PREDICTION OF BOD INDEX IN WASTEWATER TREATMENT PLANT USING GENETIC ALGORITHMS AND NEURAL NETWORKS

Abstract

The biochemical oxygen demand (BOD is defined as the rate at which microorganisms use oxygen in water or wastewater during the fixation of biodegradable organic matter under aerobic conditions. The use of BOD as a necessary parameter for the effective control and monitoring of the wastewater treatment plants remains somewhat restricted due to the long time it takes by traditional ways, which hinders its use in real time. In this paper, a hybrid model of genetic algorithms (GA) and artificial neural networks (ANN) is used for the prediction of biochemical oxygen demand. The data used in this research was collected over ten years through the daily records of the Homs waste water treatment plant. the model was built based on the approval of each of the values of (COD, SS, Q) as inputs to predict the value of the BOD, and the performance of the model was evaluated by adopting an inverse validation error for selecting the best network structure in addition to other differential criteria. The optimal structure of the neural network was determined after a number of attempts and errors, and the results showed a high efficiency of the proposed hybrid model of (genetic algorithms and neural networks) by predicting the value of inflow BOD .             As a result of this research, a neural network structure was selected to predict the value of the BOD indicator which is (2-54-1) using the Hyperbolic Tangent function in the hidden layer and the logistic function in the output layer, the Levenberg-Marquardt was used as a training algorithm for training, The value of the performance function was 0.125, and the average error value of the three groups was 1.83, while the mean maximum error of the three groups was 5.86, the value of the correlation coefficient was 0.99.