Forecast of selected quality indicators of wastewater flowing to the treatment plant using selected black box methods

In this study some black-box models are employed to model various indicators of the wastewater quality (characterized by biochemical oxygen demand, chemical oxygen demand, the content of total and ammoniacal nitrogen, suspended solids and total phosphorus) at the inflow to a sewage treatment plant using as the predictors the sewage inflow and temperature as well as the indicators measured in the previous time steps. Considering all tested methods one can see that in most cases the best predictions have been obtained by the method of multivariate adaptive regression splines. Moreover the models based only on inflow and temperature inputs have got a noticeably lower errors than those once conditioned by indicators.