基于小波神经网络的气温预测研究

Research on Temperature Forecasts Based on Wavelet Neural Network

主要参考温度信号的固有特性,以1951-2014年的重庆市温度数据为例,使用小波神经网络（WNN）对温度进行预测估计。实证结果表明,该研究建立的小波神经网络能够对未来气温进行较好的预测,进而可应用于天气衍生品定价等领域,实现对冲天气风险。

The paper sets up weather change＇ s stochastic model basing on temperature inherent characteristic. Taking temperature data of Chongqing from 1951 to 2013 as training set, it estimates the temperature in wavelet neural network model. The results of empirical simulation and model verification show that the model has relative minor error and wavelet neural network can better imitate temperature index in the future. Then, it can apply in pricing weather derivatives.