Research Of Ethylene Energy Efficiency Evaluation Method Based On ELFLN And Multi-stage DEA

Energy efficiency has always been a major concern of the high-energy-consuming manufacturing.With the development of China's petrochemical industry,improving energy efficiency is of great significance to further implement the policy of energy conservation and emission reduction,to ensure national energy security,and to achieve economic and environmental benefits.Energy efficiency evaluation is an important link to improve energy efficiency.Through data driven modeling and index analysis of production factors,the energy saving potential under inherent production conditions is revealed and the direction of energy efficiency improvement is pointed out,thus the purpose of guiding actual production is achieved.If the energy efficiency model is not reasonable,the evaluation method is not comprehensive,or the result is inadequate,the practical value of the energy efficiency evaluation method will be reduced seriously.The complex petrochemical process data is highly nonlinear data containing uncertainty and noise.The traditional energy efficiency evaluation method cannot overcome the uncertainty of actual production data,and cannot make pre-evaluation on the planned production data(with input and without output),so is restricted by many conditions and fails to be widely used.In addition,previous energy efficiency evaluation methods only consider input-output variables,ignoring the influence of exogenous environmental variables on energy efficiency calculation.In order to solve these problems,this paper proposes a novel ethylene energy efficiency evaluation model based on extreme learning fuzzy logic network(ELFLN)and multi-stage data envelopment analysis(DEA).In order to solve the problem of data's uncertainty and pre-evaluation in the complex petrochemical process,fuzzy logic and neural network algorithm are combined to construct a network whose original hidden layer is replaced by the fuzzy inference system.Then the influence of exogenous environmental variables and random factors is excluded by the combination of DEA and stochastic frontier analysis(SFA).The case study shows that the comprehensive evaluation model based on ELFLN and multi-stage DEA is better than the traditional method in efficiency identification.The accuracy of neural network modeling and energy efficiency classification is also enhanced due to the introduction of fuzzy logic and extreme learning algorithm.