基于人工神经网络算法的新疆百里风区风速预报方法研究

Wind speed prediction method based on ANN arithmetic for 100-kilometer wind area in Xinjiang

为了给铁路运输提供准确的大风预警,文中将人工神经网络算法应用于新疆百里风区的风速预报,分别采用BP神经网络与Elman神经网络建立模型,对实际历史风速数据进行仿真预测。利用实时自动站资料预测未来20分钟瞬间风速并做预报对比检验。结果表明:与BP神经网相比,Elman人工神经网络模型具有更好的拟合效果,独立样本预报及实际预报的检验结果均达到了较为精确的效果,具有实际应用意义。

In order to provide accurate gale early warning for railway transportation in Xinjiang, the Artificial Neural Network arithmetic was used to the wind speed forecast at 100 - kilometer wind area. In this paper, the forecast model was established by using BP neural network and Elman neural network respectively, and applied to simulate and forecast the real historical wind speed data. Using measured data of automatic weather station nearby railway, such as pressure, temperature and instantaneous wind speed, the instantaneous wind speeds in first 20 minutes were predicted. The simulation result show that the predicted accuracy by Elman neural network is superior to that by BP Artificial Neural Network, and its inspection result of independent sample forecast and actual forecast are very accurate. The research results indicate that the Elman neural network used for wind speed forecast is of practical.