Performance Prediction and Optimization of an Organic Rankine Cycle Using Back Propagation Neural Network for Diesel Engine Waste Heat Recovery

Abstract

This paper presents a methodology to predict and optimize performance of an organic Rankine cycle (ORC) using a back propagation neural network (BPNN) for diesel engine waste heat recovery. A test bench of an ORC with a diesel engine is established to collect experimental data. The collected data are used to train and test a BPNN model for performance prediction and optimization. After evaluating different hidden layers, a BPNN model of the ORC system is determined with the consideration of mean squared error (MSE) and correlation coefficient. The effects of key operating parameters on the power output of the ORC system and exhaust temperature at the outlet of the evaporator are evaluated using the proposed model and further discussed. Finally, a multi-objective optimization of the ORC system is conducted for maximizing power output and minimizing exhaust temperature at the outlet of the evaporator based on the proposed BPNN model. The results show that the proposed BPNN model has a high prediction accuracy and the maximum relative error of the power output is less than 5%. It also shows that when the operations are optimized based on the proposed model, the power output of the ORC system can be higher than the experimental results.