Modelling of dissolved oxygen in the Danube River using artificial neural networks and Monte Carlo Simulation uncertainty analysis
Само за регистроване кориснике
2014
Чланак у часопису (Објављена верзија)
Метаподаци
Приказ свих података о документуАпстракт
This paper describes the training, validation, testing and uncertainty analysis of general regression neural network (GRNN) models for the forecasting of dissolved oxygen (DO) in the Danube River. The main objectives of this work were to determine the optimum data normalization and input selection techniques, the determination of the relative importance of uncertainty in different input variables, as well as the uncertainty analysis of model results using the Monte Carlo Simulation (MCS) technique. Min-max, median, z-score, sigmoid and tanh were validated as normalization techniques, whilst the variance inflation factor, correlation analysis and genetic algorithm were tested as input selection techniques. As inputs, the GRNN models used 19 water quality variables, measured in the river water each month at 17 different sites over a period of 9 years. The best results were obtained using min-max normalized data and the input selection based on the correlation between DO and dependent var...iables, which provided the most accurate GRNN model, and in combination the smallest number of inputs: Temperature, pH, HCO3-, SO42-, NO3-N, Hardness, Na, Cl-, Conductivity and Alkalinity. The results show that the correlation coefficient between measured and predicted DO values is 0.85. The inputs with the greatest effect on the GRNN model (arranged in descending order) were T, pH, HCO3-, SO42- and NO3-N. Of all inputs, variability of temperature had the greatest influence on the variability of DO content in river body, with the DO decreasing at a rate similar to the theoretical DO decreasing rate relating to temperature. The uncertainty analysis of the model results demonstrate that the GRNN can effectively forecast the DO content, since the distribution of model results are very similar to the corresponding distribution of real data.
Кључне речи:
DO / GRNN / MCS / VIF / Genetic algorithm / Correlation analysisИзвор:
Journal of Hydrology, 2014, 519, 1895-1907Издавач:
- Elsevier Science Bv, Amsterdam
Финансирање / пројекти:
- Развој и примена метода и материјала за мониторинг нових загађујућих и токсичних органских материја и тешких метала (RS-MESTD-Basic Research (BR or ON)-172007)
DOI: 10.1016/j.jhydrol.2014.10.009
ISSN: 0022-1694
WoS: 000347018100054
Scopus: 2-s2.0-84908377307
Институција/група
Tehnološko-metalurški fakultetTY - JOUR AU - Antanasijević, Davor AU - Pocajt, Viktor AU - Perić-Grujić, Aleksandra AU - Ristić, Mirjana PY - 2014 UR - http://TechnoRep.tmf.bg.ac.rs/handle/123456789/2703 AB - This paper describes the training, validation, testing and uncertainty analysis of general regression neural network (GRNN) models for the forecasting of dissolved oxygen (DO) in the Danube River. The main objectives of this work were to determine the optimum data normalization and input selection techniques, the determination of the relative importance of uncertainty in different input variables, as well as the uncertainty analysis of model results using the Monte Carlo Simulation (MCS) technique. Min-max, median, z-score, sigmoid and tanh were validated as normalization techniques, whilst the variance inflation factor, correlation analysis and genetic algorithm were tested as input selection techniques. As inputs, the GRNN models used 19 water quality variables, measured in the river water each month at 17 different sites over a period of 9 years. The best results were obtained using min-max normalized data and the input selection based on the correlation between DO and dependent variables, which provided the most accurate GRNN model, and in combination the smallest number of inputs: Temperature, pH, HCO3-, SO42-, NO3-N, Hardness, Na, Cl-, Conductivity and Alkalinity. The results show that the correlation coefficient between measured and predicted DO values is 0.85. The inputs with the greatest effect on the GRNN model (arranged in descending order) were T, pH, HCO3-, SO42- and NO3-N. Of all inputs, variability of temperature had the greatest influence on the variability of DO content in river body, with the DO decreasing at a rate similar to the theoretical DO decreasing rate relating to temperature. The uncertainty analysis of the model results demonstrate that the GRNN can effectively forecast the DO content, since the distribution of model results are very similar to the corresponding distribution of real data. PB - Elsevier Science Bv, Amsterdam T2 - Journal of Hydrology T1 - Modelling of dissolved oxygen in the Danube River using artificial neural networks and Monte Carlo Simulation uncertainty analysis EP - 1907 SP - 1895 VL - 519 DO - 10.1016/j.jhydrol.2014.10.009 ER -
@article{ author = "Antanasijević, Davor and Pocajt, Viktor and Perić-Grujić, Aleksandra and Ristić, Mirjana", year = "2014", abstract = "This paper describes the training, validation, testing and uncertainty analysis of general regression neural network (GRNN) models for the forecasting of dissolved oxygen (DO) in the Danube River. The main objectives of this work were to determine the optimum data normalization and input selection techniques, the determination of the relative importance of uncertainty in different input variables, as well as the uncertainty analysis of model results using the Monte Carlo Simulation (MCS) technique. Min-max, median, z-score, sigmoid and tanh were validated as normalization techniques, whilst the variance inflation factor, correlation analysis and genetic algorithm were tested as input selection techniques. As inputs, the GRNN models used 19 water quality variables, measured in the river water each month at 17 different sites over a period of 9 years. The best results were obtained using min-max normalized data and the input selection based on the correlation between DO and dependent variables, which provided the most accurate GRNN model, and in combination the smallest number of inputs: Temperature, pH, HCO3-, SO42-, NO3-N, Hardness, Na, Cl-, Conductivity and Alkalinity. The results show that the correlation coefficient between measured and predicted DO values is 0.85. The inputs with the greatest effect on the GRNN model (arranged in descending order) were T, pH, HCO3-, SO42- and NO3-N. Of all inputs, variability of temperature had the greatest influence on the variability of DO content in river body, with the DO decreasing at a rate similar to the theoretical DO decreasing rate relating to temperature. The uncertainty analysis of the model results demonstrate that the GRNN can effectively forecast the DO content, since the distribution of model results are very similar to the corresponding distribution of real data.", publisher = "Elsevier Science Bv, Amsterdam", journal = "Journal of Hydrology", title = "Modelling of dissolved oxygen in the Danube River using artificial neural networks and Monte Carlo Simulation uncertainty analysis", pages = "1907-1895", volume = "519", doi = "10.1016/j.jhydrol.2014.10.009" }
Antanasijević, D., Pocajt, V., Perić-Grujić, A.,& Ristić, M.. (2014). Modelling of dissolved oxygen in the Danube River using artificial neural networks and Monte Carlo Simulation uncertainty analysis. in Journal of Hydrology Elsevier Science Bv, Amsterdam., 519, 1895-1907. https://doi.org/10.1016/j.jhydrol.2014.10.009
Antanasijević D, Pocajt V, Perić-Grujić A, Ristić M. Modelling of dissolved oxygen in the Danube River using artificial neural networks and Monte Carlo Simulation uncertainty analysis. in Journal of Hydrology. 2014;519:1895-1907. doi:10.1016/j.jhydrol.2014.10.009 .
Antanasijević, Davor, Pocajt, Viktor, Perić-Grujić, Aleksandra, Ristić, Mirjana, "Modelling of dissolved oxygen in the Danube River using artificial neural networks and Monte Carlo Simulation uncertainty analysis" in Journal of Hydrology, 519 (2014):1895-1907, https://doi.org/10.1016/j.jhydrol.2014.10.009 . .