ARIMA-RBFNN组合模型在白城市降水量预测中的应用  

Application of ARIMA-RBFNN Model in Prediction of Precipitation in Baicheng City

作　　者：安永凯[1] 卢文喜[1] 宋文博[1] 贺石良[1] 赵莹[1] 

机构地区：[1]吉林大学环境与资源学院,吉林长春130026

出　　处：《水电能源科学》2014年第6期25-28,共4页Water Resources and Power

基　　金：国家自然科学基金项目(41072171)

摘　　要：针对降水受大气环流、地形、气压等诸多环境因素影响致使准确预报降水量较为困难的问题,结合ARIMA模型和RBFNN模型各自优势,提出了ARIMA-RBFNN组合模型,对白城市2001~2010年降水量进行了预报,并与ARIMA模型和RBFNN模型预报结果进行了对比分析。结果表明,ARIMA-RBFNN组合模型在预测降水量时最大相对误差为27.33%,最小相对误差为0.70%,平均相对误差为8.54%,预测精度明显优于ARIMA模型和RBFNN模型,可见该组合模型发挥了ARIMA模型和RBFNN模型各自的优点,为精确预测降水量提供了一种有效方法。Precipitation is influenced by atmospheric circulation, topography, air pressure and other environmental factors, so accurate prediction of precipitation is difficult. Combining advantages of ARIMA model and RBFNN model, ARIMA-RBFNN combination model is proposed to predict the precipitation of Baicheng city from 2001 to 2010. Compared with the ARIMA model and RBFNN model, the results show that the maximum relative error of ARIMA-RBFNN model for prediction of precipitation is 27.33%, the minimum relative error is 0.70%, and the average relative error is 8.54% ; the prediction accuracy is better than that of ARIMA model and RBFNN model. Thus, combination model play the advantages of ARIMA model and RBFNN model respectively, which provides an effective method for prediction of precipitation.

关 键 词：降水量 ARIMA-RBFNN组合模型 ARIMA模型 RBFNN模型 

分 类 号：S161.61[农业科学—农业气象学]