to external interference to produce second-level abnormal wind speed due to the installation environment. The existing basic quality control methods and minute-level time-interval quality control algorithms in the meteorological field cannot accurately identify the second-level abnormal wind speed, resulting in that the monitoring wind speed can not well meet the second-level wind speed monitoring and early warning requirements of high-speed railway. To solve the problem, a second-level wind speed monitoring data quality control algorithm based on PI-BBI was proposed according to the abnormal value data characteristics of wind speed monitoring along the high-speed railway. Firstly, a two-way memory network with long-and short-term was used to predict wind speed anomalies based on physical information and obtains the prediction error. Then the prediction error was smoothed by an improved exponential weighted moving average method. Finally uses an isolated forest was used to detect outliers in the smooth error sequence,enabling the identification of outliers in the raw wind speed series. Experimental results show that this quality control algorithm can effectively improve the quality of wind speed monitoring data along the high-speed railway. The accuracy and predictability of wind speed monitoring and early warning are improved.更多还原