Optimization of mixed refrigerant systems in low temperature applications by means of group method of data handling (GMDH)

• Presenting two cascade refrigeration cycles which are able to be replaced by olefin plant of the Tabriz Petrochemical Complex.
• Presenting a multi hybrid model i.e. GMDH-type neural network along with linking between Aspen HYSYS and MATLAB software, optimized with genetic algorithm (GA), for predicting optimal consumed power.
• Comparison of NLP techniques with proposed model for finding the values of optimizing variables.

Over the past decades, increasing attention has been paid to optimal design and operation of energy intensive industries. In this paper, a multi-hybrid model with high estimation capability has been applied for prediction of optimum consumed power. Consumed power is the most important factor for cascade refrigeration systems in which efficient estimation of this factor in various operating conditions is essential. The purpose of this paper is to present a new multi-hybrid Model in which six input variables consist of methane, ethane, propane, and nitrogen components composition along with suction and discharge pressures have been employed in order to estimate and predict the optimal consumed power. Having replaced by pure ethylene cycle in the olefin plant of the Tabriz Petrochemical Complex, the one and two stage-cascade cycles are modeled continuously by the proposed model. A hybrid group method of data handling (GMDH) along with linking between Aspen HYSYS and MATLAB software, optimized with Genetic algorithm (GA), is herein proposed to obtain efficient polynomial correlation to estimate optimal consumed power for these two cascade cycles. Results show that the proposed multi-hybrid model is superior to non-linear programming techniques for obtaining the optimum consumed power of cascade refrigeration cycles and finding the values of optimizing variables.