Analysis on provincial industrial energy efficiency and its influencing factors in China based on DEA-RS-FANN. (1st January 2018)

• Record Type: Journal Article
• Authors: He, Yong; Liao, Nuo; Zhou, Ya
• Is Part Of: Energy. Volume 142(2018)
• Page Start: 79
• Page End: 89
• Publication Date: 2018-01-01
• Subjects: Industrial energy efficiency -- Influencing factors -- DEA-RS-FANN method; Power resources -- Periodicals; Power (Mechanics) -- Periodicals; Energy consumption -- Periodicals; 333.7905
• DOI: 10.1016/j.energy.2017.10.011
• Languages: English
• ISSNs: 0360-5442

Abstract

Abstract: Data envelopment analysis (DEA), rough set theory (RS) and fuzzy artificial neural network (FANN) are combined as DEA-RS-FANN procedure to explore the effects of influencing factors on energy efficiency in China's provincial industry sectors. The analysis begins with the DEA technique to evaluate energy efficiency in provincial industries, followed by fuzzy c-means (FCM) algorithm to classify energy efficiency and the influencing factors to three categories (low-, medium- and high-levels). This process facilitates the construction of the decision table from condition attribute (the influencing factors) to decision attribute (energy efficiency). Then significance analysis of attributes in RS theory is adopted to investigate the significance of the influencing factors and determine the primary factors. Finally, FANN is utilized to further analyze the marginal effect of primary factors on energy efficiency in three specific categories, comprising of those provinces with different levels of energy efficiency. The proposed method takes into consideration non-linear and lag effects between energy efficiency and the influencing factors, as well as the characteristics of the impreciseness and incompleteness of the statistical data, ultimately leading to more precise and reliable results, as compared to conventional methods. Highlights: The effects of the factors influencing energy efficiency are analysed by DEA-RS-FANN. The significance of the factors is investigated and primary factors are recognized. The impact of the four primary factors shows difference in three clusters of regions.