Simultaneous Synthesis Of Large-scale Heat Exchanger Networks Based On Uniformity Principlie

As an important field in the process systems engineering,the heat exchanger networks synthesis(HENS)plays an important role in the reduction of energy consumption.Two primary approaches have been proposed for HENS over the last 50 years,namely,thermodynamics and mathematical programming approaches.The pinch method,as a typical thermodynamics method,has a wide range of applications in industrial fields due to its clear concept and sample operation.However,the sequential nature of this method limits its ability that accurately trade-off between utility and exchanger costs,leading to suboptimal network designs.Mathematical programming methods based on MINLP model,which are suitable for simultaneous synthesis of heat exchanger network problems,can be broadly divided into deterministic and stochastic methods.Because of rigorous logic,deterministic methods can obtain the optimal solution theoretically.Nevertheless,for large-scale HENS problems,these methods easily obtain a sub-optimal network by converging to local optima.Compared with deterministic methods,stochastic methods,which do not need the derivative information of the objective function,can handle well the large-scale continuous or discrete optimization problems.However,due to the randomness,these algorithms often need a lot of computation time and constraints to ensure that the logic is correct.Thus,how to get a optimal solution for the industrial-scale heat exchanger network has become a research focus of researchers.Thus,a simultaneous synthesis approach was proposed for large-scale heat exchanger networks and its performance was demonstrated with a case study found in the literature.The main content of this paper was presented as follows.Firstly,Lagrange multiplier method was selected to handle the constraints of HENS after analyzing the drawbacks of penalty function method.The new Lagrange function equations were generated by connecting constraints and the original objective function.In order to solve the Lagrange function equations,the steepest-descent method and the Powell method solving strategy according to the deterministic approaches were proposed.Meanwhile,a structure evolution strategy,which can effectively reduce the search space and get the network with superior performance easily,was proposed to reach the aim of global optimization.Secondly,a novel Powell particle swarm optimization(PPSO)algorithm,which has both high precision of the deterministic methods and high efficiency of the stochastic algorithms was presented.For overcoming disadvantages of stochastic algorithms,the cloud memory and the opposite strategy of individuals were proposed,which can avoid premature convergence and expand the search space.In addition,two optimizing strategies of integer variables were combined with PPSO algorithm in order to simultaneously optimize continuous and integer variables.Thirdly,a novel uniformity factor of temperature difference and a uniformity factor of heat capacity flowrates were presented based on the single heat exchanger after analyzing the uniformity of temperature difference in heat exchanger networks.In order to demonstrate the accuracy of two uniformity factors,an improved approach based on the structure evolution strategy for heat exchanger network was established.The results show that the uniformity factor can accurately reflect the performance of network,the improved network has a lower value of total annual cost and the heat load has been redistributed.Finally,a simultaneous synthesis method for large-scale heat exchanger network on the basis of Lagrange multiplier method,PPSO algorithm and two uniformity factors is established.The complexity of large-scale HENS problems would be significantly reduced if the process streams can be grouped accurately.Given this,uniformity factors were applied to HENS.For a given process stream grouping,the PPSO algorithm was adopted to obtain an optimal sub-network.The effectiveness of the presented method was exemplified by a large case study involving 22 hot streams and 17 cold streams.The obtained solution is better than that published in the literature,which shows that the presented algorithm can find better designs in a reasonable time range.