Research On Structure Diversity Evaluation And Improvement Strategies For Heat Exchanger Network Optimization With Stream Splits

Heat exchanger network(HEN)is a heat recovery technique widely applied in the industrial field,its optimization can further reduce external energy consumption and improve production economy.However,in essence,the simultaneous optimization of HEN belongs to mixed integer non-linear problem,which is characterized as strongly non-linear and non-convex and tends to cause the local optimum prematurely,so it has become a common challenge for scholars to seek more effective solutions.In recent years,a new heuristic method named random walk algorithm with compulsive evolution(RWCE)is proposed and applied for HEN optimization,possessing the advantages of less control parameters,more simple operation and stronger global search ability compared to the traditional heuristic methods.However,the application of RWCE for HEN with stream splits is still blank,and there exists an inherent cluster phenomenon of heuristic methods in RWCE,which has a significant impact on the algorithm performance especially on the structure evolution ability and its mechanism needs to be discussed urgently.Moreover,RWCE also ignores the protection and development of optimums,making the optimization quality have more space for improvement.Therefore,based on the model with stream splits,this paper starts with the structure cluster phenomenon of heuristic methods and is devoted to studying the evolution mechanism of RWCE and its related modified strategies,aiming at the enhancement of algorithm performance and the exploration of optimal network structure.The detailed contents are given as follows:Firstly,the HEN stage wised superstructure(SWS)model considering stream splits and non-isothermal mixing assumption is established.To overcome the deficiency that SWS cannot flexibly place utilities and then restrict the flexible development of structures,a modified SWS including the inner utility strategy(IU-SWS)is proposed to enlarge the solution space.On the basis of these two models,the RWCE is used and the efficient optimization methods of heat exchange loads,inner utilities and split ratios are also created.The case results possessing stream splits are competitive compared with that in literature,proving the validity of the algorithm and modified model.Secondly,an evaluation method of structure diversity is established,the evaluating indicators are obtained by structure standardization,public structure identification and structure group identification and then applied to describe the structure cluster phenomenon.The cluster degree of RWCE,differential evolution algorithm and particle swarm optimization algorithm and their effect on the optimization performance are emphatically analyzed.Based on which,the evolution mechanism of RWCE is analyzed,revealing the effect of cluster and the positive effect of accepting imperfect solutions.Moreover,an effective evaluation factor of structure diversity is designed by synthesizing the evaluation indicators,which can describe the cluster degree more intuitively.Thirdly,considering that the excellent solutions with potentialities maybe substituted in the operation of accepting imperfect solutions and then cause the insufficient local search,an optimum-protection(OP)strategy is proposed.Two populations respectively named as basic population(BP)and protective population(PP)are set.BP executes the evolution process as RWCE,where plenty of excellent solutions including optimums and pseudo optimums are generated and delivered to the PP from being replaced by imperfect solutions.A dimensionality-reduction random walk technique is designed in PP to exploit the protected solutions fully,enhancing the local search performance.The new algorithm combining the OP strategy and RWCE are more efficient,the results better than that in literature are obtained based on the SWS and IU-SWS,improving the network economy.Finally,a more effective solver for HEN optimization with stream splits is given an in-depth study based on the structure diversity evaluation and the OP strategy.Two structure adjustment strategies are proposed.Due to that cluster could restrict global search,a dispersal search strategy is proposed to decrease the cluster degree by changing structures with perturbation.Due to that cluster has a positive effect on local search,a concentration search strategy is proposed to enhance the cluster degree of better structures by compulsively making the poor individuals accept better structures.The two strategies improve the exploration and exploitation abilities of RWCE in structure space.Then,a sequential scheme is designed.In step one,the structure adjustment strategies are applied for the BP to generate a candidate solution with better performance.In step two,the candidate solution is performed a fine structure adjustment to achieve the further improvement of the structure precision.The results are significantly superior to that in the existing literature,proving that the proposed scheme has obvious advantages in dealing with the HEN problems with stream splits.