摘要
基于降水过程存在周期性、随机性的特点,应用时间序列典型分解法提取原降水量序列中的趋势成分和周期性成分,对于剩余平稳序列成分,采取BP神经网络模型对其进行模拟;最后建立降水量的BP神经网络时间序列预测模型。以宿迁市近14年的月平均降水资料为实例对该模型进行了具体的应用。结果表明:基于BP神经网络时间序列预测模型可以有效地预测降水量,并和传统的时间序列加法模型进行了比较,结果显示基于BP神经网络的时间序列预测优于传统的时间序列加法模型,模型具有较高的精度和稳定性。
This paper first applied time series method to set up the classification standard of precipitation based on the fact that there are uncertainty and imprecise characteristics in the precipitation course;then presented a method which is called Markov chain with weights to predict the future precipitation state by regarding the standardized self-coefficients as weights based on the special characteristics of precipitation being a dependent stochastic variable;and applied this method to a real hydrological observation station with nearly 14 years precipitation information in Suqian City at last,an ideal result was obtained.
出处
《水资源与水工程学报》
2010年第5期156-159,共4页
Journal of Water Resources and Water Engineering
