Research And Application Of Operation Optimization Methods For Complex Chemical Processes

Safety and quality have always been two core issues of concern in the industry.The chemical industry is a vital component of the national economy.Chemical products act on all aspects of human economic and social life.Due to the variety and quantity of dangerous chemicals associated with the chemical production process,the production process and operating conditions are extremely demanding,coupled with the increasing size,continuity and automation of production equipment,the consequences will be extremely serious if an accident occurs.With the intensification of global competition,higher requirements have been placed on the production quality of enterprises.Industrial process system data,especially statistical data in the chemical industry,has the complexity of high dimensions,strong nonlinearity,noise,loss,and uncertainty.It is difficult to accurately evaluate the performance of process systems based on production data to find the best production configuration.The measurement of production quality is reflected in various aspects of the production process,such as process capacity,control performance,yield gain,and energy efficiency,which focus on different aspects of production quality.Due to the increasingly complex and severely non-linear characteristics of industrial production process systems,the optimization of complex chemical process operations has always been an important research and application area for control science and engineering.Aiming at the operation expectation of "safe,stable,long,full,and excellent" of chemical process,this paper studies the optimization methods of complex chemical process operation,based on data-driven methods such as complex network theory,fuzzy theory,analytic hierarchy process(AHP),process capability analysis,data envelope analysis,multi-objective optimization,focusing on data processing,fault diagnosis,performance evaluation,operation optimization,etc.,realizes the optimized operation of different aspects of the entire process system from a multi-dimensional perspective,overcomes the limitations of a single method and specific objects,and effectively guarantees the production process safety and stability,optimized production processes and operating conditions,and improved product quality and economic benefits.The research methods in this paper are mainly tested in the TE process and ethylene cracking furnace.They are also useful for other chemical equipment and other process industry objects.The main results and innovations in this paper are as below.(1)Aiming at the problem of complex chemical system structure and chemical data characteristics,a process safety analysis method based on horizontal visibility graph analysis is proposed.The individual variables contained in the process system are treated as a set of time series data respectively,and each time is calculated by a horizontal visibility algorithm.The sequence is modeled as a network,and each single-layer network corresponding to the time series is abstracted as a node.At the same time,the correlation between the corresponding nodes is quantified by the correlation function between the two single-layer networks.Then,a complex network structure representing the entire chemical system can be constructed based on the correlation.In addition,the matrix composed of the ratio of mutual information obtained in the fault state and the normal state is calculated,and the abnormal node is located by the variance of each node.(2)Aiming at the problems of uncertainty and sensitivity in the analysis of production conditions of complex chemical processes,a calculation method of fuzzy process capability based on kernel functions and fuzzy process performance indicators is proposed.The dynamic specification limit is determined by cubic spline interpolation.Effectively perform real-time process functions and analyze process performance of industrial data.Applying the integrated method of interpretative structural modeling and fuzzy AHP,based on experiments and practical experience,a method for evaluating fuzzy process capability and fuzzy process performance in the process industry is proposed,which can provide a relatively objective evaluation of the production process capability and process performance in continuous processes.(3)Aiming at the unreasonable distribution of input and output element weights in multi-criteria decision analysis,a performance analysis and optimal temperature selection method of ethylene cracking furnace based on data envelopment analysis cross-model(DEACM)-AHP was proposed.DEACM avoids the unreasonable distribution of input and output element weights,and can accurately identify and estimate the performance status and optimal temperature of ethylene cracking furnaces.At the same time,the improved AHP based on entropy weight can comprehensively consider the reasonable weight distribution of various input-output indicators,and obtain a consistent fusion result of ethylene cracking furnaces at different temperatures.Then again based on DEACM to obtain the best production benchmarks at different temperatures.(4)Aiming at the problem of multi-objective conflicts and operation constraints in the operation of complex chemical processes,a comprehensive evaluation method of multi-objective operation optimization strategy and solution set is proposed,which effectively solved the multi-objective operation optimization problem of ethylene cracking furnace.Based on the dynamic AHP adaptive multi-objective particle swarm optimization algorithm,a fuzzy consistent matrix is used to select the global optimal solution to ensure the correct direction of particle evolution,and the evolution state is measured to adaptively adjust the weights and learning factors.The proposed method is used to optimize the operation of an ethylene cracking furnace,and studies both fixed cracking cycles and non-fixed cracking cycles.According to the actual situation and fuzzy evaluation results,decision makers can choose suitable operating optimization conditions from the Pareto optimal solution set,providing a practical solution for the multi-objective operation optimization of the ethylene cracking furnace.(5)Aiming at the problem of accuracy and rationality of production data feature description of complex chemical processes,a data analysis method based on input-output model and graphlet-based analysis is proposed to realize the optimal allocation of input-output resources of production equipment.The maximum structural network is constructed by considering the mechanism modeling of input-output relationship and constraints,and then the solution structure network is obtained through the direct consumption coefficient matrix of specific production data.The proposed I-O graphlets introduce process information,and different nodes represent different input-output factors in the I-O graphlets.Feature description and analysis based on graphlets can be achieved through network and graph feature statistics.Through clustering analysis of the actual production data of ethylene industry,and using high-efficiency production equipment as the benchmark,the optimal configuration of input and output elements of low-efficiency production equipment can be achieved to improve product yield and energy efficiency.