Modelagem e simulação da autodepuração em rios

Abstract

The present work describes the development of a model in Artificial Neural Networks (ANN) with the objective of simulating the capacity of a river to recover the amount of dissolved oxygen (OD), after receiving a polluting load. The experiment was carried out in a river located in the municipality of Medianeira in the State of Paraná, which receives the treated effluent from a slaughterhouse agroindustry. The model was developed based on data on the water quality of the river and the effluent that is incorporated into the watercourse over the studied interval. For training and validation of the model, data from previous studies, collected in 2014, were used. The input variables that were used in the network are the parameters of water and effluent quality and the output was the OD. To this end, five RNA models were developed, which underwent modifications in the input variables, but maintained the same output and the results simulated by the RNA were compared with the results collected in loco. In addition, the results obtained were compared with the results of a study carried out in 2014, with the objective of verifying the evolution of the technologies used in ANN. Among the main results of the present study, the processing speed stands out in relation to the previous study. Among the developed models, the model named RNA3 was the one that presented results more similar to the results collected in the field, demonstrating greater capacity for simulating the self-purification of the river.