| 基于BPSO-NBayes的雷暴释用预报技术研究 |
| |
| 引用本文： | 刘亚杰,胡邦辉,王学忠,王举,黄泓.基于BPSO-NBayes的雷暴释用预报技术研究[J].气象科学,2018,38(3):370-377. |
| |
| 作者姓名： | 刘亚杰 胡邦辉 王学忠 王举 黄泓 |
| |
| 作者单位： | 国防科技大学气象海洋学院 |
| |
| 基金项目： | 国家自然科学基金资助项目（41475070；41375049） |
| |
| 摘 要： | 提出了一种新的雷暴预报法,即二进制粒子群-朴素贝叶斯分类器(Binary Particle Swarm Optimization-Naive Bayesian Classifiers,BPSO-NBayes)方法,以福州、连城、宁波3站为例,对使用T511数值预报产品站点的雷暴释用预报技术进行研究。利用2010—2014年T511数值预报产品和单站观测资料,使用BPSO-NBayes方法,建立了0~72 h雷暴预报模型,并与Fisher判别准则和Bayes判别准则进行比较。预报结果表明,BPSO-NBayes模型临界成功指数都在0.29以上,平均值达到0.33以上,是3种方法中最好的,空报率都在0.59以下,漏报率在0.60以下,而且变化幅度很小。BPSO-NBayes模型明显优于Fisher判别准则和Bayes判别准则,具有良好的稳定性和预报能力。
|
| 关 键 词： | T511数值预报产品 粒子群算法 朴素贝叶斯分类器 雷暴预报 |
| 收稿时间： | 2017/2/20 0:00:00 |
| 修稿时间： | 2017/4/24 0:00:00 |
Research on forecasting technology of thunderstorm interpretation based on BPSO-NBayes |
| |
| LIU Yajie,HU Banghui,WANG Xuezhong,WANG Ju and HUANG Hong.Research on forecasting technology of thunderstorm interpretation based on BPSO-NBayes[J].Scientia Meteorologica Sinica,2018,38(3):370-377. |
| |
| Authors: | LIU Yajie HU Banghui WANG Xuezhong WANG Ju and HUANG Hong |
| |
| Institution: | College of Meteorology and Oceanography, National University of Denfense Technology, Nanjing 211101, China,College of Meteorology and Oceanography, National University of Denfense Technology, Nanjing 211101, China,College of Meteorology and Oceanography, National University of Denfense Technology, Nanjing 211101, China,College of Meteorology and Oceanography, National University of Denfense Technology, Nanjing 211101, China and College of Meteorology and Oceanography, National University of Denfense Technology, Nanjing 211101, China |
| |
| Abstract: | In order to study the forecasting technology of thunderstorm interpretation in stations by T511 numerical forecast products, taking Fuzhou, Liancheng and Ningbo as examples, a new thunderstorm forecasting method was proposed, which is Binary Particle Swarm Optimization-Naïve Bayesian Classifiers. For the purpose of establishing the prediction model of thunderstorm within 72 hours, compared with Fisher discriminant criterion and Bayes discriminant criterion,T511 numerical forecast products and the corresponding observational station data from 2010 to 2014 were used, and Binary Particle Swarm Optimization-Naïve Bayesian Classifiers method was adopted. Results show that each critical success index of thunderstorm forecast model established by BPSO-NBayes is above 0.29 and the average is greater than 0.33. It is the best result of three methods. The false alarm rate of each station is below 0.59. The probability of false omission is under 0.60. And its magnitude of variety is very little. The BPSO-NBayes model is not only predictive but stable, superior to Fisher discriminant criterion and Bayes discriminant criterion in thunderstorm numerical forecast. |
| |
| Keywords: | T511 numerical forecast products Particle swarm optimization Naïve Bayesian Classifiers Thunderstorm forecasting |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《气象科学》浏览原始摘要信息 |
| 点击此处可从《气象科学》下载免费的PDF全文 |
|
