Analysis of automatic monitoring data of total phosphorus in drinking water source in east Taihu Lake based on improved extreme learning machine algorithm
CUI Jiayu1,2,, GUO Rong1,,, ZHANG Yue1,2, XU Liang1, ZHONG Sheng1, DONG Yuanyuan1, LI Xinyu3 1.Jiangsu Environmental Monitoring Center, Nanjing 210019, China 2.Jiangsu Suli Environmental Science and Technology Co., Ltd, Nanjing 210036, China 3.School of the Environment, Nanjing University, Nanjing 210093, China
Abstract:A comparative analysis was performed on the concentration of total phosphorus measured in the Jinshugang and Yuyangshan drinking water source areas in East Taihu Lake by automatic water monitoring and the laboratory method. Three key factors were found to influence the automatic monitoring of total phosphorus according to a correlation analysis, i.e. water turbidity, algal density and chlorophyll-a. The three parameters were introduced into the improved extreme learning machine model (IELM) for correction of automatic water monitoring data of total phosphorus. Compared with laboratory method, the total phosphorus concentrations measured by automatic water monitoring were shown to have relatively large errors with absolute error ranging between 0.05 mg·L?1 and 0.112 mg·L?1 and the mean absolute error being 0.017 mg·L?1. The training error was 0.0000735 and test error was 0.000103 after the training and testing of the IELM model. The measurement results by the application of IELM showed better performance. With 30% of relative error rate as judgment criteria, the eligible rates increased from 52.9% to 92.0%, the absolute error decreased by 0.026 mg/L on average, and the relative error rate dropped by 45%. The results in this study show the promise of application of IELM model in correction of automatic monitoring of total phosphorus in the field. Key words:total phosphorus/ automatic monitoring/ improved extreme learning machine algorithm/ hidden node/ corrected data.
图1太湖饮用水水源地监测点位分布图 Figure1.Monitoring sites in drinking water source in east Taihu Lake
FENG G, HUANG G B, LIN Q, et al. Error minimized extreme learning machine with growth of hidden nodes and incremental learning[J]. IEEE Transactions on Neural Networks, 2009, 20(8): 1352-1357. doi: 10.1109/TNN.2009.2024147
DENG W Y, BAI Z, HUANG G B, et al. A fast SVD-hidden-nodes based extreme learning machine for large-scale data analytics[J]. Neural Networks, 2016, 77(1): 14.
Analysis of automatic monitoring data of total phosphorus in drinking water source in east Taihu Lake based on improved extreme learning machine algorithm
1.Jiangsu Environmental Monitoring Center, Nanjing 210019, China 2.Jiangsu Suli Environmental Science and Technology Co., Ltd, Nanjing 210036, China 3.School of the Environment, Nanjing University, Nanjing 210093, China Received Date: 2020-09-26 Accepted Date: 2021-04-26 Available Online: 2021-06-25 Keywords:total phosphorus/ automatic monitoring/ improved extreme learning machine algorithm/ hidden node/ corrected data Abstract:A comparative analysis was performed on the concentration of total phosphorus measured in the Jinshugang and Yuyangshan drinking water source areas in East Taihu Lake by automatic water monitoring and the laboratory method. Three key factors were found to influence the automatic monitoring of total phosphorus according to a correlation analysis, i.e. water turbidity, algal density and chlorophyll-a. The three parameters were introduced into the improved extreme learning machine model (IELM) for correction of automatic water monitoring data of total phosphorus. Compared with laboratory method, the total phosphorus concentrations measured by automatic water monitoring were shown to have relatively large errors with absolute error ranging between 0.05 mg·L?1 and 0.112 mg·L?1 and the mean absolute error being 0.017 mg·L?1. The training error was 0.0000735 and test error was 0.000103 after the training and testing of the IELM model. The measurement results by the application of IELM showed better performance. With 30% of relative error rate as judgment criteria, the eligible rates increased from 52.9% to 92.0%, the absolute error decreased by 0.026 mg/L on average, and the relative error rate dropped by 45%. The results in this study show the promise of application of IELM model in correction of automatic monitoring of total phosphorus in the field.