The process of an anaerobic reactor in predicting the COD Level was modeled by the application of Neural Networks. The effluent COD is the most common factor used in all types of Wastewater Treatment Systems. This research is about the prediction of the concentration of |COD by using the ANN, a simple feed forward, Back Propagation (BP) neural network. The three layered feed forward ANN has been trained using four back propagation algorithms. The efficiency and accuracy were measured based on MSE and Regression Coefficient(R) to evaluate their performance. The model trained with Levenberg -Marquardt algorithm provided best results with MSE = 0.533 and R=0.991. The model performed up to the expectation and the result of the prediction is appropriate.