| 本文已被：浏览 1814次 下载 1537次 |
码上扫一扫！ |
| 湛江降水量的径向基神经网络预测模型 |
|
薛宇峰1, 宋丽红1, 罗泽举2
|
|
1.湛江海洋大学, 广东, 湛江, 524005;2.中山大学, 广东, 广州, 510275
|
|
| 摘要: |
| 以湛江地区50年来的月降水量为时间序列,利用高斯径向基函数,选择输入窗口(时滞)大小为6,建立了一种智能型的径向基函数神经网络预测系统,并分别对1991～2000年和2001～2003年的月降水量进行了测试预报和独立样本预测。结果显示,该模型预测效果明显优于传统的线性自回归预测模型,各月平均的平均绝对误差(MAE)和均方误差(RMSE)达到41.8和55.7。虽然该模型对降水量的预报还存在量级偏小的系统性偏差,但它完全有可能为本地区短期气候预测提供一种客观、自动的业务预报方法。 |
| 关键词: 湛江降水量 径向基神经网络 模型预测系统 |
| DOI： |
| 分类号: |
| 基金项目:国家自然科学基金项目(40205016);云南省自然科学基金项目(2003D0011)联合资助 |
|
| A RBF NEURAL NETWORK FORECASTING MODEL FOR RAINFALL IN ZHANGJIANG |
|
XUE Yu-feng1, SONG Li-hong1, LUO Ze-ju2
|
|
1.Zhanjiang Ocean University, Guangdong 524005, China;2.Sun Yat-Sen University, Guang Zhou 510275, China
|
| Abstract: |
| Taking 1951～2000 monthly rainfall data in the Zhanjiang area as the time series and using the Gaussian radial base function and a delayed input window chosen at 6,a new intelligent forecast system is developed based on radio basic function neural network(RBFNN) to predict monthly rainfall from 1991 to 2003.Results show that the RBFNN is obviously superior to the traditional linear model,and its MAE(mean absolute error) and RMSE(root mean square error) are 41.8 and 55.7,respectively.In view of it,this model may provide forecasters with valuable results in their short-term climate forecasting work. |
| Key words: Zhanjiang rainfall radio basic function neural networks forecasting model |