Spectral Feature Band Selection and Interval Partial Least Squares
Modeling of Short-Range Nitrification Process
SONG Yu1, 2, LI Wei-hua1, 2*, XUE Tong-zhan1, 2, YU Li1, 2, SHEN Hui-yan1, 2
1. School of Environment and Energy Engineering, Anhui Jianzhu University, Hefei 230601, China
2. Anhui Provincial Key Laboratory of Environmental Pollution Control and Resource Reuse, Anhui Jianzhu University, Hefei 230601, China
Abstract：Sequential Batch Reactor (SBR) is one of the most widely used active sludge treatment devices.In this experiment, the nitrate and nitrite nitrogen content changes during the startup process of the short-term nitrification reaction system were studied using an SBR reactor to treat artificially simulated high ammonia wastewater.A model was established based on the data collected usingultraviolet spectroscopy,aiming to rapidly predict the nitrate nitrogen and nitrite nitrogen content in the effluent of the SBR reactor. Using laboratory-prepared solutions with different concentrations of nitrate and nitrite nitrogen, a calibration model for standard mixtures was constructedusing the interval partial least squares(iPLS) for three different band selection methods. The research results show that the models built exhibit good correlations between the measured and predicted values for nitrate nitrogen and nitrite nitrogen in the mixed solution.To determine the reactor effluent parameters, models for ultraviolet spectroscopy and the nitrate and nitrite nitrogen content were constructed using partial least squares algorithms for three different band selections.The model results were evaluated using the calibration set correlation coefficient, the root mean square error of cross-validation (RMSECV), the correlation coefficient of the prediction set, and the root mean square error of prediction (RMSEP) evaluation metrics.Among the three models, the model built using the synergy interval partial least squares (siPLS) method, which divided the full spectrum into 24 and 19 intervals and established models for the combined sub-intervals[2 4] and [3 8], exhibited the best prediction and fitting results. Its calibration model hadr=0.939 3 and RMSECV=1.650 4, and the prediction model hadr=0.950 7 and RMSEP=0.442 1. This model showed an overall good prediction performance for nitrate nitrogen and nitrite nitrogen, indicating that establishing interval partial least squares models using ultraviolet spectroscopy can rapidly predict nitrate nitrogen and nitrite nitrogen content in the effluent of the short-term nitrification reactor.
[1] WANG Shu-ying, LI Lun, LI Ling-yun, et al(王淑莹，李 论，李凌云，等). Journal of Beijing University of Technology(北京工业大学学报), 2011, 37(7): 1067.
[2] XU Ting, WANG Li, WU Jun(徐 婷，王 丽，吴 军). Chinese Journal of Environmental Engineering(环境工程学报), 2016, 10(6): 2840.
[3] Yaqoob M, Folgado Biot B, Nabi A, et al. Luminescence, 2012, 27(5): 419.
[4] SUN Hao, GUO Hui, ZHAO Hui, et al(孙 浩，郭 慧，赵 辉，等). Chinese Journal of Environmental Engineering(环境工程学报), 2016, 10(4): 2122.
[5] Shariati-Rad M, Irandoust M, Haghighi M. International Journal of Environmental Science and Technology, 2015, 12: 3837.
[6] Nørgaard L, Saudland A, Wagner J, et al. Applied Spectroscopy, 2000, 54(3): 413.
[7] Nørgaard L, Hahn M T, Knudsen L B, et al. International Dairy Journal, 2005, 15(12): 1261.
[8] YAO Liang, SHEN Hui-yan, LI Wei-hua, et al(姚 亮，申慧彦，李卫华，等). Environmental Science and Technology(环境科学与技术), 2018, 41(7): 145.
[9] Guo Y, Liu C, Ye R, et al. Applied Sciences, 2020, 10(19): 6874.
[10] Skou P B, Berg T A, Aunsbjerg S D, et al. Applied Spectroscopy, 2017, 71(3): 410.
[11] Li J, Tong Y, Guan L, et al. Optik, 2018, 174: 591.
[12] Cai J, Chen Q, Wan X, et al. Food Chemistry, 2011, 126(3): 1354.
[13] HUANG Jian, HUANG Shan, ZHANG Hua, et al(黄 健，黄 珊，张 华，等). China Environmental Science(中国环境科学), 2015, 35(7): 2014.
[14] Chen Q, Ding J, Cai J, et al. Journal of Food Science, 2012, 77(2): C222.
