Abstract:

There are many factors influence the energy use of an ethylene plant. So it is necessary to screen the related factors with correlation analysis to ensure the comprehensiveness and effectiveness for the energy efficiency of an ethylene plant. Data envelopment analysis (DEA) is a non-parametric optimization of efficiency analysis, which would consider the multiple factors for ethylene plant. However, the improper or excessive input and output indicators would lead to the poor resolution of DEA efficiency. It needs to reduce current dimensionality of input and output indicators for energy efficiency analysis of ethylene plant. The PCA-DEA method which based on principal component analysis (PCA) is provided. Proposed PCA-DEA method applies to energy efficiency analysis of an ethylene plant with monthly production data from 2001 to 2010. More reasonable energy efficiency analysis results and more information are obtained compare with the method just selected key indicators and the method without PCA. The applicability and effectiveness of PCA-DEA have been verified by comparison to specific energy efficiency (SEC). The accuracy of energy efficiency analysis for an ethylene plant has been improved.

Key words: energy efficiency analysis, data envelopment analysis, principal component analysis, numerical analysis, dimensionality reduction, production

摘要：

由于影响乙烯装置能耗的因素较多, 有必要采用相关分析筛选与乙烯装置能效相关的因素, 以确保分析的全面性与有效性。数据包络分析(DEA)作为一种非参数的效率优化分析, 适用于多输入多输出系统的特点能够综合考虑乙烯能效评估中的多因素影响。输入输出指标过多或存在不当指标, 则会导致DEA分辨率过低, 需要对输入输出指标降维。为此, 提出采用主元分析(PCA)降维的PCA-DEA法, 并将该PCA-DEA法应用于某乙烯装置以2001～2010年的月实际生产数据为依据的能效分析中, 得到比单纯选用主要指标的DEA算法和未采取PCA降维的DEA算法更好的能效分析结果。通过与单位综合能耗(SEC)对比, 验证了PCA-DEA方法的有效性, 为乙烯装置能效分析的准确性提供了更为实用的方法。

关键词: 能效分析, 数据包络分析, 主元分析, 数值分析, 降维, 生产