Annoyance Evaluation And Prediction Of Noise Emitted By Urban Electrical Substation

The continuous popularization of electric vehicles promotes the construction and expansion of urban substations,and the noise pollution caused by urban substations is aggravated.Residents around the substations frequently complain about the noise disturbance,and the problem of noise disturbance in the substations is becoming increasingly prominent.In order to accurately evaluate the degree of annoyance caused by substation noise to the surrounding residents,this paper carried out a study on the evaluation of noise annoyance of an urban substation based on sound quality evaluation method.Combined with the characteristics of steady signal and prominent low-frequency harmonics of substation noise,the subjective and objective indicators suitable for noise evaluation of the urban substation are determined,two evaluation methods about urban substation noise annoyance are proposed,and corresponding noise annoyance prediction models is established respectively.The verification results show that both methods can correctly predict the subjective annoyance of urban substation noise and achieve high prediction accuracy.First of all,the subjective and objective evaluation were carried out for the collected urban substation noise samples in the laboratory environment,the data validity of the subjective evaluation results was analyzed,and the correlation among all the evaluation subjects and among the objective parameters was discussed respectively.The results show that after eliminating the invalid data of two evaluation subjects,there was a good consistency among the remaining evaluation subjects,and the Kendall correlation coefficients were all above 0.68.There had strong correlation among some objective parameters,and the Pearson correlation coefficient was above 0.7.Then,the main objective parameters that affect the noise annoyance are discussed and retained based on principal component analysis(PCA)method,the correlation between these parameters and the subjective evaluation results is analyzed,the objective parameters that are highly correlated with the subjective evaluation results are used as independent variables,the subjective evaluation result is used as dependent variable,and a multiple linear stepwise regression model for predicting the noise annoyance of urban substations is established.The results show that the A sound level of 200 Hz,400 Hz,500 Hz,600 Hz,900 Hz and the total A sound level are the main objective parameters affecting annoyance,they have a high correlation with the subjective evaluation result of the noise annoyance,and the Pearson correlation coefficient are all over 0.65.The 600 Hz A sound level and total A sound level are the key objective parameters that constitutes the regression model of annoyance.The regression model can correctly predict the subjective annoyance of urban substation noise.Secondly,the convolutional neural network(CNN)method based on transfer learning is studied,a convolutional neural network model for the prediction of substation noise annoyance is established,and the influence of the value of Mini Batch Size on the prediction accuracy of the model is explored.The results show that when Mini Batch Size is set to 4,8,16 and 32 respectively,all the data sets have convergent after 90 iterations.The RMSE of all validation sets is no more than 0.355,and the loss of all validation sets is no more than 0.067.With the increase of Mini Batch Size,the RMSE,loss and MAE of the verification set are also gradually increase,and the number of each iteration gradually decrease,the training duration gradually decrease.In this test,the value of Mini Batch Size of 4 is appropriate.