The Global Optimization Of Heat Exchanger Networks

With the advance of industrialization, energy reserves are growing short. Heat ex-changer networks, as an important part of process industries, have a positive impact onthe rational use of energy throughout the system, and are significant for reducing energyconsumption, and reducing product cost in the course of production. At present the localoptimization methods and stochastic optimization methods are widely used for heat ex-changer network optimization problem. Although it is easy to get optimal results bystochastic optimization methods which don’t rely on the objective function, the methodsare established based on statistical probability theory and simulation of some naturalphenomena or processes so that the heat exchanger network structure cannot be provedto be the best solution. Local optimization methods develop relatively mature, haveformed a series of classical theory, and solve the problem for a single peak function toget the optimal solution, but the methods are very dependent on the initial value strongwhen used in optimization of heat exchanger networks, and cannot obtain the global op-timal solution with the limitations of the optimization method itself. So, the determinis-tic methods of global optimization are studied for the heat exchanger network to jumpout of local optimal solution in the process of searching variables, looking for betterheat exchanger network structure and result until the global optimal solution is found inthis paper.Firstly, with the comprehensive cost as the objective function, a mathematicalmodel of heat exchanger network optimization was established. The optimization resultsof benchmark example were used to reveal the objective function of heat exchangernetwork optimization had multi-peak characteristics, and the nonlinear characteristicscaused by the integer variables and continuous variable were analyzed.Secondly, deterministic optimization methods were applied to optimization of heatexchanger networks. Filling and tunneling function algorithm of heat exchanger net-works optimization was proposed, with the appropriate filling and tunneling functionconstructed, to jump out of local minimum of objective function of heat exchanger net-works optimization and its feasibility was checked by an example. Peak-Valley Super-seding Approach was first proposed for heat exchanger networks optimization, the al-gorithm was established to jump out of the trap of local optimum of heat exchangernetworks. Then, based on idea of jumping out of local minimum of global optimization me-thod, the interpolation and extrapolation approach was proposed to jump out of the localminimum point of heat exchanger networks. The single variable one-dimensional vectorsearching, multi-variable one-dimensional vector searching and multi-dimensional po-lyhedron searching algorithm were established for heat exchanger network optimization,and their optimization results were verified and compared by a case.Furthermore, the strategy of large-scale heat exchanger networks optimization wasproposed as to its feature. Hybrid optimization algorithms of heat exchanger networkswere presented combined with advantages of random methods and deterministic me-thods. The size of heat exchanger networks’ structure was reduced with application ofstochastic methods, and then, the areas of heat exchangers were optimized by determi-nistic method.In a word, the heat exchanger network deterministic methods were deeply studiedbased on the idea of global optimization deterministic method. Several optimizationmethods jumping out of local minimum of heat exchanger networks were presented, andthey have an important theoretical and practical significance at global optimization ofheat exchanger networks.