抄録
In the previous paper, R. filter GMDH correlation (coefficient was used as a criterion function for selection.) was presented to predict effluent quality from the sewage treatment plant in terms of influent quality and running condition.
In this paper, it was recognized that good effluent quality was attained under the condition in 80 SVI and 1700 mg/l MLSS as the result of investigating the influence of the input parameter to the effluent quality.
Frequency distribution curves of the input and output data were shown to investigate utility of prediction medel. For the similar influent quality, effluent quality observed under the conditions suggested by the prediction model is better than that under the other conditions.
Prediction model by R. filter GMDH with data in a limited period could not be well applied to period without revision. However, provided that the prediction model is remodeled by the method of least squares, it could be applicable to extended objective period for the prediction TRSe (transparency of effluent), SSe (suspended solid of effluent).
