基于时间序列的山区风向预测方法研究

为更好的预测山区风向,分别使用自回归滑动平均模型(Auto-Regressive Moving Average Model,ARMA)、极端梯度提升算法(eXtreme Gradient Boosting,XGBoost)和长短期记忆网络(Long Short Memory Network,LSTM)对山区4个气象站的风向时间观测序列进行预测,比较三种算法的优劣。针对风向独特的性质,结合传统气象学知识,探索了一种风向的预测方法：引入风速序列,将风速、风向拆分为U、V风,对其分别进行预测,进而将结果合成,使用设定的风向评分检验结果。对比结果表明：1)XGBoost获得了最高的评分；2)使用双时间序列预测风向相比单序列效果更好。

Research on Wind Direction Prediction Method in Mountainous Area Based on Time Series

In order to predict the wind direction in mountainous areas better, the Auto-Regressive Moving Average Model (ARMA), the extreme gradient boosting algorithm (XGBoost) and the Long Short Memory Network (LSTM) are used respectively. Then the advantages and disadvantages of the three algorithms are compared. In view of the unique nature of wind direction, combined with traditional meteorological knowledge, a wind direction prediction method is explored: introducing wind speed series, splitting wind speed and wind direction into U and V winds, respectively predicting them, and then combining the results, using the set wind direction scoring test result. The comparison results show that: 1) XGBoost obtains the highest score; 2) Using dual time series to predict wind direction is better than a single series.