| 基于随机森林算法的WANO机组能力因子分类 |
| |
| 引用本文： | 朱 伟,侯秦脉,吴彦农,张泽宇,王娅琦,胡 江.基于随机森林算法的WANO机组能力因子分类[J].科技管理研究,2022(21). |
| |
| 作者姓名： | 朱 伟 侯秦脉 吴彦农 张泽宇 王娅琦 胡 江 |
| |
| 作者单位： | 生态环境部核与辐射安全中心,生态环境部核与辐射安全中心,生态环境部核与辐射安全中心,生态环境部核与辐射安全中心,中核战略规划研究总院有限公司,建信金融租赁有限公司 |
| |
| 摘 要： | 针对核电厂机组能力因子分类研究缺少相对简单有效的方法，基于第一至第八次《中华人民共和国核安全公约国家报告》中世界核电营运者协会（WANO）性能指标的数据，提出一种随机森林模型（random forest，RF）的机组能力因子分类方法，通过估算随机森林模型决策树的棵树、内部节点再划分所需要的最小样本数等，构建了最优的随机森林分类模型，成功实现对能力因子的快速和精细分类，为第九次国家报告中定性掌握我国核电机组发电状况及行业内机组所处状况有及其重要的意义。同时，选用解决二分类的Logistic回归作对比试验，试验结果表明RF分类算法的总体精度达到77.27%，Kappa系数为0.705 3达到高度一致性检验标准区间，明显高于Logistic回归的51.14%和0.110 1，RF表现出分类效果好、准确率高和性能稳定等优点，能够有效提高机组能力因子分类的准确度。
|
| 关 键 词： | 随机森林 Logistic回归 机组能力因子 WANO |
| 收稿时间： | 2022/5/18 0:00:00 |
| 修稿时间： | 2022/7/8 0:00:00 |
WANO Unit Capability Factor Classification Based on Random Forest Algorithm |
| |
| Abstract: | In view of the lack of a relatively simple and effective method for the classification of nuclear power plant unit capacity factor, based on the data of the performance indicators of the World Association of Nuclear Operators (WANO) in the first to eighth National Reports to the Convention on Nuclear Safety of the People''s Republic of China, put a unit capacity factor classification method of the random forest model (random forest, RF), constructs the optimal random forest classification model by estimating the tree of the decision tree of the random forest model and the minimum number of samples required for internal node subdivision. The rapid and fine classification of the capability factor is of great significance for qualitatively grasping the power generation status of my country''s nuclear power plants and the status of the units in the industry in the ninth national report. At the same time, Logistic regression which solves two classifications is selected as a comparative test. The test results show that the overall accuracy of the RF classification algorithm reaches 77.27%, and the Kappa coefficient is 0.705 3, which reaches the standard range of high consistency test, which is significantly higher than the 51.14% and 0.110 1 of Logistic regression. It shows the advantages of good classification effect, high accuracy and stable performance, which can effectively improve the classification accuracy of unit capacity factors. |
| |
| Keywords: | |
|
| 点击此处可从《科技管理研究》浏览原始摘要信息 |
| 点击此处可从《科技管理研究》下载免费的PDF全文 |
|
