Leandro V. Pavão

Automated heat exchanger network synthesis by using hybrid natural algorithms and parallel processing

Abstract Heat exchanger network (HEN) synthesis can be formulated as an optimization problem, which can be solved by meta-heuristics. These approaches account for a large computational time until convergence. In the present paper the potentialities of applying parallel processing techniques to a non-deterministic approach based on a hybridization between Genetic Algorithms (GA) and Particle Swarm Optimization (PSO) were investigated. Six literature examples were used as benchmarks for the solutions obtained. Comparative experiments were carried out to investigate the time efficiency of the method while implemented using series or parallel processing. The solutions obtained led to lower Total Annual Costs (TAC) than those presented by the literature. As expected, parallel processing usage multiplied the algorithm speed by the number of cores used. Hence, it can be concluded that the proposed method is capable of finding excellent local optimal solutions, and the application of multiprocessing techniques represented a substantial reduction in execution time.

Heat Exchanger Network Synthesis using a Genetic Algorithms-Particle Swarm Optimization Hybrid Method and Parallel Processing

Abstract In Heat Exchanger Network (HEN) synthesis capital savings and pollutant emission reduction can be achieved. The mathematical modeling of the HEN synthesis problem requires elaborated solution strategies given the particularities of their non-linear formulations and non-convex problems. The use of heuristic approach accounts for a large computational load, and hence a high processing time until convergence. In the present paper a hybrid model for HEN synthesis using Genetic Algorithms (GA) and Particle Swarm Optimization (PSO) is presented. The potentialities of applying parallel processing techniques to solve the problem were studied. Two examples from the literature were used as benchmarks for the solutions obtained. Comparative experiments were carried out to investigate the time efficiency of the method while implemented using series or parallel processing. The solutions obtained in both cases with the proposed methodology led to Total Annual Costs (TAC) equal or lower to those presented by the literature. As one could expect, parallel processing usage multiplied the algorithm speed by the number of cores used (processing time was close to 75% lower by using 4 processing cores). Hence, it can be concluded that the hybrid algorithm proposed has potential to find near-optimal solutions, and the application of multiprocessing techniques to such non-deterministic approaches represents a substantial reduction in the execution time.

Heat exchanger network synthesis using genetic algorithm and differential evolution

Abstract Heat Exchanger Networks (HEN) synthesis is a process engineering problem that can be mathematically characterized as highly combinatory, non-linear and non-convex. All these aspects bottleneck the identification of locally optimal solutions at acceptable computational time. This work proposes an optimization algorithm based on a superstructure considering non-isothermal mixing and stream splitting. HENs are optimized through the application of a bi-level new hybrid method that works at an upper level with Genetic Algorithm (GA) to optimize discrete variables and at a lower level with Differential Evolution (DE) for optimizing heat loads and stream split fractions in order to find solutions with low total annual costs (TAC). The proposed method was applied to six literature case studies and was efficient in obtaining solutions with TAC comparable or lower than those previously reported.

A meta-heuristic approach for financial risks management in heat exchanger networks

Abstract Mathematical programming models for heat exchanger networks (HEN) synthesis generally do not take into account the fluctuations in the costs of commodities related to the plant operation. This work proposes to include such features in the HEN synthesis model by assuming utility production costs as uncertain and considering them as stochastic variables. These are discretized via Monte Carlo Simulation (MCS) with the Geometric Brownian Motion (GBM) model, which creates an appropriate number of scenarios throughout plant lifetime based on natural gas and electricity historical price data. The downside risk, which is a target based metric, is used for financial risks management. Multi-objective optimization (MOO) can be performed to better address trade-offs between solutions’ expected total annual costs (ETAC) and the chosen metric. A meta-heuristic two-level approach is adapted to handle such MOO. The developed scheme is based on the e-constraint method. Two costs targets are tested with downside risk. The number of scenarios with costs higher than the targets is reduced when compared with single ETAC optimization. The developed meta-heuristic solution method was able to efficiently perform MOO in a model with rather large number of scenarios, and is a good option for applying risk metrics including uncertainties in HEN synthesis.