Intelligent Optimization Of High Energy Consumption Processes

Operation optimization of high energy consumption production process in continuous industry is an important technical means to achieve energy saving and emission reduction,as well as quality rising and efficiency improvement.However,the complexity of continuous industry creates many difficulties in operation optimization of high energy consumption production process.For exam-ple,the complicated physical and chemical reactions,the large lag from distribution to tapping,the lack of detection means and the strong coupling of process index parameters in blast furnace,which seriously restrict the modeling and prediction of key process quality index and process operation op-timization;In the high energy consumption process of papermaking and pulping,the multi-source heterogeneous data problem caused by multi-stage pulping creates difficulties for in-time operation optimization;In addition,there are some problems in the continuous industry,such as information island,decentralized management,difficult to solidify and share operation knowledge.It is ur-gent to build an industrial cloud platform to realize the software component and sharing reuse of operation optimization knowledge in high energy consumption production processes.In view of the above problems,this thesis focuses on the modeling and prediction of key pro-cess quality indicators and operation optimization methods of large-scale blast furnace ironmaking process,the operation optimization of paper pulping process with multi-source heterogeneous data,and the construction of industrial cloud platform to deploy and apply these high energy consump-tion process operation optimization methods in the form of micro service.Exploring a new mode of knowledge sharing for operation optimization in continuous processes.The detailed researches and achievements are introduced in the following four parts:1)Addressing the difficulty in modeling the industrial process,a fuzzy neural network with sliding window is proposed to predict the molten iron quality in the blast furnace ironmaking process.First,the mutual information method is used to select the operating variables that have the clos-est relation with the quality indexes.Then,the T-S fuzzy neural network with sliding window is utilized to learn the dynamics of the iron quality in the production process.Compared with traditional mechanism-based modeling methods,this algorithm does not have to consider the complex physical transformations and chemical reactions inside the blast furnace,and can ef-fectively deal with the prediction problem of iron quality indexes in the ironmaking process under different working conditions.2)Addressing the difficulty in optimizing the industrial process,a multi-objective optimization control method consisting of evolutionary algorithm and deep learning is proposed to optimize the ironmaking process.Firstly,the mapping relationship between input and output arguments is described by recurrent neural network to establish a data-driven model for ironmaking process.Secondly,under the constraints of production conditions,quality conservation and operation conditions,a multi-objective optimization algorithm based on genetic algorithm is proposed to optimize process indicators and their corresponding operating parameters.By regarding the proposed deep learning model as the fitness function of genetic algorithm,the construction of an integrated model for modeling and optimizing is realized.3)Addressing the multi-sampling problem caused by heterogeneous data sources,an energy opti-mization model based on an intelligent optimization framework is proposed.First,a mixed data sampling regression model is established for the refining system,in which the low sampling rate data is predicted by high sampling one.Then,the optimal objective function value and its corresponding input parameters are found by genetic algorithm based on the established mixed data sampling regression model.At last,the energy consumption of high consistency refiner in the refining system is optimized without sacrificing the pulping quality or yield.4)Addressing the difficulties of algorithmic applications,an optimal scheduling method based on industrial internet is proposed to optimize the industrial processes.At first,a cloud computing cluster is constructed by four super servers.Then,Rancher and Harbor softwares are imple-mented in conjunction to provide cloud services.After that,the cloud system is constructed based on industrial internet platform.Finally,the optimal operation algorithm is packaged as a service in Docker.With the parallel running in industrial internet platform,the operation level of practical production plant has been significantly improved.In the end,the academic contents and experimental results of this thesis are summarized,and the future research direction of intelligent optimization for high energy consumption processes is discussed.