Multi-Objective Dynamic Robust Operation Optimization For Naphtha Pyrolysis Process

The purpose of production operation optimization is to provide the optimal operating parameters for the practical production,which is very important for improving factory efficiency and enhancing product quality.There are many uncertain disturbances in the process of naphtha pyrolysis and the traditional static robust operation optimization only sets a fixed degree of robustness according to the historical disturbance data.Although the influence of disturbance can be reduced,this method is based on human experience and consequently cannot guarantee the maximization of production efficiency.For this problem,this dissertation proposes a dynamic robust operation optimization model based on time series prediction and solution method.Firstly,a time series forecasting model based on Least Squares Support Vector Machine(LSSVR)and differential evolution algorithm(DE)is established to realize the real-time prediction of disturbance.Secondly,the ethylene and propylene prediction models are established by comparing the regression effect of several stable ensemble models.Finally,the improved multi-objective differential evolution algorithm(IGDE3)is used to solve the dynamic robust operation optimization model.Specific research includes:(1)Aiming at the actual situation of naphtha pyrolysis,the optimization objective and operation parameters are determined.Then considering the influence of disturbance and the insufficiency of the static robust optimization model,a multi-objective dynamic robust operation optimization model is established based the time series forecasting.(2)Aiming at the dynamic uncertainties in the process of naphtha pyrolysis,a time series forecasting model of dynamic disturbance is established by combining Least Square Support Vector Machine(LSSVR)with improved differential evolution algorithm(DE)which takes parameters of phase space reconstruction and model parameters as decision variables.The time series forecasting model is used to predict the fluctuation range of the parameters of the next optimization period,which can be used to realize the dynamic adjustment of operation optimization robustness.(3)Aiming at the problem that the yield of ethylene and propylene can not be detected on-line in the process of naphtha pyrolysis,some efficient ensemble regression models are established which take LSSVR_DE as the sub-learning machine.Then the high-performance models for predicting the yields of ethylene and propylene are selected.(4)Aiming at dynamic robust operation optimization model of naphtha pyrolysis process,this dissertation proposes the improved GDE3 algorithm whose features mainly consist of the parameter adaptive ability and local search ability.The experimental results based on simulation data show that the multi-objective dynamic robust operation optimization method based on time series prediction is superior to the traditional static robust operation optimization method.