摘要
由博斯腾湖1956～1998年的水位动态变化特征可知,其水位具有明显的季节性变化特征.根据1956～1996年的实测水位数据,分别建立了ARIMA(1,1,2)×(0,1,1)12模型和BP神经网络模型,对1997～1998年的博斯腾湖水位进行了预测.结果表明,ARIMA模型和BP网络模型可以进行比较准确的短期预测,而且BP网络的预测更准确,但长期预测误差较大.
The water levels in Bosten Lake, Xinjiang Region vary significantly from season to season. An ARIMA (1,1,2)×(0,1,1)12 model and BP neutral network model were built to predict its water levels, based on the data measured from 1956～1996. The results obtained indicated that these models can provide accurate data for short term prediction, and the BP neutral model has better accuracy in prediction, but its long,term prediction has less accurate data.
出处
《水利水电技术》
CSCD
北大核心
2004年第5期5-7,共3页
Water Resources and Hydropower Engineering
基金
高等学校博士学科点专项科研基金(20020491011)项目.
