Abstract:

Considering the complicated process of biochemical sewage treatment,difficulty in precisely forecastingeffluent quality and relatively serious prediction error,a prediction model for the effluent quality in wastewater treat-ment is proposed on the basis of relevance vector machine.In this method,an attribute reduction is,first and fore-most,performed for input data by using fuzzy monotonic increasing dependence algorithm,and the final inputattributes are determined in combination with experience.Then,an effluent quality prediction model is establishedwith the help of relevance vector machine and the model parameters are subsequently optimized.Experimentalresults indicate that the proposed prediction model well meets the requirements of effluent quality forecasting due toits high prediction accuracy and strong generalization ability.

Key words: wastewater treatment, relevance vector machine, fuzzy monotonic increasing, forecasting