Parameter analysis and multi-objective optimization of organic Rankine cycle coupled vapor compression cycle using PSO-BPNN model

Abstract

The accuracy of predictive models is crucial for evaluating and optimizing the performance of organic Rankine cycle combined with vapor compression refrigeration (ORC-VCR) cycles. This paper establishes a PSO-BPNN prediction model by optimizing the weight and threshold of the back-propagation neural network (BPNN) using particle swarm optimization (PSO). A small ORC-VCR device with a cooling capacity of 3 kW is constructed, and 142 steady-state experimental data are obtained for training the developed model. Then, the influence of operating parameters on the system performance is investigated. In addition, operating parameters are optimized to maximize the cooling capacity and coefficient of performance (COP). The average absolute error of PSO-BPNN model for cooling capacity and COP is about 2.2 %, which is 34 % and 46 % lower than the BPNN model. Compared with the flow rate of cooling water, its temperature has a greater impact on the system performance. The vapor compression cycle has an optimal flow rate to obtain the maximum cooling capacity and COP of the combined system. Through multi-objective optimization, the optimal cooling capacity and COP of the ORC-VCR system are 4.41 kW and 0.32, which are 32 % and 14 % higher than the maximum values observed in the experimental data.