摘要
短期精确降水预报一直是天气预报的难点,如何提高预报准确率也是一直被关注的热点。提出了一种新的基于加权最近邻算法,利用某地区降雨量资料和NCEP天气资料,将降雨量作为类,将NCEP天气资料的各种因子场都作为分类因子,计算出不同天气样本间分类因子的相似离度,利用分类因子与类的皮尔逊矩阵相关系数来确定分类因子的权重,通过因子场的逐步引入实现最优分类,最终确定分类因子的数目及其权重来建立最优预报方程,即预报模型。实验中用改进模型对南京市7、8月份进行了24小时降雨预报,实验结果表明,改进模型具有较好的预报效果。
Short -time precipitation forecast has been a difficult problem of weather forecast, so how to improve the forecast accuracy is a critical issue. In this paper, a novel weighted nearest neighbor algorithm was proposed for making similarity forecast. Using the precipitation data in a region as the class and the factor fields of NCEP weather data as classification factors, we firstly calculated the analogue deviation among different classification factors and the Pearson product - moment correlation coefficient between the classification factors and the classes to determine the weight of classification factors. Then we recommended the classification factors step by step to achieve the optional classification, the number and the weight of the classification factors. Finally, an optimal equation, namely forecasting model which aims at making similarity forecast was established based on the number and the weight of the classification. Using this forecasting model, the experimental results of Nanjing City＇s 24h＇s precipitation forecast in July and August show that the model is effective.
出处
《计算机仿真》
CSCD
北大核心
2014年第6期325-328,共4页
Computer Simulation
关键词
加权最近邻算法
相似预报
降水量
Weighted nearest neighbor algorithm
Similarity forecast
Precipitation
