Prediction of the Anaerobic Systems Based on Neural Network with Multipopulation Parallel Genetic Algorithm

The performance changes database of the Up-flow anaerobic sludge blanket reactor shocked by the test loadings was obtained according to the measure per hour. Artificial neural network (ANN) was applied to predict parameters change of the anaerobic system. The multipopulation parallel genetic algorithm (MPGA) based on real coding was engaged to optimize weights of ANN. The correlation coefficients of observed data and predicted value were 0.916, 0.853, and 0.892 for volatile fatty acid, volume gas production and CH₄ content, respectively. The results showed that ANN with MPGA can be a valuable tool for predicting the performance change of anaerobic system, and has greatly adaptability to the variations of environmental conditions. It can be also further extended to the other wastewater treatment system.