Experimental research and artificial neural network prediction of free piston expander-linear generator

Abstract

For the purpose of better matching the performance of the organic Rankine cycle (ORC) system concerning the vehicle engine waste heat recovery, this paper studies the output performance of free piston expander-linear generator (FPE-LG). A test bench of FPE-LG is established for small scale ORC system, and timing and displacement control strategy is proposed. Furthermore, the impact of the intake pressure and the torque on motion characteristics and output performance of FPE-LG are analyzed. According to evaluating different learning rates, number of hidden artificial neural networks and training functions, a prediction model of FPE-LG based on artificial neural network is established. Genetic algorithm is used to optimize the key operating parameters, to maximize the power output of FPE-LG. In consideration of the mean square error and determination coefficient, the artificial neural network model is verified and tested by experimental data. Finally, combining genetic algorithm with artificial neural network model, the maximum power output of FPE-LG is optimized and its performance is predicted. The results show that the maximum value of electric current, voltage and power output are 2.8 A, 14.75 V and 28.5 W, respectively. Based on artificial neural network, this method can provide useful guidance for performance prediction and coordinated optimization, with advantages of minimum deviation and high precision.