Premium
Development and application of reduced‐order neural network model based on proper orthogonal decomposition for BOD 5 monitoring in river systems: Uncertainty analysis
Author(s) -
Noori Roohollah,
Ashrafi Khosro,
Karbassi Abdolreza,
Ardestani Mojtaba,
Mehrdadi Naser
Publication year - 2013
Publication title -
environmental progress and sustainable energy
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.495
H-Index - 66
eISSN - 1944-7450
pISSN - 1944-7442
DOI - 10.1002/ep.11610
Subject(s) - artificial neural network , computer science , monte carlo method , decomposition , uncertainty analysis , data mining , machine learning , artificial intelligence , statistics , mathematics , simulation , chemistry , organic chemistry
Uncertainty of the reduced‐order neural network (RONN) models is one of the main challenges for developing a proper framework based on their results in water quality management. Hence, the main objective of the research is to determine and compare the uncertainty of both RONN and neural network (NN) models for online estimation of the 5‐day biochemical oxygen demand (BOD 5 ). To achieve this goal, the Monte–Carlo method is used for determining uncertainty analysis of these models. Results indicated that bracketed predictions by 95% confidence bound in the testing steps for selected RONN and NN models are 90.5% and 71.4%, respectively. However, d‐factor of the selected RONN model is better than NN model. Furthermore, obtained results based on comparison between RONN and NN models for online estimation of BOD 5 revealed that uncertainty of RONN models were less than NN model. Generally, results of the present research are another confirmation on the authors' previous study. © 2013 American Institute of Chemical Engineers Environ Prog, 32: 344‐349, 2013