摘要
针对海量气象观测数据间存在大量的物理噪声、与气温无关的冗余特征以及时间相关性,提出了一种将一维卷积神经网络(1DCNN)和长短期记忆神经网络(LSTM)相结合的多信息融合气温预报方法。首先,运用差分法对气象观测数据进行预处理,得到平稳时间序列数据;其次,运用1DCNN提取与气温变化相关的特征变量作为神经网络模型的输入变量;最后,运用1DCNN和LSTM构建多信息融合气温预报模型1DCNNLSTM,并以云南省昆明市历史气象观测数据为例,与传统的LSTM、1DCNN和反向传播神经网络(BP)对未来24小时的逐时气温预报进行了比较研究。研究结果表明,1DCNN-LSTM的均方根误差(RMSE)相较于LSTM、1DCNN和BP最大降低了5.221%、19.350%和9.253%,平均绝对误差(MAE)最大降低了4.419%、17.520%和8.089%。为气温的精准预报提供了参考依据。
To deal with the large amount of physical noise, redundant features unrelated to temperature,and time correlation between massive meteorological observational data, a multi-information fusion temperature forecast method combining one-dimensional convolutional neural network(1DCNN) and long short-term memory neural network(LSTM) is proposed. First, the difference method is used to preprocess the meteorological observational data to obtain stationary time series data. Second, 1DCNN is used to extract feature variables related to temperature changes as the input variables of the neural network model.Finally, 1DCNN and LSTM are used to establish a multi-information fusion temperature prediction model1DCNN-LSTM. Taking the historical meteorological observational data of Kunming City in Yunnan Province as an example, the model is compared with the traditional LSTM, 1DCNN and Back Propagation Neural Network(BP) of the hourly temperature forecast in the next 24 hours. The results show that compared with those of LSTM, 1DCNN and BP, the root mean square error(RMSE) of 1DCNN-LSTM is reduced by 5.221%, 19.350%, and 9.253%, and the mean absolute error(MAE) is reduced by 4.419%,17.520% and 8.089%, respectively. This research method provides a reference for the accurate prediction of air temperature.
作者
李晶
唐全莉
LI Jing;TANG Quanli(Ningbo University of Technology,Ningbo,Zhejiang 315000,China;Faculty of Science,Kunming University of Science and Technology,Kunming 650500,China)
出处
《热带气象学报》
CSCD
北大核心
2022年第6期800-811,共12页
Journal of Tropical Meteorology
基金
浙江省统计局重点项目(22TJZZ25)
国家自然科学基金(71964018)
云南省社会科学基金(YB2019036)
昆明理工大学学生课外学术科技创新基金(2020YB207)共同资助。