Affiliation:
1. Nanjing Vocational University of Industry Technology
2. Automotive College, Sanmenxia Polytechnic
Abstract
Abstract
In recent years, air pollution has become an increasingly important issue in the sustainable development of cities. Monitoring air pollutants is of great significance for government departments to effectively control air pollution. The development of micro air quality monitors provides the possibility for grid monitoring and real-time monitoring of air pollutants. However, affected by many factors, the measurement accuracy of the micro air quality monitors need to be improved. In this paper, a combined prediction model of Multiple Linear Regression, Boosted Regression Tree and AutoRegressive Integrated Moving Average model (MLR-BRT-ARIMA) is proposed to calibrate the measurement data of the micro air quality monitor. First, the multiple linear regression model is used to find the linear relationship between the concentration of various pollutants and the measurement data of the micro air quality monitor and obtain the predicted value of the concentration of various pollutants. Second, take the measurement data of the micro air quality monitor and the prediction value of the multiple regression model as the input, and use the boosted regression tree to find the nonlinear relationship between the concentrations of various pollutants and the input variables. Finally, the autoregressive integrated moving average model is used to extract the information hidden in the residual sequence, and finally the establishment of the MLR-BRT-ARIMA model is completed. This combined model combines the advantages of multiple linear regression and boosted regression trees. It can not only give the quantitative relationship between the explained variables and their influencing factors, but also the prediction accuracy is higher than the multiple linear regression and boosted regression tree models alone. Using the ARIMA model to correct the residuals can further improve the prediction accuracy of the model. Root mean square error, mean absolute error and relative mean absolute percent error are used to compare the calibration effect of MLR-BRT-ARIMA model and other commonly used models such as multilayer perceptron neural network, support vector regression machine and nonlinear autoregressive models with exogenous input. The results show that no matter what kind of pollutant, the MLR-BRT-ARIMA combined model proposed in this paper has the best performance in the three indicators. Using this model to calibrate the measurement value of the micro air quality monitor can improve the accuracy by 82.4%~95.4%.
Publisher
Research Square Platform LLC