摘要
提出了一种面向热工过程海量运行数据的高质量样本提取方法,通过主成分分析(PCA)提取系统隐变量,采用基于稳态权重的合成少数类过采样(SWSMOTE)来补充少数类工况样本。以燃气轮机为工程算例,验证所提算法的有效性。结果表明:提出的高质量样本提取方法可将原始数据数量压缩到10%左右,模型平均均方根误差从0.042下降至0.031,模型训练时间减少90%。
A method of high-quality sample selection was proposed for mass operating data of thermal processes. The latent variables of the system were selected through principal component analysis(PCA), and the samples were supplied by the steady weights synthetic minority over sampling technique(SWSMOTE) for the operating condition with fewer samples. After which, taking a gas turbine as an engineering example, the availability of the proposed method was verified. Results show that, the original data can be compressed to about 10% by the proposed method for high-quality sample selection, and the average root mean square error of the model can be reduced from 0.042 to 0.031. The training time of the model can be reduced by 90%.
作者
何康
汪勇
陈荣泽
任少君
司风琪
He Kang;Wang Yong;Chen Rongze;Ren Shaojun;Si Fengqi(School of Energy and Environment,Southeast University,Nanjing 210096,China;Shanghai Power Equipment Research Institute Co.Ltd.,Shanghai 200240,China)
出处
《发电设备》
2023年第1期59-64,共6页
Power Equipment
基金
国家自然科学基金资助项目(51976031)
国家电力投资集团有限公司统筹研发经费支持项目(TC2019HD10)
上海发电设备成套设计研究院有限责任公司科技发展基金(201909009C)。
关键词
热工过程
样本提取
数据约简
thermal process
sample selection
data reduction
