Neural network prediction of fluidized bed bioreactor performance for sulfide oxidation

Prediction and quantifying parameter importance in simultaneous anaerobic sulfide and nitrate removal process using artificial neural network

The present investigation deals with the prediction of the performance of simultaneous anaerobic sulfide and nitrate removal in an upflow anaerobic sludge bed (UASB) reactor through an artificial neural network (ANN). Influent sulfide concentration, influent nitrate concentration, S/N mole ratio, pH, and hydraulic retention time (HRT) for 144 days' steady-state condition were the inputs of the model; whereas output parameters were sulfide removal percentage, nitrate removal percentage, sulfate production percentage, and nitrogen production percentage. The prediction performance was evaluated by calculating root mean square error (RMSE), mean absolute error (MAE), mean absolute relative error (MARE), and determination coefficient (R (2)) values. Generally, the ANN model exhibited good prediction of the simultaneous sulfide and nitrate removal process. The effect of five input parameters to the performance of the reactor was quantified and compared using the connection weights method, Garson's algorithm method, and partial derivatives (PaD) method. The results showed that HRT markedly affects the performance of the reactor.