Hybrid artificial neural network genetic algorithm technique for modeling chemical oxygen demand removal in anoxic/oxic process

Abstract

In this paper, a hybrid artificial neural network (ANN) - genetic algorithm (GA) numerical technique was successfully developed to deal with complicated problems that cannot be solved by conventional solutions. ANNs and Gas were used to model and simulate the process of removing chemical oxygen demand (COD) in an anoxic/oxic system. The minimization of the error function with respect to the network parameters (weights and biases) has been considered as training of the network. Real-coded genetic algorithm was used to train the network in an unsupervised manner. Meanwhile the important process parameters, such as the influent COD (COD_in), reflux ratio (R _r ), carbon-nitrogen ratio (C/N) and the effluent COD (COD_out) were considered. The result shows that compared with the performance of ANN model, the performance of the GA-ANN (genetic algorithm - artificial neural network) network was found to be more impressive. Using ANN, the mean absolute percentage error (MAPE), mean squared error (MSE) and correlation coefficient (R) were 9.33×10⁻⁴, 2.82 and 0.98596, respectively; while for the GA-ANN, they were converged to be 4.18×10⁻⁴, 1.12 and 0.99476, respectively.

Keywords:

• Anoxic/oxic process
• artificial neural networks
• genetic algorithms
• COD removal

Acknowledgments

This work was supported by Guangdong Provincial Department of Science (No.2008 2008A080800003) and State Key Laboratory of Pulp and Paper Engineering in China (201003). The authors are thankful to the anonymous reviewers for their insightful comments and suggestions.