Compared to the internal combustion engine, the electric powertrain is composed of the motor, the motor controller and the transmission. Its radiated noise has the characteristics of high-frequency dispersion and outstanding tonality, which makes the sound quality as a key evaluation criteria. Take a low-speed and high-load electric powertrain from an Electric Bus (E-bus) as an example. According to the structural composition and sound characteristics of the electric powertrain, the multiple dimensional objective evaluation parameters are selected, which are more suitable for evaluating the sound quality for E-bus electric powertrain, including the psychoacoustics and physical acoustics characteristics, such as A-weighted sound pressure level, loudness, tonality, etc. Based on noise structure and operating condition distribution of the noise samples, a Rating-comparison Method (RCM) is proposed to carry out jury test. And the correctness of subjective evaluation results are verified. The Kernel Principal Component Analysis (KPCA) is used to analyze and extract the key objective evaluation parameters suitable for evaluating its sound quality characteristics, and its target kernel function and the optimal kernel parameters are determined, which simplify the objective evaluation dimension. The Lasso Regression Method (LRM) is used to solve the multi-collinearity characteristics problem of the multi-dimensional objective evaluation parameters. Finally, the multilinear-quantitative relationship between the objective evaluation parameters and the subjective evaluation results is constructed. The sound quality prediction model for low-speed and high-load electric powertrain has been obtained, which provides theoretical basis for evaluating the sound quality for different electric powertrain.