A refined rank set pair analysis model based on wavelet analysis for predicting temperature series

作者：Zhang, Jian; Yang, Xiao-Hua^*; Li, Yu-Qi

来源：International Journal of Numerical Methods for Heat and Fluid Flow, 2015, 25(5): 974-985.

DOI：10.1108/HFF-05-2014-0140

• Temperature
• Prediction
• Refined rank set pair analysis
• Wavelet analysis

Purpose - The purpose of this paper is to accurately simulate and predict the daily extreme temperature in Beijing Reservoir and the monthly extreme temperature in Tianjin Reservoir using wavelet refined rank set pair analysis (WRRSPA).
Design/methodology/approach - The new method, called WRRSPA, which combines wavelet analysis and refined rank set pair analysis (RRSPA), was proposed for use in this study because of the non-linear and multi-time scale characteristics of the temperature series. The model includes the advantages of the multi-resolution feature of wavelet analysis and the non-parametric data-driven prediction from refined rank set air analysis.
Findings - Based on the daily extreme temperature of Beijing Reservoir, the predictions of the last 18 days reveal that WRRSPA is more appropriate because the percentage of the relative errors that are smaller than 10 percent increased from 78 percent by Back Propagation (BP) and 78 percent by RRSPA to 100 percent by WRRSPA in Beijing Reservoir. In addition, WRRSPA has lower values of root mean squared error (RMSE) and mean absolute error (MAE) and a higher coefficient of efficiency (modified coefficient of efficiency (MCE)). The last 12 monthly extreme temperature predictions of Tianjin Reservoir demonstrate that WRRSPA produces prediction results: the percentage of relative errors that are smaller than 10 percent are improved from 34 percent by BP and 58 percent by RRSPA to 67 percent by WRRSPA. In addition, WRRSPA also has lower values of RMSE and MAE and a higher coefficient of efficiency (MCE).
Research limitations/implications - The analysis results ignore the physical processes and may be affected by the limited observation data. In addition, the WRRSPA method is still in its early stages of application and must be further tested.
Practical implications - The results of the study are helpful for the study of the complex features and accurate prediction of temperature series.
Social implications - This paper contributes to further the process of research of climate change.
Originality/value - This study represents the first use of the WRRSPA method to analyze the multi-scale characteristics and forecast the future values of the extreme temperature series from Beijing Reservoir and Tianjin Reservoir. This paper provides an important theoretical support for extreme temperature prediction.

• 出版日期2015
• 单位北京师范大学