基于神经网络逐级分类建模的北京地区能见度预报

Visibility forecast in Beijing through artificial neural network based on hierarchical classification method

在研究了北京市近10年来城市能见度变化特征及低能见度发生的主要影响因素的基础上,利用北京市气象站观测资料筛选出主要的预报因子,使用学习向量量化算法对样本进行分类,再利用列文夸特算法逐级建立预报模型,并进行预报试验.结果表明神经网络逐级分类模型具有良好的预报能力,其关键在于模型的识别分类能力,试验得到分类准确率达到87.0%,采用列文夸特优化算法的逐级分类模型较统计回归预报残差平方和下降了57.1%;低能见度、中等能见度、高能见度预报准确率分别为60.0%,68.2%,79.3%,均高于统计回归预报方法(12.5%,41.7%,52.4%).

Based on the study of visibility variation and the main causes for low visibility in Beijing during the past 10 years, the appropriate predictors were selected from Beijing observation station in this study. The selected samples were classified through learning vector quantization （LVQ） artifical neural network, and then Levenberg-Marquardt （L-M） optimization scheme was used to establish a prediction model to conduct the forecasting test. The results indicate that the increase of the accuracy of L-M optimization model lies in its ability of identification and classification, with the accuracy rate of model classification being 87.0% in this test. The sum squared resin was decreased by 57.1% through the use of L-M optimization model, in contrast with statistical regression models. The prediction accuracy of low-value visibility, moderate-value visibility and high-value visibility was 60.0%, 68.2% and 79.3% respectively, much higher than that produced by the statistical model （12.5%, 41.7%, 52.4%）.