The sound quality (SQ) and sound perception assessments of electric vehicles (EVs) clearly differ from those of conventional internal combustion engine vehicles (ICEVs). Therefore, it is essential to describe and evaluate the SQ of EVs. To evaluate the SQ in EVs, it is necessary to organize evaluators for conducting subjective jury tests, which are time-consuming and labor-intensive. In addition, the evaluation results are subject to the evaluators themselves and other external interferences. With the advancement of machine learning and artificial neural networks (ANNs), this problem can be well solved. This paper outlines a model for SQ estimation in EVs based on a genetic algorithm-optimized back propagation artificial neural network (GA-BP ANN). Moreover, the correlation between the physical-psychoacoustical parameters and the subjective SQ estimations obtained from the jury tests was investigated in this study. It was found that the GA-BP ANN SQ model has many advantages in comparison with the multiple linear regression (MLR) model in terms of precision and generalization. In addition, this method is ready to be applied for rapidly evaluating the SQ in EVs without jury tests, and it can also be of high significance in dealing with the acoustical designs and improvements of EVs in the future.
State-based load profile generation for modeling energetic flexibility
