Desenvolvimento de técnicas de otimização da programação das limpezas de redes de trocadores de calor

Abstract

Heat exchangers are devices widely used in many processes with the goal of changing the temperature and/or the physical state of the streams. A heat exchanger network (HEN) can be defined as a group of interconnected heat exchangers in order to reduce the energy requirements of a system and is widely used in process industries. However, during the operation of a HEN, the thermal effectiveness of the heat exchangers decreases due to fouling. This effect brings an increase of the fuel costs and carbon emissions. A possible approach to mitigate this problem is based on the establishment of a heat exchanger cleaning schedule during the refinery operation. This kind of intervention brings a tradeoff: the cleaning of a heat exchanger can restore its thermal effectiveness, but implies in additional costs, such as, manpower, chemicals, etc. Additionally, during the cleaning, the heat exchanger must be bypassed which temporally increases the energy consumption. In this context, the present work aims to explore different optimization techniques involving stochastic and heuristic methods. For this purpose, it was developed a set of integrated computational codes involving a pseudo-stationary simulation of HEN behavior related to fouling and schedule optimization of cleaning this type of system. The solution of the problem indicates for the operating staff the time periods for the cleaning of each heat exchanger in the crude preheat train. In the stochastic approach used, the parameters of the genetic algorithm, as the probability of crossover and mutation, have been calibrated to the present problem. The developed heuristic approach was made through the sequence of motions set of zero, one and two. Alternatively, the methodology developed in which the recursive heuristic move sets one and two were used recursively. The hybrid approach consisting of different combinations of stochastic and heuristic methodology was also developed. A comparative analysis of methodologies aimed at assessing the most appropriate approach to this problem of scheduling the cleaning in terms of computational effort and objective function. The performance of the proposed approach was explored by a series of examples including a real Brazilian refinery. The results were promising, suggesting that the optimization techniques of this work may be interesting approaches to operations involving HENs. Among the analyzed optimization approaches, the heuristic methodology developed in this work showed the best results showing competitive front approaches compared to the literature