关键词:
Reactor networks synthesis
Optimization
Imperialist competition algorithm
Quasi linear programming
Continuous variables
Structural variables
摘要:
The Reactor network (RN) synthesis of optimization problems leads to mixed-integer non-linear programming (MINLP) models that are very difficult to solve, especially for non-isothermal cases. In this research, a new and simple robust approach is presented for RN synthesis. To reduce the complexity of the model, instead of solving discrete and continuous variables simultaneously, a combination of a stochastic algorithm (i.e. imperialist competition algorithm (ICA)) and mathematical methods (i.e. modified Quasi Linear Programming (QLP)) has been used. The ICA is used to produce structural configuration, whereas continuous variables are handled using a modified QLP formulation. Each structure of the RN generated by ICA includes continuous stirred tank reactor (CSTR), plug flow reactor (PFR), PFR with recycle stream, and differential side-stream reactor (DSR) which are addressed by generating numerical codes in a vector. The RNs structures are then sent to the modified QLP to compute the amount of overall objective function (OOF). In the modified QLP model, first, the continuous variables are divided into two groups I and II, and the optimal values of these variables are obtained at internal and external levels respectively, taking into account the OOF. With this method, the complex MINLP model of RN synthesis is converted to an IAC + modified QLP which is much easier to solve. The cases studied showed that this method can reach better solutions compared to the literature. ? 2021 Elsevier Ltd. All rights reserved.